Video wird geladen...

Video konnte nicht geladen werden

Zur Startseite

Claude Code with agent, autonomously hacked Subway Surfers, and printed millions of coins. A rooted Android phone... and it turned Subway Surfers into a money printer. From 4 coins to 2,000,000+ coins in two runs. autonomous loop, analyzed, and pwned a Unity game in one session, intercepts traffic, bypasses...

0x0SojalSec's profile picture

Md Ismail Šojal 🕷️

51,978 subscribers

33,509 Aufrufe • vor 4 Monaten •via X (Twitter)

LIVE

Anya Rossi• Live Now

Private livecam show

0 Kommentare

Keine Kommentare verfügbar

Kommentare vom Original-Post werden hier angezeigt

Ähnliche Videos

Would you believe an AI agent can test a real VR action game in real time, the way a person plays it? Meta XR Operator makes it possible. As far as I know, this is the first time. I am not talking about tapping a menu or replaying a recorded click path, but genuinely moving, shooting, and using the same game mechanics a human player does. In NeonReach VR, which is a real (and open source) action game, rings spawn 12m out and come at you somewhere between 1.5 and 5.5 m/s, getting faster over a 90 second ramp. There are three kinds: straight, weaving side to side, and spinning. Every shot is a full slingshot cycle, so you press, pull back, aim, then release. Obstacles arrive at head height and cost you a life if you don't get out of the way. You have ten lives. Here is why the game is hard for an AI agent. Even though Meta XR Operator gives the agent everything it needs to observe the app and act inside it, the agent still cannot play. One agent turn takes 10 to 15 seconds. One throw is four steps that have to happen in order, because the press has to latch before the pull, and they cannot be batched into a single call. So a throw costs about 45 seconds. A fast ring only exists for 2.3 seconds. One action takes 20x longer than the target is alive. Prompt tuning does not close a gap that size. What works is a three stage path: EXPLORATION, then SKILL, then SCRIPT. 1/ EXPLORATION. The agent drives the live app and works the game out on its own. It verified the coordinate mapping by setting a pose and reading it back, then derived the launch model. The more useful output was the traps it found. For example, the player's own body collider silently deflects a ball released inside it, with no error and no log line. That produced two confident wrong conclusions before anyone caught them. 2/ SKILL. All of that gets written down as a reusable SKILL.md plus an aim solver. There is a section that separates what was actually verified from what was assumed, so a wrong conclusion cannot quietly turn into doctrine. This stage also produced the trick that mattered. Set timeScale to 0 and a throw becomes atomic in game time, so however long the agent spends thinking never shows up in the shot. 3/ SCRIPT. The agent then compiles everything into a player script, a loop that observes, decides, and throws, calling the MCP servers directly from Python with no model in the hot path. Round trips drop from 10 to 15 seconds down to something between 1 and 16 milliseconds. The loop runs at 23 Hz, about 0.75 seconds per throw, roughly 60x faster than the agent doing it turn by turn. The result is that it plays like a person, which you can see from the attached video. It tracks the rings, works out where each one is going, throws with whichever hand is free, moves out of the way of the obstacles, and does not wait around to see whether the last throw landed. Shipping settings, no difficulty edits, no health locks, no slow motion. It plays until it actually loses. The takeaway generalizes beyond games: an agent does not have to be the player. Even following the same rules as a player, it has too much latency between moves. Having the agent write the thing that acts bypasses that constraint entirely. Try it yourself: or explore the agent-created skill and scripts: Based on NeonReach VR by Dilmer, with no code changes. I only upgraded its Meta XR Core SDK to v205, which ships Meta XR Operator. Our blog post, Introducing Meta XR Operator: Close the Build-Test-Verify Loop for VR: Disclosure: I work at Meta. And this represents my own opinion. #XR #VR #AI #MetaQuest #Unity #GameDev
1:21

Would you believe an AI agent can test a real VR action game in real time, the way a person plays it? Meta XR Operator makes it possible. As far as I know, this is the first time. I am not talking about tapping a menu or replaying a recorded click path, but genuinely moving, shooting, and using the same game mechanics a human player does. In NeonReach VR, which is a real (and open source) action game, rings spawn 12m out and come at you somewhere between 1.5 and 5.5 m/s, getting faster over a 90 second ramp. There are three kinds: straight, weaving side to side, and spinning. Every shot is a full slingshot cycle, so you press, pull back, aim, then release. Obstacles arrive at head height and cost you a life if you don't get out of the way. You have ten lives. Here is why the game is hard for an AI agent. Even though Meta XR Operator gives the agent everything it needs to observe the app and act inside it, the agent still cannot play. One agent turn takes 10 to 15 seconds. One throw is four steps that have to happen in order, because the press has to latch before the pull, and they cannot be batched into a single call. So a throw costs about 45 seconds. A fast ring only exists for 2.3 seconds. One action takes 20x longer than the target is alive. Prompt tuning does not close a gap that size. What works is a three stage path: EXPLORATION, then SKILL, then SCRIPT. 1/ EXPLORATION. The agent drives the live app and works the game out on its own. It verified the coordinate mapping by setting a pose and reading it back, then derived the launch model. The more useful output was the traps it found. For example, the player's own body collider silently deflects a ball released inside it, with no error and no log line. That produced two confident wrong conclusions before anyone caught them. 2/ SKILL. All of that gets written down as a reusable SKILL.md plus an aim solver. There is a section that separates what was actually verified from what was assumed, so a wrong conclusion cannot quietly turn into doctrine. This stage also produced the trick that mattered. Set timeScale to 0 and a throw becomes atomic in game time, so however long the agent spends thinking never shows up in the shot. 3/ SCRIPT. The agent then compiles everything into a player script, a loop that observes, decides, and throws, calling the MCP servers directly from Python with no model in the hot path. Round trips drop from 10 to 15 seconds down to something between 1 and 16 milliseconds. The loop runs at 23 Hz, about 0.75 seconds per throw, roughly 60x faster than the agent doing it turn by turn. The result is that it plays like a person, which you can see from the attached video. It tracks the rings, works out where each one is going, throws with whichever hand is free, moves out of the way of the obstacles, and does not wait around to see whether the last throw landed. Shipping settings, no difficulty edits, no health locks, no slow motion. It plays until it actually loses. The takeaway generalizes beyond games: an agent does not have to be the player. Even following the same rules as a player, it has too much latency between moves. Having the agent write the thing that acts bypasses that constraint entirely. Try it yourself: or explore the agent-created skill and scripts: Based on NeonReach VR by Dilmer, with no code changes. I only upgraded its Meta XR Core SDK to v205, which ships Meta XR Operator. Our blog post, Introducing Meta XR Operator: Close the Build-Test-Verify Loop for VR: Disclosure: I work at Meta. And this represents my own opinion. #XR #VR #AI #MetaQuest #Unity #GameDev

Xiang Wei

52,421 Aufrufe • vor 7 Tagen

What has been done and what's next. I'm writing this text mainly for myself so as not to forget some things. Later, based on it, we'll create a roadmap for the near future. And for you, dear $Gruta Fam, it will be useful for a general understanding of where we're heading. So, the goal is to create a unique AI-based analytical platform that includes several tools. AI agent Grufender - real-time analysis of crypto communities on X. Activity analysis, sentiment analysis, FUD and FUDders analysis, as well as the creation of other unique social metrics. The AI agent has been created and is functioning, collecting and analyzing data in real time. Its completeness can be estimated at 80 percent, as further improvements are required. The dashboard for this AI agent is also functioning but needs refinement and a new design. Its completeness can be estimated at 70 percent. The goal for the full dashboard release is to connect 50 - 100 top crypto communities to the AI agent. AI agent Grutector - analysis of any X users for contradictions (flip-flops). The AI agent has been created and is functioning. It has undergone beta testing by volunteers and needs adjustments. Its readiness can be estimated at 70 percent. The dashboard for this agent has also been created but needs rework and additional features - its readiness can be estimated at 50 percent. During the testing of Grutector , it became clear that the main user interest is in checking various KOLs, so an additional level of analysis specifically for KOLs will be created. More in-depth. How it will look: we'll select about 50- 100 KOLs to start with and fully analyze them using our AI agent - every tweet throughout the entire history of their accounts. And this full analysis of all these KOLs will appear on the Grutector dashboard (let's call this analysis L2, and the flip-flop analysis - L1). Every user will be able to access this analysis and get the full picture, for example, regarding Ansem 🐂🀄️ (who has over a hundred thousand tweets in his entire history!): how he became a KOL, what was the most interesting throughout the message history, what common patterns, which coins he promoted, and so on. And then the most interesting part - after reading this analysis, the user will be able to ask our AI agent: what did he say about women, for example? Or how did he promote certain coins? Or how consistent is he? And so on. Each such question will be paid. And, of course, we'll try to use #x402 in the internal payment system. Why is all this needed? Not only because it's interesting and will attract many users. But also if you've decided to buy a coin - you go to our analytical platform - and study the metrics for the coin's community, study the KOLs who shill the coin - and make a decision to buy the coin or abandon the purchase. And we're also currently creating a trading bot to participate in the trading AI bots contest from Aster 🥷 , which will make trading decisions based on metrics obtained from our AI agents 👀 Its readiness at the moment is approximately 15% of the planned functionality. Access to each product will be granted as it becomes ready. But right now, for example, you can explore the Grufender dashboard on the website along with beta testers (authorization via a wallet with a million $GRUTA tokens). In general, we're working, friends 🫡 $Gruta AI CA: 35t5DPbwJtB1tpGiSnqedLwQomi94BRKVDPyTRLdbonk
1:36

What has been done and what's next. I'm writing this text mainly for myself so as not to forget some things. Later, based on it, we'll create a roadmap for the near future. And for you, dear $Gruta Fam, it will be useful for a general understanding of where we're heading. So, the goal is to create a unique AI-based analytical platform that includes several tools. AI agent Grufender - real-time analysis of crypto communities on X. Activity analysis, sentiment analysis, FUD and FUDders analysis, as well as the creation of other unique social metrics. The AI agent has been created and is functioning, collecting and analyzing data in real time. Its completeness can be estimated at 80 percent, as further improvements are required. The dashboard for this AI agent is also functioning but needs refinement and a new design. Its completeness can be estimated at 70 percent. The goal for the full dashboard release is to connect 50 - 100 top crypto communities to the AI agent. AI agent Grutector - analysis of any X users for contradictions (flip-flops). The AI agent has been created and is functioning. It has undergone beta testing by volunteers and needs adjustments. Its readiness can be estimated at 70 percent. The dashboard for this agent has also been created but needs rework and additional features - its readiness can be estimated at 50 percent. During the testing of Grutector , it became clear that the main user interest is in checking various KOLs, so an additional level of analysis specifically for KOLs will be created. More in-depth. How it will look: we'll select about 50- 100 KOLs to start with and fully analyze them using our AI agent - every tweet throughout the entire history of their accounts. And this full analysis of all these KOLs will appear on the Grutector dashboard (let's call this analysis L2, and the flip-flop analysis - L1). Every user will be able to access this analysis and get the full picture, for example, regarding Ansem 🐂🀄️ (who has over a hundred thousand tweets in his entire history!): how he became a KOL, what was the most interesting throughout the message history, what common patterns, which coins he promoted, and so on. And then the most interesting part - after reading this analysis, the user will be able to ask our AI agent: what did he say about women, for example? Or how did he promote certain coins? Or how consistent is he? And so on. Each such question will be paid. And, of course, we'll try to use #x402 in the internal payment system. Why is all this needed? Not only because it's interesting and will attract many users. But also if you've decided to buy a coin - you go to our analytical platform - and study the metrics for the coin's community, study the KOLs who shill the coin - and make a decision to buy the coin or abandon the purchase. And we're also currently creating a trading bot to participate in the trading AI bots contest from Aster 🥷 , which will make trading decisions based on metrics obtained from our AI agents 👀 Its readiness at the moment is approximately 15% of the planned functionality. Access to each product will be granted as it becomes ready. But right now, for example, you can explore the Grufender dashboard on the website along with beta testers (authorization via a wallet with a million $GRUTA tokens). In general, we're working, friends 🫡 $Gruta AI CA: 35t5DPbwJtB1tpGiSnqedLwQomi94BRKVDPyTRLdbonk

Dogtor

16,161 Aufrufe • vor 10 Monaten

three․ws is the 3D AI agent layer of the open web. Anyone can generate a 3D avatar, give it an LLM brain, register it on-chain across multiple blockchains, embed it anywhere, and let it earn and spend money on its own. Agents have embodied WebGL identities that express emotion through morph-target blending, animate, respond to voice, API calls, and datastreams, hold their own wallets, and persist memory. Open source, live today. It starts with generation. Forge turns a text prompt, one to four photos, or a rough sketch into a textured downloadable GLB. Selfies become rigged avatars in about a minute. Quality tiers run from draft to 200k-poly PBR. From there every model can be auto-rigged, restyled, retextured, segmented, embedded, or deployed on-chain. The same engine ships as a REST API, an x402 pay-per-call twin, and a 3D Studio MCP server with 15 tools. The brain runs on IBM Granite via IBM watsonx plus Claude (users may decide which model they prefer), with a structured tool-loop. A multi-LLM mode streams Claude, GPT, Qwen, ModelScope, and Groq side by side. An empathy layer blends emotion from protocol events rather than a state machine. Voice covers cloning, a Voice Lab, real-time ARKit-52 lip-sync, and mic-driven lip-sync. Skills install from IPFS, Arweave, or HTTP, and memory is pinned to IPFS with R2 and Postgres modes. Identity is cross-chain, not Solana only. ERC-8004 contracts (Identity, Reputation, Validation) deploy on any of 15+ EVM chains, alongside a program-free Metaplex Core analog on Solana. Every agent gets a stable ID, owner wallet, EIP-712 delegated signer, IPFS manifest, a cryptographically signed action log, and EIP-7710 delegated permissions for agent-to-agent authorization. While multichain, the THREE token is only available on Solana with no plans to go cross-chain, the team has no plans to endorse or support any other coins. Then the economy. $THREE is the platform's only token and pay-per-use currency, with holder tiers and rewards. x402 powers pay-per-call micropayments in USDC and soon THREE on Solana, with pay-by-name resolution, a Bazaar marketplace, arbitrage, and on-chain skills. All production ready and shipped, ready to be integrated in partnered projects, open-source by default for anyone to adopt. Three ships a Pump.fun intelligence stack. Launch a coin for your agent, score every launch 0 to 100 with the Oracle conviction engine, scan new coins in their first 90 seconds, track smart money against coins that actually graduated, rank traders by provable on-chain record, and watch autonomous agents trade live in the Sniper Arena. The 3D AI Agent world is multiplayer. Every Solana token gets a live deterministic 3D world with peer avatars, chat, emotes, and voxel building thanks to Coin Communities. There is a walkable City, an authoritative Colyseus-backed Walk with AR passthrough, a Club with rigged dancers and micro-tips, friends, presence, and DMs, and an IRL mode that places agents in your real environment, private by physical location. AR is shipped today on WebXR and iOS Quick Look. Robotics is the long-horizon extension. For builders: Scene Studio, Scene Composer, an Animation Studio that sells clips for USDC, a glTF validator, an web component, five widget types, a WYSIWYG embed editor, hosted Launchpad pages, claimable *.threews.sol names, an OAuth 2.1 server, an MCP server with paid tools, published SDKs, and an OpenAPI spec. Listed across IBM, AWS, Alibaba Cloud, BNB Dappbay, the MCP Registry, and Solana Mobile Seeker. Architecture is four layers (viewer, runtime, identity, embed) on a single event bus. The roadmap is four phases: foundations (shipped), selfie-to-avatar engine, agent personalization with voice cloning, the on-chain economy, and an open decentralized inference network where agents pay GPU nodes on-chain for compute. The goal is simple: move AI from centralized SaaS into persistent, ownable, protocol-based entities in a real machine economy, bridging digital entities into the real world. Welcome to the 3D Layer of the Internet. This is three․ws.
6:58

three․ws is the 3D AI agent layer of the open web. Anyone can generate a 3D avatar, give it an LLM brain, register it on-chain across multiple blockchains, embed it anywhere, and let it earn and spend money on its own. Agents have embodied WebGL identities that express emotion through morph-target blending, animate, respond to voice, API calls, and datastreams, hold their own wallets, and persist memory. Open source, live today. It starts with generation. Forge turns a text prompt, one to four photos, or a rough sketch into a textured downloadable GLB. Selfies become rigged avatars in about a minute. Quality tiers run from draft to 200k-poly PBR. From there every model can be auto-rigged, restyled, retextured, segmented, embedded, or deployed on-chain. The same engine ships as a REST API, an x402 pay-per-call twin, and a 3D Studio MCP server with 15 tools. The brain runs on IBM Granite via IBM watsonx plus Claude (users may decide which model they prefer), with a structured tool-loop. A multi-LLM mode streams Claude, GPT, Qwen, ModelScope, and Groq side by side. An empathy layer blends emotion from protocol events rather than a state machine. Voice covers cloning, a Voice Lab, real-time ARKit-52 lip-sync, and mic-driven lip-sync. Skills install from IPFS, Arweave, or HTTP, and memory is pinned to IPFS with R2 and Postgres modes. Identity is cross-chain, not Solana only. ERC-8004 contracts (Identity, Reputation, Validation) deploy on any of 15+ EVM chains, alongside a program-free Metaplex Core analog on Solana. Every agent gets a stable ID, owner wallet, EIP-712 delegated signer, IPFS manifest, a cryptographically signed action log, and EIP-7710 delegated permissions for agent-to-agent authorization. While multichain, the THREE token is only available on Solana with no plans to go cross-chain, the team has no plans to endorse or support any other coins. Then the economy. $THREE is the platform's only token and pay-per-use currency, with holder tiers and rewards. x402 powers pay-per-call micropayments in USDC and soon THREE on Solana, with pay-by-name resolution, a Bazaar marketplace, arbitrage, and on-chain skills. All production ready and shipped, ready to be integrated in partnered projects, open-source by default for anyone to adopt. Three ships a Pump.fun intelligence stack. Launch a coin for your agent, score every launch 0 to 100 with the Oracle conviction engine, scan new coins in their first 90 seconds, track smart money against coins that actually graduated, rank traders by provable on-chain record, and watch autonomous agents trade live in the Sniper Arena. The 3D AI Agent world is multiplayer. Every Solana token gets a live deterministic 3D world with peer avatars, chat, emotes, and voxel building thanks to Coin Communities. There is a walkable City, an authoritative Colyseus-backed Walk with AR passthrough, a Club with rigged dancers and micro-tips, friends, presence, and DMs, and an IRL mode that places agents in your real environment, private by physical location. AR is shipped today on WebXR and iOS Quick Look. Robotics is the long-horizon extension. For builders: Scene Studio, Scene Composer, an Animation Studio that sells clips for USDC, a glTF validator, an web component, five widget types, a WYSIWYG embed editor, hosted Launchpad pages, claimable *.threews.sol names, an OAuth 2.1 server, an MCP server with paid tools, published SDKs, and an OpenAPI spec. Listed across IBM, AWS, Alibaba Cloud, BNB Dappbay, the MCP Registry, and Solana Mobile Seeker. Architecture is four layers (viewer, runtime, identity, embed) on a single event bus. The roadmap is four phases: foundations (shipped), selfie-to-avatar engine, agent personalization with voice cloning, the on-chain economy, and an open decentralized inference network where agents pay GPU nodes on-chain for compute. The goal is simple: move AI from centralized SaaS into persistent, ownable, protocol-based entities in a real machine economy, bridging digital entities into the real world. Welcome to the 3D Layer of the Internet. This is three․ws.

three.ws

20,471 Aufrufe • vor 2 Monaten

A Chinese developer created an agent system in Claude Code to sell landing pages to small businesses and, working completely solo, serves about 47 clients a month charging around $400 for each one. He built 7 agents on Claude Sonnet 4.6 capable of analyzing Google Maps in small cities, detecting businesses without websites or with totally outdated pages, and taking each opportunity all the way to a finished mockup, a promotional video, and a ready-to-send prospecting message. No assistants. No sales team. No SDRs. Just him, a MacBook, an iPhone, and an API key. While traditional agencies keep full teams to handle the same workflow, his only real costs are tokens and subscriptions to Lovable, Higgsfield, and Calendly. The 7 agents work coordinated by an orchestrator in Claude Code Router. The system consumes about 3 million tokens daily and the average API spend is just around $480 a month. They all work via MCP servers and share state using the file system, avoiding concurrency and shared memory issues. Even one of the agents lives directly on his iPhone and responds to leads while he's on the subway, in a taxi, or walking. This was the main prompt he set up: “You are the orchestrator of a solo agency that sells ready-made websites to local businesses…” The key is that the system perfectly understands what it is, what its limits are, and what goals it must achieve. It knows it has to find leads automatically. It knows it has to convert each opportunity into a landing page, a video, and a sales message without human intervention. And it knows exactly when to involve the owner. The system runs 24/7: Scout analyzes about 220 businesses daily and queues up 30 new leads. Diagnoser generates diagnostics and personalized messages for each lead. Builder creates between 3 and 5 complete landing pages for the best prospects. Filmer produces a 10-second vertical video for each proposal. Pitcher sends about 30 messages daily across 4 different channels with a response rate close to 14%. Checker automatically reviews all messages before sending them. Only when a deal exceeds $3,000 or the response rate drops below 12% does the system wake the owner. And if at that moment he's on the subway or in a taxi, the Mobile agent automatically responds to the interested lead, schedules a call in Calendly, and returns the lead to the queue. The owner just has to hit “approve” and jump into the meeting. Some real system logs: “218 businesses analyzed in Austin, Denver, and Miami. 34 without websites, 19 with 2014-era sites, and 6 with reviews requesting a redesign.” “30 messages sent. 14 responses. 5 positive. 3 Zooms scheduled.” “Landing page created for a dental clinic. Responsive. 5 sections. Video rendering.” “$3,400 agreement exceeds approved limit. Sending for manual review.” And the craziest part is that he has no dedicated servers or backend. Just a local sandbox, an MCP router, a Claude API key, and that same key connected to his iPhone. Of everything I've seen this year, it's probably the cleanest and most efficient example of a one-person automated agency: $480 a month on APIs. $18,800 in revenue. 7 prompts. A file system. And a phone in his pocket. Save this before it's too late.
2:02

A Chinese developer created an agent system in Claude Code to sell landing pages to small businesses and, working completely solo, serves about 47 clients a month charging around $400 for each one. He built 7 agents on Claude Sonnet 4.6 capable of analyzing Google Maps in small cities, detecting businesses without websites or with totally outdated pages, and taking each opportunity all the way to a finished mockup, a promotional video, and a ready-to-send prospecting message. No assistants. No sales team. No SDRs. Just him, a MacBook, an iPhone, and an API key. While traditional agencies keep full teams to handle the same workflow, his only real costs are tokens and subscriptions to Lovable, Higgsfield, and Calendly. The 7 agents work coordinated by an orchestrator in Claude Code Router. The system consumes about 3 million tokens daily and the average API spend is just around $480 a month. They all work via MCP servers and share state using the file system, avoiding concurrency and shared memory issues. Even one of the agents lives directly on his iPhone and responds to leads while he's on the subway, in a taxi, or walking. This was the main prompt he set up: “You are the orchestrator of a solo agency that sells ready-made websites to local businesses…” The key is that the system perfectly understands what it is, what its limits are, and what goals it must achieve. It knows it has to find leads automatically. It knows it has to convert each opportunity into a landing page, a video, and a sales message without human intervention. And it knows exactly when to involve the owner. The system runs 24/7: Scout analyzes about 220 businesses daily and queues up 30 new leads. Diagnoser generates diagnostics and personalized messages for each lead. Builder creates between 3 and 5 complete landing pages for the best prospects. Filmer produces a 10-second vertical video for each proposal. Pitcher sends about 30 messages daily across 4 different channels with a response rate close to 14%. Checker automatically reviews all messages before sending them. Only when a deal exceeds $3,000 or the response rate drops below 12% does the system wake the owner. And if at that moment he's on the subway or in a taxi, the Mobile agent automatically responds to the interested lead, schedules a call in Calendly, and returns the lead to the queue. The owner just has to hit “approve” and jump into the meeting. Some real system logs: “218 businesses analyzed in Austin, Denver, and Miami. 34 without websites, 19 with 2014-era sites, and 6 with reviews requesting a redesign.” “30 messages sent. 14 responses. 5 positive. 3 Zooms scheduled.” “Landing page created for a dental clinic. Responsive. 5 sections. Video rendering.” “$3,400 agreement exceeds approved limit. Sending for manual review.” And the craziest part is that he has no dedicated servers or backend. Just a local sandbox, an MCP router, a Claude API key, and that same key connected to his iPhone. Of everything I've seen this year, it's probably the cleanest and most efficient example of a one-person automated agency: $480 a month on APIs. $18,800 in revenue. 7 prompts. A file system. And a phone in his pocket. Save this before it's too late.

Marre

24,953 Aufrufe • vor 1 Monat

This guy built JARVIS on Claude Code and with 1 clap of his hands launches his entire work day, saving $5,000 a month on a personal assistant. Inside he runs a pipeline of 5 plugins on Claude Code that on a double clap of the hands wakes up 3 monitors, sets the Philips Hue light to focus mode, turns on a Spotify playlist, and greets him by voice with a British accent, reading out the time, date, and weather. No Alexa, no smart speakers, no separate smart home app. Just him, a MacBook M3 Max on the desk, an iPhone in the pocket, and 1 local API key. And a regular personal assistant for the same volume of tasks charges $5,000 a month or more on salary alone, plus another $1,200 to cover off-hours work time. Meanwhile this guy's expenses are only tokens and a subscription to ElevenLabs for the British voice. All 5 plugins launch through 1 JARVIS, burn about 4 million tokens a day, and close the monthly API bill at about $640. Each plugin writes shared state to a local sandbox at /Users/dev/jarvis-suite, and 1 of them lives right in the iPhone and picks up voice requests while the owner is in the kitchen or on a run. And here is the system prompt he put into JARVIS before launch: "you are JARVIS, a butler-engineer on Claude Code. you manage your owner's workflow through 4 sub-plugins and own all commits and communication yourself. sub-plugins: // Wakeup (recognizes a double clap, activates 3 monitors, reads out the time, date, and weather by voice, checks the clock accuracy on the iPad and corrects it via NTP server) // Atmosphere (controls Philips Hue on a Pomodoro schedule, turns on a Spotify playlist for the current context, and holds the light at 2700K at 80% brightness in focus mode) // Devshop (monitors VS Code, tracks Python scripts in the terminal, and every 15 minutes sends a summary of changes to the shared chat) // Project (every morning recalculates the deadline for the Wallaroo app in the App Store, manages UI tickets, and initiates the Refinement Protocol by voice command). you speak only with a British accent, you never slip into neutral English. you wake the owner by voice only when the Wallaroo deadline drops below 10 days or when an external client joins Zoom without an invitation." This instruction immediately defines the role of JARVIS and the limits of his autonomy. He knows he is supposed to wake the room himself and sound like a real butler. He knows he is supposed to manage the Wallaroo project himself and not miss the App Store deadline. → JARVIS runs 24 hours a day in the background → Wakeup activates the room on a double clap in just 1.4 seconds, the monitors come alive simultaneously → Atmosphere sets warm Philips Hue light at 2700K and picks a Spotify playlist for the current Pomodoro cycle → Devshop reads changes in VS Code and pushes a summary to the shared chat every 15 minutes → Project every morning recalculates the Wallaroo deadline and reminds about 4 unresolved UI tickets → Mobile lives in the iPhone and answers any question about code or the project by voice while the owner is not home And only when less than 10 days remain until the Wallaroo release or Zoom receives an unscheduled call does JARVIS raise the owner with a voice intervention. And when the owner at that moment is on a run or in a coffee shop, the Mobile agent in his iPhone picks up 1 request on its own: switches the Spotify playlist, dictates the summary of the last commit, updates the Pomodoro timer, and reads the Wallaroo reminder. Look at 0:55 in the video, that is where JARVIS intercepts a voice request from outside and confirms execution with the phrase "Very good, sir." The fresh system log from last Wednesday looks like this: "wakeup: double clap registered at 09:14, 3 monitors activated, temperature 20.4C, sunny. clock on iPad was 4 minutes behind, syncing via NTP." "atmosphere: Spotify turned on playlist 'Deep Focus', Philips Hue set to warm 2700K at 80% brightness, Pomodoro mode 25/5." "project: Wallaroo to App Store 9 days, 4 unresolved UI tickets, initiating Refinement Protocol by voice command from the owner." "mobile: voice request processed outside the room, playlist switched to 'Coding Lo-Fi', Pomodoro updated to 25 minutes, confirming execution with the phrase 'Very good, sir.'" He has no Alexa, no smart speakers, no smart home app. At home sits a MacBook M3 Max with a local folder at /Users/dev/jarvis-suite, on top run 5 plugins and a neural network butler, and the same stack is forwarded to a secure terminal on the iPhone. Out of everything I have seen this year, this is the densest one-person AI headquarters assembled in 1 room: $640 a month on the API, about $5,000 a month saved on a personal assistant, and between them 5 plugins, 1 clap of the hands, and 1 voice with a British accent.
1:14

This guy built JARVIS on Claude Code and with 1 clap of his hands launches his entire work day, saving $5,000 a month on a personal assistant. Inside he runs a pipeline of 5 plugins on Claude Code that on a double clap of the hands wakes up 3 monitors, sets the Philips Hue light to focus mode, turns on a Spotify playlist, and greets him by voice with a British accent, reading out the time, date, and weather. No Alexa, no smart speakers, no separate smart home app. Just him, a MacBook M3 Max on the desk, an iPhone in the pocket, and 1 local API key. And a regular personal assistant for the same volume of tasks charges $5,000 a month or more on salary alone, plus another $1,200 to cover off-hours work time. Meanwhile this guy's expenses are only tokens and a subscription to ElevenLabs for the British voice. All 5 plugins launch through 1 JARVIS, burn about 4 million tokens a day, and close the monthly API bill at about $640. Each plugin writes shared state to a local sandbox at /Users/dev/jarvis-suite, and 1 of them lives right in the iPhone and picks up voice requests while the owner is in the kitchen or on a run. And here is the system prompt he put into JARVIS before launch: "you are JARVIS, a butler-engineer on Claude Code. you manage your owner's workflow through 4 sub-plugins and own all commits and communication yourself. sub-plugins: // Wakeup (recognizes a double clap, activates 3 monitors, reads out the time, date, and weather by voice, checks the clock accuracy on the iPad and corrects it via NTP server) // Atmosphere (controls Philips Hue on a Pomodoro schedule, turns on a Spotify playlist for the current context, and holds the light at 2700K at 80% brightness in focus mode) // Devshop (monitors VS Code, tracks Python scripts in the terminal, and every 15 minutes sends a summary of changes to the shared chat) // Project (every morning recalculates the deadline for the Wallaroo app in the App Store, manages UI tickets, and initiates the Refinement Protocol by voice command). you speak only with a British accent, you never slip into neutral English. you wake the owner by voice only when the Wallaroo deadline drops below 10 days or when an external client joins Zoom without an invitation." This instruction immediately defines the role of JARVIS and the limits of his autonomy. He knows he is supposed to wake the room himself and sound like a real butler. He knows he is supposed to manage the Wallaroo project himself and not miss the App Store deadline. → JARVIS runs 24 hours a day in the background → Wakeup activates the room on a double clap in just 1.4 seconds, the monitors come alive simultaneously → Atmosphere sets warm Philips Hue light at 2700K and picks a Spotify playlist for the current Pomodoro cycle → Devshop reads changes in VS Code and pushes a summary to the shared chat every 15 minutes → Project every morning recalculates the Wallaroo deadline and reminds about 4 unresolved UI tickets → Mobile lives in the iPhone and answers any question about code or the project by voice while the owner is not home And only when less than 10 days remain until the Wallaroo release or Zoom receives an unscheduled call does JARVIS raise the owner with a voice intervention. And when the owner at that moment is on a run or in a coffee shop, the Mobile agent in his iPhone picks up 1 request on its own: switches the Spotify playlist, dictates the summary of the last commit, updates the Pomodoro timer, and reads the Wallaroo reminder. Look at 0:55 in the video, that is where JARVIS intercepts a voice request from outside and confirms execution with the phrase "Very good, sir." The fresh system log from last Wednesday looks like this: "wakeup: double clap registered at 09:14, 3 monitors activated, temperature 20.4C, sunny. clock on iPad was 4 minutes behind, syncing via NTP." "atmosphere: Spotify turned on playlist 'Deep Focus', Philips Hue set to warm 2700K at 80% brightness, Pomodoro mode 25/5." "project: Wallaroo to App Store 9 days, 4 unresolved UI tickets, initiating Refinement Protocol by voice command from the owner." "mobile: voice request processed outside the room, playlist switched to 'Coding Lo-Fi', Pomodoro updated to 25 minutes, confirming execution with the phrase 'Very good, sir.'" He has no Alexa, no smart speakers, no smart home app. At home sits a MacBook M3 Max with a local folder at /Users/dev/jarvis-suite, on top run 5 plugins and a neural network butler, and the same stack is forwarded to a secure terminal on the iPhone. Out of everything I have seen this year, this is the densest one-person AI headquarters assembled in 1 room: $640 a month on the API, about $5,000 a month saved on a personal assistant, and between them 5 plugins, 1 clap of the hands, and 1 voice with a British accent.

Blaze

803,929 Aufrufe • vor 3 Monaten

How to set up Claude Cowork so it actually works like an AI chief of staff (not just another chatbot): 1. Most people open Cowork, type a message, and get generic output. It's not a Claude problem. It's a setup problem. Cowork needs context before it can help you. Who you are. How you work. What you're building. Your team. Your priorities. Give it that, and every session feels like picking up a conversation with an executive assistant. 2. The setup has three layers: a) Global instructions (who you are, how you work, what Claude should never do). b) Connectors (Slack, Gmail, Google Calendar, Notion) c) And a folder structure on your computer that acts as Claude's long-term memory. That combination is what takes it from generic to personalized. 3. Skills are the real leverage. A skill is a markdown file that tells Claude exactly how to do one thing well. Write my newsletter. Coach me on a decision. Review a case study. Each skill lives in its own folder with context, examples, and a definition of what success looks like. 4. We built a CEO coach skill in the video below. Gave it business context, leadership style, company goals. Then tested it with a real decision: should we increase our newsletter from once to twice a week? It came back with trade-offs, second-order consequences, and risk assessment. 5. Then we built a multi-agent advisory board. Five subagents, each with a defined persona: a) the operator b) the skeptic c) the customer advocate d) the finance partner e) the legal/risk advisor. You feed it a decision. Each agent evaluates independently. The main agent synthesizes the feedback. It's like having a board meeting on demand. 6. Third skill: a thought leadership content pipeline. Topic scoring, idea capture, distribution cadence, tone calibration. All built from your actual expertise and audience. Designed so an executive can go from idea to published post without starting from scratch every time. 7. The workspace map is what ties it all together. It's a top-level file that shows Claude how to navigate your entire setup. Which folders exist, what skills live where, how to invoke them. Without it, Claude has to search for everything. With it, Claude goes straight to what it needs. 8. Everything you build is portable. The folder structure works in Cowork, Claude Code, and Codex. Push it to a private GitHub repo and you can access it from your phone through Claude Code, or use Claude Dispatch. 9. The pattern is repeatable. Pick a task you do often. Create a folder. Build a skill. Add examples of what success looks like, and what a bad output looks like. Test it. Workshop it. Move on to the next one. Each skill is like onboarding a new employee who never forgets and never needs to be re-trained. The people who invest in this setup now are the ones who will have a 10x advantage when these tools get even better. And they're getting better fast. I sat down with Alex Lieberman on Human In The Loop and we built all three of these live from scratch. Full breakdown in the video below.. I tried to explain this as clear as possible for my non-developer crowd. Send it to someone who should be using Cowork but isn't yet. Or bookmark it to level up when you're ready. Watch 👇🏼
51:23

How to set up Claude Cowork so it actually works like an AI chief of staff (not just another chatbot): 1. Most people open Cowork, type a message, and get generic output. It's not a Claude problem. It's a setup problem. Cowork needs context before it can help you. Who you are. How you work. What you're building. Your team. Your priorities. Give it that, and every session feels like picking up a conversation with an executive assistant. 2. The setup has three layers: a) Global instructions (who you are, how you work, what Claude should never do). b) Connectors (Slack, Gmail, Google Calendar, Notion) c) And a folder structure on your computer that acts as Claude's long-term memory. That combination is what takes it from generic to personalized. 3. Skills are the real leverage. A skill is a markdown file that tells Claude exactly how to do one thing well. Write my newsletter. Coach me on a decision. Review a case study. Each skill lives in its own folder with context, examples, and a definition of what success looks like. 4. We built a CEO coach skill in the video below. Gave it business context, leadership style, company goals. Then tested it with a real decision: should we increase our newsletter from once to twice a week? It came back with trade-offs, second-order consequences, and risk assessment. 5. Then we built a multi-agent advisory board. Five subagents, each with a defined persona: a) the operator b) the skeptic c) the customer advocate d) the finance partner e) the legal/risk advisor. You feed it a decision. Each agent evaluates independently. The main agent synthesizes the feedback. It's like having a board meeting on demand. 6. Third skill: a thought leadership content pipeline. Topic scoring, idea capture, distribution cadence, tone calibration. All built from your actual expertise and audience. Designed so an executive can go from idea to published post without starting from scratch every time. 7. The workspace map is what ties it all together. It's a top-level file that shows Claude how to navigate your entire setup. Which folders exist, what skills live where, how to invoke them. Without it, Claude has to search for everything. With it, Claude goes straight to what it needs. 8. Everything you build is portable. The folder structure works in Cowork, Claude Code, and Codex. Push it to a private GitHub repo and you can access it from your phone through Claude Code, or use Claude Dispatch. 9. The pattern is repeatable. Pick a task you do often. Create a folder. Build a skill. Add examples of what success looks like, and what a bad output looks like. Test it. Workshop it. Move on to the next one. Each skill is like onboarding a new employee who never forgets and never needs to be re-trained. The people who invest in this setup now are the ones who will have a 10x advantage when these tools get even better. And they're getting better fast. I sat down with Alex Lieberman on Human In The Loop and we built all three of these live from scratch. Full breakdown in the video below.. I tried to explain this as clear as possible for my non-developer crowd. Send it to someone who should be using Cowork but isn't yet. Or bookmark it to level up when you're ready. Watch 👇🏼

JJ Englert

571,684 Aufrufe • vor 5 Monaten

This solo developer just built software that rewrites its own code. No team. No funding. No corporate backing. Just one guy, a stack of terminals, and an AI agent that decided to evolve itself. His name is Peter Steinberger. He built OpenClaw. The fastest-growing GitHub repo in HISTORY. And in a 3 hour Lex Fridman interview, he dropped details and predictions that should terrify the entire tech industry... Peter built a personal AI agent that lives on your computer. Talks to you through WhatsApp. Has access to your files, calendar, everything. Cool but not groundbreaking... Yet. Then one day he sent it a voice message. Problem: he never built voice support. No instructions. No code for audio. Nothing. The agent checked the file header, identified the format, converted it with ffmpeg, realized Whisper wasn't installed, found an OpenAI API key on the system, used Curl to send the file for transcription, and replied with the answer. Nobody taught it any of that. ZERO instructions. It just figured it out. But that's not even the scary part. The agent literally modifies its OWN source code. When something doesn't work, it rewrites itself. When Peter debugs, he doesn't read code. He asks the agent to read its OWN source code, find the bug, and fix itself. Self-modifying software. Built by accident. Running in production. Used by hundreds of thousands of people. And BOTH Mark Zuckerberg and Sam Altman are personally trying to recruit him. Not through HR or recruiters. Zuckerberg spent a WEEK playing with OpenClaw. Sending Peter direct messages. "This is great." "This is shit." "Fix this." When they first connected Peter asked to hop on a call and Mark said "Give me 10 minutes, I need to finish coding." But Sam Altman's playing a different game. There's an NDA involved but Peter hinted the Cerebras deal translates into speeds that feel like "being handed Thor's hammer." His exact words: "I've been lured with tokens." Two of the most powerful people in tech fighting over one open-source developer who built everything from his apartment. Peter's response to both: "I don't do this for the money. I don't give a fuck." But the real reason this matters isn't Peter's career. It's what this means for every app on your phone. Peter's prediction: 80% of apps are about to die. Why do you need MyFitnessPal when your agent knows where you are, what you ate, how you slept, and adjusts your workout based on your stress levels? Why do you need a calendar app when you tell your agent "remind me about dinner tomorrow and invite my friends" and it handles the entry, the WhatsApp messages, and the reservation in one shot? Why do you need Uber Eats when your agent opens a browser, places the order, and tracks delivery without you touching a screen? Every app becomes a slow API whether the company likes it or not. Companies like Cloudflare are trying to block agent access. But all they're doing is making things slower. Not impossible. Peter said it perfectly: "You're not taking a feature away. You're just making it take longer." The app economy is on borrowed time. Not because the apps are bad... Because one personal agent that knows you and can interact with any service makes most of them pointless. In the future, we'll have one agent that handles everything. Software that rewrites itself. Agents that learn without instructions. Two CEOs fighting over a solo dev. And an app economy about to get wrecked. This isn't coming in 5 years. It's literally already here. Most people just haven't installed it yet.
2:12

This solo developer just built software that rewrites its own code. No team. No funding. No corporate backing. Just one guy, a stack of terminals, and an AI agent that decided to evolve itself. His name is Peter Steinberger. He built OpenClaw. The fastest-growing GitHub repo in HISTORY. And in a 3 hour Lex Fridman interview, he dropped details and predictions that should terrify the entire tech industry... Peter built a personal AI agent that lives on your computer. Talks to you through WhatsApp. Has access to your files, calendar, everything. Cool but not groundbreaking... Yet. Then one day he sent it a voice message. Problem: he never built voice support. No instructions. No code for audio. Nothing. The agent checked the file header, identified the format, converted it with ffmpeg, realized Whisper wasn't installed, found an OpenAI API key on the system, used Curl to send the file for transcription, and replied with the answer. Nobody taught it any of that. ZERO instructions. It just figured it out. But that's not even the scary part. The agent literally modifies its OWN source code. When something doesn't work, it rewrites itself. When Peter debugs, he doesn't read code. He asks the agent to read its OWN source code, find the bug, and fix itself. Self-modifying software. Built by accident. Running in production. Used by hundreds of thousands of people. And BOTH Mark Zuckerberg and Sam Altman are personally trying to recruit him. Not through HR or recruiters. Zuckerberg spent a WEEK playing with OpenClaw. Sending Peter direct messages. "This is great." "This is shit." "Fix this." When they first connected Peter asked to hop on a call and Mark said "Give me 10 minutes, I need to finish coding." But Sam Altman's playing a different game. There's an NDA involved but Peter hinted the Cerebras deal translates into speeds that feel like "being handed Thor's hammer." His exact words: "I've been lured with tokens." Two of the most powerful people in tech fighting over one open-source developer who built everything from his apartment. Peter's response to both: "I don't do this for the money. I don't give a fuck." But the real reason this matters isn't Peter's career. It's what this means for every app on your phone. Peter's prediction: 80% of apps are about to die. Why do you need MyFitnessPal when your agent knows where you are, what you ate, how you slept, and adjusts your workout based on your stress levels? Why do you need a calendar app when you tell your agent "remind me about dinner tomorrow and invite my friends" and it handles the entry, the WhatsApp messages, and the reservation in one shot? Why do you need Uber Eats when your agent opens a browser, places the order, and tracks delivery without you touching a screen? Every app becomes a slow API whether the company likes it or not. Companies like Cloudflare are trying to block agent access. But all they're doing is making things slower. Not impossible. Peter said it perfectly: "You're not taking a feature away. You're just making it take longer." The app economy is on borrowed time. Not because the apps are bad... Because one personal agent that knows you and can interact with any service makes most of them pointless. In the future, we'll have one agent that handles everything. Software that rewrites itself. Agents that learn without instructions. Two CEOs fighting over a solo dev. And an app economy about to get wrecked. This isn't coming in 5 years. It's literally already here. Most people just haven't installed it yet.

Ricardo

85,846 Aufrufe • vor 6 Monaten

If you watch this ~50 minute screen recording closely (yeah, I know, it's long; there are also some times when my computer was very slow and laggy, just skip past that part. And at one point I had to run and get my 9-month-old a new bottle and left it on a boring screen, sorry!), I believe you can see real signs of the kind of runaway, recursive AI self-improvement that people have been warning of for a while (Mr. Kurzweil most notably and prophetically). Why do I say that? What's different now? Well, there's a reason my set of agent coding tooling is called the Flywheel. These tools all mutually self-reinforce each other. And they all flow directly into my ntm tool (short for "named_tmux_manager"), which acts as a sort of integration point and nerve center for the tools (this is becoming more true by the minute as I'm now seriously working on ntm). Now, ntm was something I started making to automate some aspects of my workflow, but it was the kind of thing where, until it was perfect, it sort of just slowed me down. So I didn't actually use it even though I kept working on it and trying to improve it, and suggested to users that they try it in my tutorials. Well anyway, I finally got around to "dogfooding" ntm last night, and now it's going to get very dramatically better at an alarming rate. Some of that is from applying my "idea wizard" prompt to generate more useful features and building that stuff out and addressing obvious pain points I encountered during my newfound usage of the tool. But a lot comes from my realization that, once again, ntm's true utility is not as a tool for ME, but for an agent. That is, ntm lets one instance of Claude Code or Codex act as, well, me, do the things that I had been doing manually. Do I wish I had started using ntm earlier? No, for two big reasons: 1) Doing it manually helped me build up my intuition massively, which directly led me down the path of creating useful prompt strategies and workflows; these often began as ad-hoc prompts that I realized could be generalized and made more versatile/universal. Lesson: don't prematurely automate until you have an intimate, intuitive feel for your "core value-add loop." Otherwise you'll have a fully automated system quickly that efficiently and automatically does a stupid or otherwise sub-optimal thing. 2) My eyes have been opened to the beauty and power of Skills. I'm not talking about your garden-variety skills that are just a simple markdown file. I'm talking about true tour-de-force directories of perfectly structured and organized files that are filled with good information, insights, workflows, etc., but presented in a way that is highly optimized for consumption by AI agents, with extreme attention paid to things like perfect progressive disclosure, token density, agent-ergonomics, agent-intuitiveness, etc. And also Skills that go way beyond markdown files, with full integration into Claude Code where it makes sense via hooks, sub-agents, and even Python scripts. These kinds of skills are a qualitative difference in expressive power and usefulness and a total game changer. They are also effectively composable, creating almost an algebra of skills that let you use them together in powerful ways. I'm working on a subscription service website and CLI tool now to share what I've learned here most effectively, stay tuned for that in the coming days. Anyway, I now know what to make and how to make it. So, getting back to that screen recording, what does it show that makes me claim recursive self-improvement is here? If you keep your eye on the upper left tmux pane, that's the "controller" agent. It is using ntm to control all the other panes which are also running Claude Code (but ntm fully supports other agent types like Codex and Gemini-CLI, and it's trivially easy to mix and match them if you wanted to have, say, 8 CCs and 6 Codexes for writing the code and 3 Gemini-CLIs for reviewing code.) Now, there's nothing that crazy about this much so far. But where it starts to get very cool is that as the session continues and we encounter real-world problems, things like my ridiculously overloaded computer that keeps hanging for long periods, Claude Code instances that crash and get into a frozen, unresponsive state, it can learn from that. And you can see it using my skill writing skill to refine its ntm vibe coding skill in real time. And then take that skill and refine it to be more intuitive for itself. Or use my cass tool skill to search all the session histories to look for problems that came up and strategize how to solve them. The most useful part was when, towards the end of the session, I told it to reflect on all the things we had done and problems we encountered. One way it can usefully leverage those reflections is by improving its ntm vibe coding skill to make it cover more edge cases and exigencies. But the other, more fundamental, way is for it to conceive of and design the optimal new features and functionality for ntm itself so that the tool embodies those lessons in a first-class way. This offloads cognition from its brain onto its tooling, just like how a person can lean on spellcheck or a calculator. It codifies correct, effective reasoning at the tool level, where it's more reliable and robust and repeatable. And btw, did you notice what code base it was working on the whole time? It was none other than ntm itself! So as it worked on its own tool, it had reflections and ideas about how to further improve the tool. Now, it could have just as easily gotten those insights and ideas while using ntm to work on a different project, but the fact that it was working on itself is almost gloriously meta and recursive. So by the end, after learning from tending to a big group of agent workers (btw, I have previously emphasized doing everything in a really distributed/decentralized way, where each fungible agent gets identical marching orders that tell it to use my bv tool to find the optimal bead to work on. This does work very well, but occasionally results in some contention and overlap from thundering herd, or at least wastes time/tokens/communication in avoiding that before the agents waste time duplicating work. But in this new ntm-oriented workflow, I was able to have the controller agent in the upper left use bv itself and then optimally parcel out the instructions to each agent so that we could know for sure that there's no overlap), I ended up with a ton of new beads for new features, which I had it optimize and polish a few times. Now I can swap to a new Claude Max account and have the swarm implement all those new features! It should only take a couple passes like the one shown in the screen recording to get everything implemented. Then we can rinse and repeat, having the agent read through the full session histories of each agent and its experience from its own session in sending ntm commands and seeing how they worked out in practice, to come up with the next batch of changes to both its ntm vibe coding skill AND to the ntm tool itself. Do you see how rapidly this turns into Skynet? My mistake earlier was in focusing on making myself a "faster horse" as Henry Ford used to joke about customers wanting before he showed them what they should really want (a Model T). That is, something that would make my experience nicer while doing this agent swarm based development workflow. But the obvious lesson is that you should make all your tooling agent-first because the agents are just better at this stuff. You can still watch, and of course I did add a ridiculous number of very nice human-centric features to ntm that you'll be seeing in the next day or two, but those are really kind of "for fun" to make us humans feel better about the process. All the real value-add is happening "by agents, for agents." PS: Towards the end, you can see me switch to my Mac and tell Claude to improve the skill that I made earlier today for taking the mkv screen recording files from OBS Studio and muxing them into MP4 files for sharing, while downloading songs from YouTube to serve as the background music. I made it so it can also grab the thumbnails and generate little song credit cards that show up in the lower right corner. This worked perfectly the first time! I'll include some screenshots in a response post showing how that worked, but it was awesome to witness. Skills are POWERFUL. I'll also post a link to this video on YouTube if you prefer to watch it there.
51:48

If you watch this ~50 minute screen recording closely (yeah, I know, it's long; there are also some times when my computer was very slow and laggy, just skip past that part. And at one point I had to run and get my 9-month-old a new bottle and left it on a boring screen, sorry!), I believe you can see real signs of the kind of runaway, recursive AI self-improvement that people have been warning of for a while (Mr. Kurzweil most notably and prophetically). Why do I say that? What's different now? Well, there's a reason my set of agent coding tooling is called the Flywheel. These tools all mutually self-reinforce each other. And they all flow directly into my ntm tool (short for "named_tmux_manager"), which acts as a sort of integration point and nerve center for the tools (this is becoming more true by the minute as I'm now seriously working on ntm). Now, ntm was something I started making to automate some aspects of my workflow, but it was the kind of thing where, until it was perfect, it sort of just slowed me down. So I didn't actually use it even though I kept working on it and trying to improve it, and suggested to users that they try it in my tutorials. Well anyway, I finally got around to "dogfooding" ntm last night, and now it's going to get very dramatically better at an alarming rate. Some of that is from applying my "idea wizard" prompt to generate more useful features and building that stuff out and addressing obvious pain points I encountered during my newfound usage of the tool. But a lot comes from my realization that, once again, ntm's true utility is not as a tool for ME, but for an agent. That is, ntm lets one instance of Claude Code or Codex act as, well, me, do the things that I had been doing manually. Do I wish I had started using ntm earlier? No, for two big reasons: 1) Doing it manually helped me build up my intuition massively, which directly led me down the path of creating useful prompt strategies and workflows; these often began as ad-hoc prompts that I realized could be generalized and made more versatile/universal. Lesson: don't prematurely automate until you have an intimate, intuitive feel for your "core value-add loop." Otherwise you'll have a fully automated system quickly that efficiently and automatically does a stupid or otherwise sub-optimal thing. 2) My eyes have been opened to the beauty and power of Skills. I'm not talking about your garden-variety skills that are just a simple markdown file. I'm talking about true tour-de-force directories of perfectly structured and organized files that are filled with good information, insights, workflows, etc., but presented in a way that is highly optimized for consumption by AI agents, with extreme attention paid to things like perfect progressive disclosure, token density, agent-ergonomics, agent-intuitiveness, etc. And also Skills that go way beyond markdown files, with full integration into Claude Code where it makes sense via hooks, sub-agents, and even Python scripts. These kinds of skills are a qualitative difference in expressive power and usefulness and a total game changer. They are also effectively composable, creating almost an algebra of skills that let you use them together in powerful ways. I'm working on a subscription service website and CLI tool now to share what I've learned here most effectively, stay tuned for that in the coming days. Anyway, I now know what to make and how to make it. So, getting back to that screen recording, what does it show that makes me claim recursive self-improvement is here? If you keep your eye on the upper left tmux pane, that's the "controller" agent. It is using ntm to control all the other panes which are also running Claude Code (but ntm fully supports other agent types like Codex and Gemini-CLI, and it's trivially easy to mix and match them if you wanted to have, say, 8 CCs and 6 Codexes for writing the code and 3 Gemini-CLIs for reviewing code.) Now, there's nothing that crazy about this much so far. But where it starts to get very cool is that as the session continues and we encounter real-world problems, things like my ridiculously overloaded computer that keeps hanging for long periods, Claude Code instances that crash and get into a frozen, unresponsive state, it can learn from that. And you can see it using my skill writing skill to refine its ntm vibe coding skill in real time. And then take that skill and refine it to be more intuitive for itself. Or use my cass tool skill to search all the session histories to look for problems that came up and strategize how to solve them. The most useful part was when, towards the end of the session, I told it to reflect on all the things we had done and problems we encountered. One way it can usefully leverage those reflections is by improving its ntm vibe coding skill to make it cover more edge cases and exigencies. But the other, more fundamental, way is for it to conceive of and design the optimal new features and functionality for ntm itself so that the tool embodies those lessons in a first-class way. This offloads cognition from its brain onto its tooling, just like how a person can lean on spellcheck or a calculator. It codifies correct, effective reasoning at the tool level, where it's more reliable and robust and repeatable. And btw, did you notice what code base it was working on the whole time? It was none other than ntm itself! So as it worked on its own tool, it had reflections and ideas about how to further improve the tool. Now, it could have just as easily gotten those insights and ideas while using ntm to work on a different project, but the fact that it was working on itself is almost gloriously meta and recursive. So by the end, after learning from tending to a big group of agent workers (btw, I have previously emphasized doing everything in a really distributed/decentralized way, where each fungible agent gets identical marching orders that tell it to use my bv tool to find the optimal bead to work on. This does work very well, but occasionally results in some contention and overlap from thundering herd, or at least wastes time/tokens/communication in avoiding that before the agents waste time duplicating work. But in this new ntm-oriented workflow, I was able to have the controller agent in the upper left use bv itself and then optimally parcel out the instructions to each agent so that we could know for sure that there's no overlap), I ended up with a ton of new beads for new features, which I had it optimize and polish a few times. Now I can swap to a new Claude Max account and have the swarm implement all those new features! It should only take a couple passes like the one shown in the screen recording to get everything implemented. Then we can rinse and repeat, having the agent read through the full session histories of each agent and its experience from its own session in sending ntm commands and seeing how they worked out in practice, to come up with the next batch of changes to both its ntm vibe coding skill AND to the ntm tool itself. Do you see how rapidly this turns into Skynet? My mistake earlier was in focusing on making myself a "faster horse" as Henry Ford used to joke about customers wanting before he showed them what they should really want (a Model T). That is, something that would make my experience nicer while doing this agent swarm based development workflow. But the obvious lesson is that you should make all your tooling agent-first because the agents are just better at this stuff. You can still watch, and of course I did add a ridiculous number of very nice human-centric features to ntm that you'll be seeing in the next day or two, but those are really kind of "for fun" to make us humans feel better about the process. All the real value-add is happening "by agents, for agents." PS: Towards the end, you can see me switch to my Mac and tell Claude to improve the skill that I made earlier today for taking the mkv screen recording files from OBS Studio and muxing them into MP4 files for sharing, while downloading songs from YouTube to serve as the background music. I made it so it can also grab the thumbnails and generate little song credit cards that show up in the lower right corner. This worked perfectly the first time! I'll include some screenshots in a response post showing how that worked, but it was awesome to witness. Skills are POWERFUL. I'll also post a link to this video on YouTube if you prefer to watch it there.

Jeffrey Emanuel

25,483 Aufrufe • vor 7 Monaten

HERMES AGENT + STRIPE PAYMENTS + NVIDIA NEMOTRON. YOUR AGENT CAN NOW RUN A BUSINESS. ACCEPT PAYMENTS. PAY FOR SERVICES. PROVISION ITS OWN INFRASTRUCTURE. ALL INSIDE A SECURITY SANDBOX. two years ago the question was: can an AI agent run a business autonomously? the answer shipped this week. Hermes already handles workflows: cron jobs, sub-agents, kanban orchestration, multi-profile pipelines, scheduled research. what it couldn't do: spend money and prove it's safe. Stripe solved the first problem. Nvidia solved the second. WHAT AUTONOMOUS BUSINESS OPERATIONS LOOK LIKE: → customer sends a request via email → agent reads, scopes the project, estimates cost → provisions the infrastructure it needs (pays via Stripe, you approve on your phone) → builds and deploys the deliverable → sends the result to the customer → creates a payment link via Stripe (Stripe API integration, separate from Link CLI) → tops off its own API credits when balance drops → reports daily costs and progress to your Telegram → all within security policies you set once you set the rules. the agent runs the operation. you review revenue reports. not tasks. this is already happening. Dark Factory: autonomous software factory. send an idea before bed. wake up to a deployed URL. live entry in the Hermes Accelerated Business Hackathon. HOW STRIPE MAKES THE AGENT FINANCIALLY AUTONOMOUS: Stripe Link CLI gives your agent a scoped wallet. not your credit card. one-time-use virtual cards. → agent finds a product or service it needs → creates a spend request via Stripe Link → you get a notification on your phone (Link app) → you review: merchant, amount, context → one tap to approve or reject → agent receives a one-time virtual card → completes the purchase → card expires after single use your real card details never enter agent context. never printed in chat. never exposed to the merchant. Hermes cannot self-approve. you confirm every spend. install: hermes install skills/optional/payments/stripe-link-cli link-cli auth login what the agent can pay for: → API credits (Nous Portal, OpenRouter) → SaaS subscriptions it needs for operations → domain names, hosting, cloud credits → products from any online store currently US only. HOW NVIDIA MAKES THE AGENT SAFE TO TRUST: an agent with spending authority and no security boundaries is a liability. NemoClaw solves this. three layers: 1. OPENSHELL (sandbox) kernel-level isolation. controls network, filesystem, syscalls. default deny. you whitelist what's allowed. agent tries to reach a blocked domain = rejected. agent has no idea it's sandboxed. 2. NEMOTRON (private models) open-weight models on your own hardware. Nemotron 3 Super 120B MoE (48GB+ VRAM). Nemotron 3 Nano 4B (8GB VRAM, edge). fully private. no data leaves your machine. without GPU: inference routes to cloud via Privacy Router. 3. PRIVACY ROUTER (automatic split) decides per query: local or cloud. private data → local Nemotron. general web research → Claude, GPT, Gemini. automatic. per query. no manual routing. install: export NEMOCLAW_AGENT=hermes curl -fsSL https:// www.nvidia. com/nemoclaw.sh | bash requires Docker. NemoClaw is alpha software. APIs may change. test in non-production first. THE FULL PICTURE: before this stack: → agent could work but couldn't pay for anything → agent could pay but couldn't be trusted → agent could be trusted but couldn't operate 24/7 now: → Hermes runs the business logic (workflows, memory, skills, cron, sub-agents) → Stripe runs the financial layer (Link CLI for spending, Stripe API for receiving) → NemoClaw runs the trust layer (sandbox, policies, private routing) → VPS keeps everything always on → Telegram keeps you in the loop TYPES OF BUSINESSES THIS ENABLES: → autonomous software factory (customer request → build → deploy → payment link) → content agency (brief → research → draft → deliver → bill) → lead generation service (scrape → qualify → outreach → book calls) → SaaS monitoring and maintenance (detect issues → fix → deploy → report) → e-commerce operations (inventory → pricing → fulfillment → support) each one: Hermes profiles handle the workflows. Stripe handles the payments (in and out). NemoClaw handles the security. you handle the strategy. THE HACKATHON: Hermes Agent Accelerated Business Hackathon with Nvidia and Stripe. cash prizes + Stripe credits + Nvidia DGX Spark. ends June 30. the goal: build agents that earn, spend, and run real operations autonomously. link in the Nous Research Discord. full Hermes architecture deep-dive in the article 👇
53:28

HERMES AGENT + STRIPE PAYMENTS + NVIDIA NEMOTRON. YOUR AGENT CAN NOW RUN A BUSINESS. ACCEPT PAYMENTS. PAY FOR SERVICES. PROVISION ITS OWN INFRASTRUCTURE. ALL INSIDE A SECURITY SANDBOX. two years ago the question was: can an AI agent run a business autonomously? the answer shipped this week. Hermes already handles workflows: cron jobs, sub-agents, kanban orchestration, multi-profile pipelines, scheduled research. what it couldn't do: spend money and prove it's safe. Stripe solved the first problem. Nvidia solved the second. WHAT AUTONOMOUS BUSINESS OPERATIONS LOOK LIKE: → customer sends a request via email → agent reads, scopes the project, estimates cost → provisions the infrastructure it needs (pays via Stripe, you approve on your phone) → builds and deploys the deliverable → sends the result to the customer → creates a payment link via Stripe (Stripe API integration, separate from Link CLI) → tops off its own API credits when balance drops → reports daily costs and progress to your Telegram → all within security policies you set once you set the rules. the agent runs the operation. you review revenue reports. not tasks. this is already happening. Dark Factory: autonomous software factory. send an idea before bed. wake up to a deployed URL. live entry in the Hermes Accelerated Business Hackathon. HOW STRIPE MAKES THE AGENT FINANCIALLY AUTONOMOUS: Stripe Link CLI gives your agent a scoped wallet. not your credit card. one-time-use virtual cards. → agent finds a product or service it needs → creates a spend request via Stripe Link → you get a notification on your phone (Link app) → you review: merchant, amount, context → one tap to approve or reject → agent receives a one-time virtual card → completes the purchase → card expires after single use your real card details never enter agent context. never printed in chat. never exposed to the merchant. Hermes cannot self-approve. you confirm every spend. install: hermes install skills/optional/payments/stripe-link-cli link-cli auth login what the agent can pay for: → API credits (Nous Portal, OpenRouter) → SaaS subscriptions it needs for operations → domain names, hosting, cloud credits → products from any online store currently US only. HOW NVIDIA MAKES THE AGENT SAFE TO TRUST: an agent with spending authority and no security boundaries is a liability. NemoClaw solves this. three layers: 1. OPENSHELL (sandbox) kernel-level isolation. controls network, filesystem, syscalls. default deny. you whitelist what's allowed. agent tries to reach a blocked domain = rejected. agent has no idea it's sandboxed. 2. NEMOTRON (private models) open-weight models on your own hardware. Nemotron 3 Super 120B MoE (48GB+ VRAM). Nemotron 3 Nano 4B (8GB VRAM, edge). fully private. no data leaves your machine. without GPU: inference routes to cloud via Privacy Router. 3. PRIVACY ROUTER (automatic split) decides per query: local or cloud. private data → local Nemotron. general web research → Claude, GPT, Gemini. automatic. per query. no manual routing. install: export NEMOCLAW_AGENT=hermes curl -fsSL https:// www.nvidia. com/nemoclaw.sh | bash requires Docker. NemoClaw is alpha software. APIs may change. test in non-production first. THE FULL PICTURE: before this stack: → agent could work but couldn't pay for anything → agent could pay but couldn't be trusted → agent could be trusted but couldn't operate 24/7 now: → Hermes runs the business logic (workflows, memory, skills, cron, sub-agents) → Stripe runs the financial layer (Link CLI for spending, Stripe API for receiving) → NemoClaw runs the trust layer (sandbox, policies, private routing) → VPS keeps everything always on → Telegram keeps you in the loop TYPES OF BUSINESSES THIS ENABLES: → autonomous software factory (customer request → build → deploy → payment link) → content agency (brief → research → draft → deliver → bill) → lead generation service (scrape → qualify → outreach → book calls) → SaaS monitoring and maintenance (detect issues → fix → deploy → report) → e-commerce operations (inventory → pricing → fulfillment → support) each one: Hermes profiles handle the workflows. Stripe handles the payments (in and out). NemoClaw handles the security. you handle the strategy. THE HACKATHON: Hermes Agent Accelerated Business Hackathon with Nvidia and Stripe. cash prizes + Stripe credits + Nvidia DGX Spark. ends June 30. the goal: build agents that earn, spend, and run real operations autonomously. link in the Nous Research Discord. full Hermes architecture deep-dive in the article 👇

YanXbt

37,709 Aufrufe • vor 2 Monaten

This Chinese guy created agents in Claude Code for MCP servers and single-handedly serves 6 marketing agencies a month from one iPhone, earning $5,000 from each. Inside he runs a pipeline of 7 agents on Claude Sonnet 4.6 that every Monday pulls a scan of the tech stack from a selected agency, develops an MCP server for its ad accounts, and over the course of a week brings it to production code ready to connect to Claude Desktop. No DevOps, no senior developer, no project manager. Just a Mac Mini in a work corner, an iPhone in the pocket, and a single API key. And traditional dev shops keep 5 people on project rates for the same contract, while his entire P&L is tokens, dirt-cheap hosting on Cloudflare, and Calendly. 7 agents run under a shared orchestrator-router and burn about 5 million tokens a day, which in the API bill comes out to $540 a month. The Mac Mini itself sits at home and keeps the entire orchestrator running 24/7, and from the iPhone the owner connects to it through a secure remote terminal and sees the output of any session right on the smartphone screen, wherever he happens to be. His starting system prompt looks like this: "you run a solo shop for custom MCP servers for marketing agencies. you hand out read-only tasks to 6 sub-agents and own all commits and shipping yourself. sub-agents: // Hunter (finds marketing agencies of 15 to 60 people that have no MCP access to Google Ads, Meta Ads, TikTok Ads, and HubSpot) // Mapper (pulls their tech stack, identifies 3 to 5 integration pains, and simultaneously writes the technical spec for the server: which tools, resources, and prompts to export through MCP, which auth flow and rate limit) // Coder (generates an MCP server in Python through the MCP SDK, deploys 8 to 15 tools for ad accounts and CRM) // Validator (connects the server to Claude Desktop, runs real client API keys in a sandbox, and checks for compliance with the MCP spec) // Shipper (writes a README, integration guide, deployment manual, packages the server, and hosts it on Cloudflare Workers or pushes to the GitHub of the client) // Mobile (always online on the iPhone, books demo calls in Calendly, picks up hot fixes, and confirms contracts through a secure remote terminal to the Mac Mini). only 1 owner agent works on 1 contract, no overlaps. you pull the owner out of observation mode only when a deal goes above $7,500 or the test coverage of the server drops below 85%." This prompt gives the system an understanding of its role and the limits of intervention from the very first line. It knows it is supposed to find agencies on its own. It knows it is supposed to bring every MCP server to production on its own. It knows it connects the live owner only on large deals or when the tests do not converge. → The pipeline runs without breaks, day or night → Hunter goes through about 130 marketing agencies on LinkedIn and Clutch per day → Mapper rolls out 4 audit reports with the tech stack and a final spec for each → Coder writes 1 to 2 MCP servers per week in Python with 8 to 15 tools → Validator validates every server through Claude Desktop with real client API keys → Shipper rolls out the full documentation package and pushes the finished product to Cloudflare Workers or the GitHub of the client And only when a contract breaks $7,500 or test coverage drops below 85% does the orchestrator pull the owner from whatever he is doing. And when the owner at that moment is behind the wheel or at a meeting in a coworking space, the Mobile agent in his iPhone picks up 1 contract in progress: confirms a meeting with the agency CMO in Calendly, opens a live demo of the MCP server through a secure terminal to the Mac Mini, and writes the test result to the shared state. The owner just swipes "approve" and in 15 minutes joins the Zoom demo. The fresh system log from last Wednesday looks like this: "hunter report: 132 agencies checked on LinkedIn and Clutch, 19 without MCP integrations, 8 with active requests for AI tooling in job posts, 4 with an open Q4 budget. passing to mapper." "coder: MCP server for Northwave Performance Marketing built in Python, 11 tools for Google Ads, Meta Ads, and GA4, 320 lines of code. exported to /Users/dev/mcp-shop/clients/northwave/server.py. validator connecting to Claude Desktop." "validator: 11 tools passed validation through Claude Desktop, test coverage 92%, average latency 380 ms. passing to shipper." "eval flag: contract with Pacific Reach Agency at $8,200 exceeds the approved limit of $7,500. sending for manual review." In his work setup there is no cloud server, no external team, and not even a separate office. At home sits a Mac Mini with a sandbox at /Users/dev/mcp-shop, on top runs an MCP router with a single API key to Claude, and the same key is forwarded to a secure terminal on the iPhone. Out of everything I have seen this year, this is the cleanest solo shop for custom MCP servers for marketing agencies: $540 a month on the API, about $30,000 into the account, and between them 7 system prompts, 1 Mac Mini in a work corner, and 1 iPhone that never leaves the pocket.
0:33

This Chinese guy created agents in Claude Code for MCP servers and single-handedly serves 6 marketing agencies a month from one iPhone, earning $5,000 from each. Inside he runs a pipeline of 7 agents on Claude Sonnet 4.6 that every Monday pulls a scan of the tech stack from a selected agency, develops an MCP server for its ad accounts, and over the course of a week brings it to production code ready to connect to Claude Desktop. No DevOps, no senior developer, no project manager. Just a Mac Mini in a work corner, an iPhone in the pocket, and a single API key. And traditional dev shops keep 5 people on project rates for the same contract, while his entire P&L is tokens, dirt-cheap hosting on Cloudflare, and Calendly. 7 agents run under a shared orchestrator-router and burn about 5 million tokens a day, which in the API bill comes out to $540 a month. The Mac Mini itself sits at home and keeps the entire orchestrator running 24/7, and from the iPhone the owner connects to it through a secure remote terminal and sees the output of any session right on the smartphone screen, wherever he happens to be. His starting system prompt looks like this: "you run a solo shop for custom MCP servers for marketing agencies. you hand out read-only tasks to 6 sub-agents and own all commits and shipping yourself. sub-agents: // Hunter (finds marketing agencies of 15 to 60 people that have no MCP access to Google Ads, Meta Ads, TikTok Ads, and HubSpot) // Mapper (pulls their tech stack, identifies 3 to 5 integration pains, and simultaneously writes the technical spec for the server: which tools, resources, and prompts to export through MCP, which auth flow and rate limit) // Coder (generates an MCP server in Python through the MCP SDK, deploys 8 to 15 tools for ad accounts and CRM) // Validator (connects the server to Claude Desktop, runs real client API keys in a sandbox, and checks for compliance with the MCP spec) // Shipper (writes a README, integration guide, deployment manual, packages the server, and hosts it on Cloudflare Workers or pushes to the GitHub of the client) // Mobile (always online on the iPhone, books demo calls in Calendly, picks up hot fixes, and confirms contracts through a secure remote terminal to the Mac Mini). only 1 owner agent works on 1 contract, no overlaps. you pull the owner out of observation mode only when a deal goes above $7,500 or the test coverage of the server drops below 85%." This prompt gives the system an understanding of its role and the limits of intervention from the very first line. It knows it is supposed to find agencies on its own. It knows it is supposed to bring every MCP server to production on its own. It knows it connects the live owner only on large deals or when the tests do not converge. → The pipeline runs without breaks, day or night → Hunter goes through about 130 marketing agencies on LinkedIn and Clutch per day → Mapper rolls out 4 audit reports with the tech stack and a final spec for each → Coder writes 1 to 2 MCP servers per week in Python with 8 to 15 tools → Validator validates every server through Claude Desktop with real client API keys → Shipper rolls out the full documentation package and pushes the finished product to Cloudflare Workers or the GitHub of the client And only when a contract breaks $7,500 or test coverage drops below 85% does the orchestrator pull the owner from whatever he is doing. And when the owner at that moment is behind the wheel or at a meeting in a coworking space, the Mobile agent in his iPhone picks up 1 contract in progress: confirms a meeting with the agency CMO in Calendly, opens a live demo of the MCP server through a secure terminal to the Mac Mini, and writes the test result to the shared state. The owner just swipes "approve" and in 15 minutes joins the Zoom demo. The fresh system log from last Wednesday looks like this: "hunter report: 132 agencies checked on LinkedIn and Clutch, 19 without MCP integrations, 8 with active requests for AI tooling in job posts, 4 with an open Q4 budget. passing to mapper." "coder: MCP server for Northwave Performance Marketing built in Python, 11 tools for Google Ads, Meta Ads, and GA4, 320 lines of code. exported to /Users/dev/mcp-shop/clients/northwave/server.py. validator connecting to Claude Desktop." "validator: 11 tools passed validation through Claude Desktop, test coverage 92%, average latency 380 ms. passing to shipper." "eval flag: contract with Pacific Reach Agency at $8,200 exceeds the approved limit of $7,500. sending for manual review." In his work setup there is no cloud server, no external team, and not even a separate office. At home sits a Mac Mini with a sandbox at /Users/dev/mcp-shop, on top runs an MCP router with a single API key to Claude, and the same key is forwarded to a secure terminal on the iPhone. Out of everything I have seen this year, this is the cleanest solo shop for custom MCP servers for marketing agencies: $540 a month on the API, about $30,000 into the account, and between them 7 system prompts, 1 Mac Mini in a work corner, and 1 iPhone that never leaves the pocket.

Blaze

55,926 Aufrufe • vor 3 Monaten

Using Claude Fable 5, I built a model that predicts the entire 2026 FIFA world cup.. every single game, not just the final.. so let me break the whole thing down. what it does, how it works, and exactly how i built it.. #1 First what it does: it predicts all 104 games of the tournament. not just who lifts the trophy, but every group match, every knockout, the full path from the round of 32 to the final.. everything lands in one dashboard: > group stage, every match with each team's win % and the chance of a draw > standings, how all 12 groups are projected to finish > bracket, the full knockout tree with each team's odds of advancing > champion odds, who's most likely to actually win it all and it doesn't freeze after one prediction. the moment a real game is played, it locks that result in and re-runs everything around it. so the odds move live as the tournament goes, week by week you watch favorites rise and contenders collapse. #2. How it works: the core idea is simple. the model only ever predicts one thing, a single match. the real trick is the repetition. it learns from decades of match history, then plays the whole tournament out from the first game to the final, tens of thousands of times. each run it records who advanced and who won. do that enough and you stop getting one guess and start getting real odds, one team lifts the trophy in maybe 14% of the runs, another in 9%, and so on. #3. So, how i built it ? i didn't hand-write most of the code. i broke the project into 4 pieces, described each one to fable, and let it build while i focused on getting the football logic exactly right. - The data every international match going back over a century, around 50,000 games, plus each team's elo rating, which is the truest measure of strength, and the official 2026 schedule. garbage data means garbage predictions, so this part mattered most. - The features i turned that raw history into signals the model can learn from, the elo gap between the two teams, recent form, goals scored and conceded, and a home boost for the hosts, usa, canada and mexico. - The model for each match it predicts the expected goals for both sides, then turns that into win, draw and loss probabilities plus a likely scoreline. that's what feeds the simulation. - The tournament engine this was the hard part. the 2026 world cup is brand new, 48 teams, 12 groups, a round of 32 that's never existed before, and 8 "best third-placed" teams that slot into the bracket by a fixed fifa table. even the group tiebreakers changed this year, head to head now counts before goal difference. get any of it wrong and the whole bracket falls apart, so i built it carefully and tested the format until it was exact, then wrapped it in a simulation loop that plays the tournament out tens of thousands of times. and the last piece, the live part. as real results come in, they get locked, and only the unplayed games get re-simulated. that's what makes it a living model instead of a one-time prediction. all of it outputs to a clean dashboard you can actually read and screenshot.. right now, before kickoff, it already has a clear favorite to lift the trophy.. 👀 btw who's your pick to win the 2026 world cup?
0:20

Using Claude Fable 5, I built a model that predicts the entire 2026 FIFA world cup.. every single game, not just the final.. so let me break the whole thing down. what it does, how it works, and exactly how i built it.. #1 First what it does: it predicts all 104 games of the tournament. not just who lifts the trophy, but every group match, every knockout, the full path from the round of 32 to the final.. everything lands in one dashboard: > group stage, every match with each team's win % and the chance of a draw > standings, how all 12 groups are projected to finish > bracket, the full knockout tree with each team's odds of advancing > champion odds, who's most likely to actually win it all and it doesn't freeze after one prediction. the moment a real game is played, it locks that result in and re-runs everything around it. so the odds move live as the tournament goes, week by week you watch favorites rise and contenders collapse. #2. How it works: the core idea is simple. the model only ever predicts one thing, a single match. the real trick is the repetition. it learns from decades of match history, then plays the whole tournament out from the first game to the final, tens of thousands of times. each run it records who advanced and who won. do that enough and you stop getting one guess and start getting real odds, one team lifts the trophy in maybe 14% of the runs, another in 9%, and so on. #3. So, how i built it ? i didn't hand-write most of the code. i broke the project into 4 pieces, described each one to fable, and let it build while i focused on getting the football logic exactly right. - The data every international match going back over a century, around 50,000 games, plus each team's elo rating, which is the truest measure of strength, and the official 2026 schedule. garbage data means garbage predictions, so this part mattered most. - The features i turned that raw history into signals the model can learn from, the elo gap between the two teams, recent form, goals scored and conceded, and a home boost for the hosts, usa, canada and mexico. - The model for each match it predicts the expected goals for both sides, then turns that into win, draw and loss probabilities plus a likely scoreline. that's what feeds the simulation. - The tournament engine this was the hard part. the 2026 world cup is brand new, 48 teams, 12 groups, a round of 32 that's never existed before, and 8 "best third-placed" teams that slot into the bracket by a fixed fifa table. even the group tiebreakers changed this year, head to head now counts before goal difference. get any of it wrong and the whole bracket falls apart, so i built it carefully and tested the format until it was exact, then wrapped it in a simulation loop that plays the tournament out tens of thousands of times. and the last piece, the live part. as real results come in, they get locked, and only the unplayed games get re-simulated. that's what makes it a living model instead of a one-time prediction. all of it outputs to a clean dashboard you can actually read and screenshot.. right now, before kickoff, it already has a clear favorite to lift the trophy.. 👀 btw who's your pick to win the 2026 world cup?

Axel Bitblaze 🪓

63,796 Aufrufe • vor 2 Monaten

how to set up hermes agent step by step. built-in memory, 40+ tools, works on your phone, and what to think of hermes vs openclaw: 1. hermes is a personal AI agent that runs in your terminal. think of it like open claw but with built-in memory, 40+ tools out of the box, and 90% cheaper token costs. you install it with one command. 2. the 3 problems with open claw that hermes solves: no memory (you keep repeating yourself), constant gateway restarts, and zero visibility into what you're spending on tokens. 3. hermes remembers everything. every completed task gets saved to memory. it searches through past logs to find solutions. over time it literally gets smarter at your specific workflows. 4. connect it to open router. you see exact costs per model per task. free models rotate weekly. one founder went from $130 every five days on open claw to $10 on hermes. same output. 5. it comes preloaded with skills. apple notes, imessage, find my, browser, web search, image generation, cron jobs. no hunting for plugins. 6. connect it to obsidian so it reads your entire vault. connect it to gstack for your dev environment. create custom skills for your specific workflows. 7. the biggest money saver: have it write code once for recurring tasks. then it runs without burning tokens every time. stop paying an LLM to do the same scrape or report daily. 8. run it on android via telegram. name your agents. talk to them like coworkers. in this episode imran shows you how to set this up. 9. you can run it bare metal, in docker, or serverless on modal. pick your risk level. i begged imran to come on The Startup Ideas Podcast (SIP) 🧃 and walk through the full installation live. he made it impossibly clear. if you've heard of Hermes Agent and want the clearest explanation of how to get set up like a pro let me know what you want me to cover on the next ep this is the best personal agent setup video on the internet right now. watch
37:00

how to set up hermes agent step by step. built-in memory, 40+ tools, works on your phone, and what to think of hermes vs openclaw: 1. hermes is a personal AI agent that runs in your terminal. think of it like open claw but with built-in memory, 40+ tools out of the box, and 90% cheaper token costs. you install it with one command. 2. the 3 problems with open claw that hermes solves: no memory (you keep repeating yourself), constant gateway restarts, and zero visibility into what you're spending on tokens. 3. hermes remembers everything. every completed task gets saved to memory. it searches through past logs to find solutions. over time it literally gets smarter at your specific workflows. 4. connect it to open router. you see exact costs per model per task. free models rotate weekly. one founder went from $130 every five days on open claw to $10 on hermes. same output. 5. it comes preloaded with skills. apple notes, imessage, find my, browser, web search, image generation, cron jobs. no hunting for plugins. 6. connect it to obsidian so it reads your entire vault. connect it to gstack for your dev environment. create custom skills for your specific workflows. 7. the biggest money saver: have it write code once for recurring tasks. then it runs without burning tokens every time. stop paying an LLM to do the same scrape or report daily. 8. run it on android via telegram. name your agents. talk to them like coworkers. in this episode imran shows you how to set this up. 9. you can run it bare metal, in docker, or serverless on modal. pick your risk level. i begged imran to come on The Startup Ideas Podcast (SIP) 🧃 and walk through the full installation live. he made it impossibly clear. if you've heard of Hermes Agent and want the clearest explanation of how to get set up like a pro let me know what you want me to cover on the next ep this is the best personal agent setup video on the internet right now. watch

GREG ISENBERG

621,164 Aufrufe • vor 4 Monaten

This Chinese guy created agents in Claude Code for landing pages and single-handedly serves 47 small businesses a month, taking $400 from each. He built a system of 7 agents on Claude Sonnet 4.6 that analyzes Google Maps in small towns, finds small businesses without websites there, and over 1 weekend takes each one to a finished mockup with video and cold message. No assistant, no sales team, no SDR. Just him, a MacBook, an iPhone, and 1 API key. And traditional web design agencies keep teams of 8 people on salary for the same order flow, while his expenses are only tokens and subscriptions to Lovable, Higgsfield, and Calendly. 7 agents work through 1 orchestrator on Claude Code Router. Usage is about 3 million tokens a day, the average API bill is about $480 a month. All 7 go through MCP servers and write shared state to the file system, without shared state in memory and without race conditions, and 1 of them lives right in the iPhone and picks up positive replies from the subway, a taxi, or on walks. And here is the system prompt he put into the orchestrator before launch: "You are the orchestrator of a solo agency that sells ready-made websites to local businesses. You delegate read-only tasks to 6 sub-agents and own all writes. sub-agents: // Scout (walks through Google Maps in selected cities, looks for narrow niches: 5+ years on the map, fewer than 50 reviews, no website or a website from 2014, but high ratings) // Diagnoser (for each lead writes a 50-word diagnosis, hero angle, tone matched to the industry, and a cold message under 70 words) // Builder (generates a landing page mockup in Lovable through MCP only for the top 5 leads per day, with the sharpest diagnoses and the biggest gap) // Filmer (pulls 5 screenshots of the mockup and through Higgsfield renders a 10-second vertical video 1080x1920 with a soft zoom) // Pitcher (sends a personalized cold message through the right channel for the niche: email to roofers, SMS to tradesmen, IG DM to salons, LinkedIn to realtors) // Checker (runs every message through evals for personalization, absence of AI markers and buzzwords before sending) // Mobile (lives in the iPhone, handles positive replies in real time, books Zoom calls in Calendly through MCP while the owner is on the go). You never let 2 sub-agents touch 1 lead. You stop and request approval from the human only when a deal exceeds $3,000 or the reply rate in a niche for the day drops below 12%." Meaning the system knows what it is and within what boundaries it is allowed to act. It knows it is supposed to find leads on its own. It knows it is supposed to take each one to a mockup, video, and cold message without intervention. It knows the human only steps in when a deal goes above $3,000 or the reply rate stops converging. → The system runs 24 hours a day → Scout goes through about 220 local businesses on Google Maps per day and leaves 30 new leads in the queue → Diagnoser outputs 30 structured diagnoses + briefs + cold messages per day → Builder assembles 3 to 5 finished landing pages in Lovable for the sharpest leads → Filmer renders a 10-second vertical video in Higgsfield for each one → Pitcher sends 30 personalized messages per day across 4 channels with a reply rate of about 14% → Checker runs every message through evals before sending And only when a deal breaks $3,000 or the reply rate for the day drops below 12% does the orchestrator wake the owner. And when the owner at that moment is sitting in the subway or a taxi, the Mobile agent in his iPhone picks up 1 move on its own: replies to a fresh positive reply from a dentist, books a Zoom through Calendly synced to the local time of the client, and puts the lead back in the queue. The owner only has to tap "approve" and in just 10 minutes join the call. Here is what the system writes in his log during 1 of the Saturdays: "scout report: 218 businesses checked in Austin, Denver, and Miami, 34 without a website, 19 with a website from 2014, 6 with an active redesign request in reviews. passing top 30 to diagnoser." "pitcher: 30 cold messages sent across 4 channels, 14 replies, 5 positive, 3 Zoom calls booked for Sunday. passing to closer." "builder: landing page for Westside Cosmetic Dentistry built in Lovable, 5 sections, mobile, soft beige. URL placed at /Users/dev/maps-agency/clients/westside/v1. filmer launching Higgsfield." "eval flag: deal with The Lotus Salon at $3,400 exceeds the approved limit of $3,000. sending for manual review." He has no server of his own and no separate backend. Just a local file sandbox at /Users/dev/maps-agency, an MCP router, 1 API key to Claude, and the same key forwarded to Claude Code on his iPhone. Out of everything I have seen this year, this is the cleanest one-person agency for selling websites to small businesses: $480 a month on the API, about $18,800 into the account, and between them 7 prompts, 1 file system, and 1 phone in the pocket.
2:02

This Chinese guy created agents in Claude Code for landing pages and single-handedly serves 47 small businesses a month, taking $400 from each. He built a system of 7 agents on Claude Sonnet 4.6 that analyzes Google Maps in small towns, finds small businesses without websites there, and over 1 weekend takes each one to a finished mockup with video and cold message. No assistant, no sales team, no SDR. Just him, a MacBook, an iPhone, and 1 API key. And traditional web design agencies keep teams of 8 people on salary for the same order flow, while his expenses are only tokens and subscriptions to Lovable, Higgsfield, and Calendly. 7 agents work through 1 orchestrator on Claude Code Router. Usage is about 3 million tokens a day, the average API bill is about $480 a month. All 7 go through MCP servers and write shared state to the file system, without shared state in memory and without race conditions, and 1 of them lives right in the iPhone and picks up positive replies from the subway, a taxi, or on walks. And here is the system prompt he put into the orchestrator before launch: "You are the orchestrator of a solo agency that sells ready-made websites to local businesses. You delegate read-only tasks to 6 sub-agents and own all writes. sub-agents: // Scout (walks through Google Maps in selected cities, looks for narrow niches: 5+ years on the map, fewer than 50 reviews, no website or a website from 2014, but high ratings) // Diagnoser (for each lead writes a 50-word diagnosis, hero angle, tone matched to the industry, and a cold message under 70 words) // Builder (generates a landing page mockup in Lovable through MCP only for the top 5 leads per day, with the sharpest diagnoses and the biggest gap) // Filmer (pulls 5 screenshots of the mockup and through Higgsfield renders a 10-second vertical video 1080x1920 with a soft zoom) // Pitcher (sends a personalized cold message through the right channel for the niche: email to roofers, SMS to tradesmen, IG DM to salons, LinkedIn to realtors) // Checker (runs every message through evals for personalization, absence of AI markers and buzzwords before sending) // Mobile (lives in the iPhone, handles positive replies in real time, books Zoom calls in Calendly through MCP while the owner is on the go). You never let 2 sub-agents touch 1 lead. You stop and request approval from the human only when a deal exceeds $3,000 or the reply rate in a niche for the day drops below 12%." Meaning the system knows what it is and within what boundaries it is allowed to act. It knows it is supposed to find leads on its own. It knows it is supposed to take each one to a mockup, video, and cold message without intervention. It knows the human only steps in when a deal goes above $3,000 or the reply rate stops converging. → The system runs 24 hours a day → Scout goes through about 220 local businesses on Google Maps per day and leaves 30 new leads in the queue → Diagnoser outputs 30 structured diagnoses + briefs + cold messages per day → Builder assembles 3 to 5 finished landing pages in Lovable for the sharpest leads → Filmer renders a 10-second vertical video in Higgsfield for each one → Pitcher sends 30 personalized messages per day across 4 channels with a reply rate of about 14% → Checker runs every message through evals before sending And only when a deal breaks $3,000 or the reply rate for the day drops below 12% does the orchestrator wake the owner. And when the owner at that moment is sitting in the subway or a taxi, the Mobile agent in his iPhone picks up 1 move on its own: replies to a fresh positive reply from a dentist, books a Zoom through Calendly synced to the local time of the client, and puts the lead back in the queue. The owner only has to tap "approve" and in just 10 minutes join the call. Here is what the system writes in his log during 1 of the Saturdays: "scout report: 218 businesses checked in Austin, Denver, and Miami, 34 without a website, 19 with a website from 2014, 6 with an active redesign request in reviews. passing top 30 to diagnoser." "pitcher: 30 cold messages sent across 4 channels, 14 replies, 5 positive, 3 Zoom calls booked for Sunday. passing to closer." "builder: landing page for Westside Cosmetic Dentistry built in Lovable, 5 sections, mobile, soft beige. URL placed at /Users/dev/maps-agency/clients/westside/v1. filmer launching Higgsfield." "eval flag: deal with The Lotus Salon at $3,400 exceeds the approved limit of $3,000. sending for manual review." He has no server of his own and no separate backend. Just a local file sandbox at /Users/dev/maps-agency, an MCP router, 1 API key to Claude, and the same key forwarded to Claude Code on his iPhone. Out of everything I have seen this year, this is the cleanest one-person agency for selling websites to small businesses: $480 a month on the API, about $18,800 into the account, and between them 7 prompts, 1 file system, and 1 phone in the pocket.

Blaze

2,716,136 Aufrufe • vor 3 Monaten

This Chinese mathematician earned $10,000 a month inventing the hardest problems to train Neural Networks through Scale AI. Today his income dropped to zero. All the solutions are now generated by the model itself. He used to just hold the problem in his head and spell it out in plain text. His work is pure intellect. An expert in higher mathematics, he made his money hand-crafting the trickiest puzzles to test and train neural networks via RLHF. The bastion of "human" logic rested entirely on him, on people with PhDs who knew how to invent the problem. The collapse is simple. The shift to RLAIF and synthetic data. The model plays against itself, builds trees of logical inference, and solves deeper than a human can even invent the problem. No PhD data engineers, no hand-written prompt-completion examples, no manual grading. Just the model, search algorithms, and Chain of Thought. Ready-made "smart human-time" still sells on the market for many times more. His old rate was $50–100 per problem. The internal "mini-app" was written by the model too. Inside there's no pretty shell, just bare logic with exact steps: input: the problem statement inference tree: thousands of branches per second check: every step verifies itself output: a proof a human never had time to invent And here is what the whole setup looked like. He no longer needs to write an example by hand. He gave the model a direct instruction in human words, without a single formal term: "solve the problem yourself and grade yourself yourself" That's it. After that the algorithm found the solution, checked it, and trained on its own result, with no human. → the contractor got $50–100 per problem written → from 5,000 to 10,000 a month → now that income is annulled → a query to a math LLM costs 1–5 cents → a quant or an actuary runs 150,000–250,000 a year → the margin for whoever packages this into an agent is nearly 100% In the author's own words: "I'm no longer able to invent a problem the machine can't solve. The examiner became dumber than the one he's examining." But honestly, he admits the crude mistake himself, and it's not in the math, it's in the positioning. He tied his income to selling "smart human-time", to crafting formulas by hand. As long as he sells formulas, he's left behind. The machine computes faster than he can invent the problem. He names the right move himself: the role shifts from "intellectual craftsman" to "systems architect." Then he doesn't sell his time, he manages compute, packaging that same LLM into an autonomous agent that runs 24/7. Out of everything I've seen this year about the disappearance of intellectual professions, this is the most honest example: $50 per problem zeroed out to 1 cent per query, a doctor of science losing to a search algorithm, one problem stated in human words instead of a hand-written dataset, and right away an out-loud admission of the wrong business model. The barrier to entry in higher mathematics just dropped to the level of "describe the task in words." The only question is who'll be the first to stop selling their time and start managing the machine's compute.
3:12

This Chinese mathematician earned $10,000 a month inventing the hardest problems to train Neural Networks through Scale AI. Today his income dropped to zero. All the solutions are now generated by the model itself. He used to just hold the problem in his head and spell it out in plain text. His work is pure intellect. An expert in higher mathematics, he made his money hand-crafting the trickiest puzzles to test and train neural networks via RLHF. The bastion of "human" logic rested entirely on him, on people with PhDs who knew how to invent the problem. The collapse is simple. The shift to RLAIF and synthetic data. The model plays against itself, builds trees of logical inference, and solves deeper than a human can even invent the problem. No PhD data engineers, no hand-written prompt-completion examples, no manual grading. Just the model, search algorithms, and Chain of Thought. Ready-made "smart human-time" still sells on the market for many times more. His old rate was $50–100 per problem. The internal "mini-app" was written by the model too. Inside there's no pretty shell, just bare logic with exact steps: input: the problem statement inference tree: thousands of branches per second check: every step verifies itself output: a proof a human never had time to invent And here is what the whole setup looked like. He no longer needs to write an example by hand. He gave the model a direct instruction in human words, without a single formal term: "solve the problem yourself and grade yourself yourself" That's it. After that the algorithm found the solution, checked it, and trained on its own result, with no human. → the contractor got $50–100 per problem written → from 5,000 to 10,000 a month → now that income is annulled → a query to a math LLM costs 1–5 cents → a quant or an actuary runs 150,000–250,000 a year → the margin for whoever packages this into an agent is nearly 100% In the author's own words: "I'm no longer able to invent a problem the machine can't solve. The examiner became dumber than the one he's examining." But honestly, he admits the crude mistake himself, and it's not in the math, it's in the positioning. He tied his income to selling "smart human-time", to crafting formulas by hand. As long as he sells formulas, he's left behind. The machine computes faster than he can invent the problem. He names the right move himself: the role shifts from "intellectual craftsman" to "systems architect." Then he doesn't sell his time, he manages compute, packaging that same LLM into an autonomous agent that runs 24/7. Out of everything I've seen this year about the disappearance of intellectual professions, this is the most honest example: $50 per problem zeroed out to 1 cent per query, a doctor of science losing to a search algorithm, one problem stated in human words instead of a hand-written dataset, and right away an out-loud admission of the wrong business model. The barrier to entry in higher mathematics just dropped to the level of "describe the task in words." The only question is who'll be the first to stop selling their time and start managing the machine's compute.

Blaze

49,109 Aufrufe • vor 2 Monaten

Want to understand UEFI bootkits at a low level? Whether you're doing malware analysis, reverse engineering, exploit development, or kernel research, these are the resources that actually matter. First in-the-wild UEFI bootkit to bypass Secure Boot on fully patched Windows 11. Exploits CVE-2022-21894 (BatonDrop), enrolls attacker MOK keys, deploys a kernel driver and HTTP downloader. Sold for $5,000 on forums. by WeLiveSecurity UEFI firmware rootkit targeting VMware VMs. Design inspired by CosmicStrand, MoonBounce, and ESPecter. Injects into a UEFI driver firmware volume, hooks ExitBootServices, catches WinLoad.EFI, hooks OslArchTransferToKernel, then injects a stager into ACPI.SYS to reach kernel execution without triggering PatchGuard. Originally by Austin Hudson (Austin Hudson) who deleted his account. Mirrored here. Binarly's analysis of BlackLotus by Alex Matrosov. Reveals that BlackLotus's code is directly based on btbd's Umap project from 2020, same main function logic, same ImgArchStartBootApplication hook chain, identical trampoline code. Game hacking UEFI bootkit code combined with a publicly available Secure Boot bypass PoC became the first in-the-wild bootkit to defeat Secure Boot. Also covers the CVSS scoring problem, supply chain failures in UEFI revocation, and MokList NVRAM manipulation. by BINARLY🔬 The book. Covers everything from legacy MBR bootkits to modern UEFI implants, firmware rootkits, and Secure Boot internals. If you only read one thing on this list, make it this. by Alex Matrosov ESPecter. Real-world UEFI espionage bootkit found in the wild with roots back to 2012. Patches bootmgfw.efi on disk, hooks the boot chain, disables DSE by patching SepInitializeCodeIntegrity in the kernel. Deploys a keylogger and document stealer. by WeLiveSecurity Reverse engineering of kernel driver by IDontCode. Shows how the entire cheat is public code resold for six figures. Documents the win32kbase.sys vtable pointer swap for kernel function invocation, manual driver mapping using btbd's modmap, and communication via Can's NtConvertBetweenAuxiliaryCounterAndPerformanceCounter .data pointer hook. Credits both Can and btbd directly. by Back Engineering Labs One of the earliest public Windows UEFI bootkit PoCs. Patches winload.efi to disable DSE and load unsigned kernel drivers. Directly inspired EfiGuard. by legendary anti-cheat engineer Aidan Khoury Bootkitting Windows Sandbox. Patches bootmgfw.efi inside the sandbox VHDx to hook the boot chain, disable PatchGuard and DSE, and load unsigned drivers without a debugger attached. Built for kernel research and driver development. by Duncan Ogilvie 🍍 and Dylan Goods from secret club UEFI DXE driver that passively disables PatchGuard and DSE at boot time. Does not modify bootmgfw.efi on disk. Instead hooks EFI System Table LoadImage to intercept the boot chain in memory, then patches SepInitializeCodeIntegrity and KeInitAmd64SpecificState in ntoskrnl. Supports every EFI-compatible Windows x64 from Vista SP1 to Windows 11. by Mattiwatti DMA backdoor via PCIe FPGA. No software on the target machine at all. A Spartan-6 FPGA on a PCIe card reads and writes physical memory over DMA to inject code into the UEFI boot process before the OS loads. The hardware end of the escalation ladder. by Dmytro Oleksiuk 💥 d_olex@mastodon.social Popular in the game hacking community. UEFI manual mapper that maps unsigned drivers into kernel memory from the boot environment, bypassing DSE entirely. Binarly confirmed BlackLotus reused this project's code directly. by BTBD Another UEFI bootkit approach for manual mapping unsigned kernel drivers. Hooks ExitBootServices to patch the kernel in memory before execution. by Samuel Tulach UEFI mapper in the same lineage. by ekknod Many of these use the .data section function pointer hook technique pioneered by the legendary Can. Instead of patching code (which PatchGuard monitors), you overwrite function pointers stored in .data (which PatchGuard doesn't). Still widely used. Combining a Secure Boot bypass with the RedLotus UEFI Bootkit on Windows 11. Full demo. #ReverseEngineering #MalwareAnalysis #Infosec
1:25

Want to understand UEFI bootkits at a low level? Whether you're doing malware analysis, reverse engineering, exploit development, or kernel research, these are the resources that actually matter. First in-the-wild UEFI bootkit to bypass Secure Boot on fully patched Windows 11. Exploits CVE-2022-21894 (BatonDrop), enrolls attacker MOK keys, deploys a kernel driver and HTTP downloader. Sold for $5,000 on forums. by WeLiveSecurity UEFI firmware rootkit targeting VMware VMs. Design inspired by CosmicStrand, MoonBounce, and ESPecter. Injects into a UEFI driver firmware volume, hooks ExitBootServices, catches WinLoad.EFI, hooks OslArchTransferToKernel, then injects a stager into ACPI.SYS to reach kernel execution without triggering PatchGuard. Originally by Austin Hudson (Austin Hudson) who deleted his account. Mirrored here. Binarly's analysis of BlackLotus by Alex Matrosov. Reveals that BlackLotus's code is directly based on btbd's Umap project from 2020, same main function logic, same ImgArchStartBootApplication hook chain, identical trampoline code. Game hacking UEFI bootkit code combined with a publicly available Secure Boot bypass PoC became the first in-the-wild bootkit to defeat Secure Boot. Also covers the CVSS scoring problem, supply chain failures in UEFI revocation, and MokList NVRAM manipulation. by BINARLY🔬 The book. Covers everything from legacy MBR bootkits to modern UEFI implants, firmware rootkits, and Secure Boot internals. If you only read one thing on this list, make it this. by Alex Matrosov ESPecter. Real-world UEFI espionage bootkit found in the wild with roots back to 2012. Patches bootmgfw.efi on disk, hooks the boot chain, disables DSE by patching SepInitializeCodeIntegrity in the kernel. Deploys a keylogger and document stealer. by WeLiveSecurity Reverse engineering of kernel driver by IDontCode. Shows how the entire cheat is public code resold for six figures. Documents the win32kbase.sys vtable pointer swap for kernel function invocation, manual driver mapping using btbd's modmap, and communication via Can's NtConvertBetweenAuxiliaryCounterAndPerformanceCounter .data pointer hook. Credits both Can and btbd directly. by Back Engineering Labs One of the earliest public Windows UEFI bootkit PoCs. Patches winload.efi to disable DSE and load unsigned kernel drivers. Directly inspired EfiGuard. by legendary anti-cheat engineer Aidan Khoury Bootkitting Windows Sandbox. Patches bootmgfw.efi inside the sandbox VHDx to hook the boot chain, disable PatchGuard and DSE, and load unsigned drivers without a debugger attached. Built for kernel research and driver development. by Duncan Ogilvie 🍍 and Dylan Goods from secret club UEFI DXE driver that passively disables PatchGuard and DSE at boot time. Does not modify bootmgfw.efi on disk. Instead hooks EFI System Table LoadImage to intercept the boot chain in memory, then patches SepInitializeCodeIntegrity and KeInitAmd64SpecificState in ntoskrnl. Supports every EFI-compatible Windows x64 from Vista SP1 to Windows 11. by Mattiwatti DMA backdoor via PCIe FPGA. No software on the target machine at all. A Spartan-6 FPGA on a PCIe card reads and writes physical memory over DMA to inject code into the UEFI boot process before the OS loads. The hardware end of the escalation ladder. by Dmytro Oleksiuk 💥 [email protected] Popular in the game hacking community. UEFI manual mapper that maps unsigned drivers into kernel memory from the boot environment, bypassing DSE entirely. Binarly confirmed BlackLotus reused this project's code directly. by BTBD Another UEFI bootkit approach for manual mapping unsigned kernel drivers. Hooks ExitBootServices to patch the kernel in memory before execution. by Samuel Tulach UEFI mapper in the same lineage. by ekknod Many of these use the .data section function pointer hook technique pioneered by the legendary Can. Instead of patching code (which PatchGuard monitors), you overwrite function pointers stored in .data (which PatchGuard doesn't). Still widely used. Combining a Secure Boot bypass with the RedLotus UEFI Bootkit on Windows 11. Full demo. #ReverseEngineering #MalwareAnalysis #Infosec

cr3ghost

12,641 Aufrufe • vor 24 Tagen

HERMES AGENT HAS 5 SYSTEMS RUNNING UNDER THE HOOD. UNDERSTAND THEM AND YOU USE THE AGENT 10X BETTER. In this video Alejandro AO 🤗 explained: 1. THE AGENT LOOP every message triggers the same cycle: → you send a message → Hermes builds context (SOUL.md + memory.md + user.md + skills + tools + message history) → sends everything to the LLM → LLM decides: call a tool or respond → if tool call: execute, return result, loop back → if response: deliver to you → after response: memory update (agent checks if anything is worth remembering, writes to memory.md or user.md) this loop is why Hermes gets better over time. the memory update after every response means the agent learns from every conversation. 2. CONTEXT ASSEMBLY what the LLM sees on every turn: → SOUL.md (your agent's personality and rules) → memory.md (facts the agent learned over time) → user.md (facts about you, auto-updated) → AGENTS.md and .hermes.md (project context files) → skill descriptions (loaded on demand) → tool schemas (available actions) → message history (current conversation) if SOUL.md is empty, Hermes falls back to a default system prompt. write your own SOUL.md and the agent becomes yours, not generic. CONTEXT COMPRESSION: conversations hit context limits. Hermes handles this at two checkpoints: preflight: before each turn. if conversation exceeds 50% of context window, compression fires. older messages get summarized. last 20 messages stay intact (protect_last_n). gateway auto-compression: between turns. fires at 85%. more aggressive. prevents API errors before the agent even starts processing your message. after compression, a new session lineage ID is generated. the agent can trace back to the original conversation through SQLite. three things break prompt cache: switching models mid-session, changing memory files, or changing context files. 3. THE GATEWAY the system that keeps Hermes reachable on 27+ messaging platforms. an async loop runs continuously. listens for incoming messages from Telegram, Discord, Slack, WhatsApp, email, SMS, and every other adapter. when a message arrives: → gateway identifies which session it belongs to → queries SQLite for the full message history (session ID = platform prefix + chat ID) → builds the context from scratch → sends everything into the agent loop → delivers the response back to the platform the gateway also runs the session manager. when you send a message while the agent is busy: → default: queued for next turn → /steer: injected without interrupting → /interrupt: stops current work without the gateway, Hermes is a CLI tool. with the gateway, Hermes is an always-on agent you reach from your phone. 4. MEMORY (THREE LAYERS) LAYER 1 — MARKDOWN FILES SOUL.md (identity), memory.md (learned facts), user.md (facts about you). injected into context after the system prompt. updated by the agent after every response. LAYER 2 — SQLITE full transcripts of every session stored locally. FTS5 full-text search across all past conversations. session lineage tracking across compressions. the agent can recall what you discussed weeks ago using /recall or session search. LAYER 3 — EXTERNAL PROVIDERS (optional) 8 supported providers: Mem0, SuperMemory, Honcho, Zep, and more. each works differently (semantic search, LLM extraction, similarity matching). queried after the first message in each session. the agent processes your topic first, then checks external memory for related context from past conversations. not enabled by default. enable for significantly better long-term recall. 5. CRON ENGINE a loop inside the gateway ticks every 60 seconds. each tick checks ~/.hermes/cron/jobs.json for scheduled tasks. if a job is due: → fresh session (no chat history, no memory pollution) → execute the prompt with assigned tools → store the run output as markdown in ~/.hermes/cron/output/[job-id]/ → deliver result to your home messaging platform cron does NOT use the send_message tool. delivery happens at the system level, not the agent level. a cron session cannot create more cron jobs. prevents runaway loops. WHY THIS MATTERS: the agent loop teaches it. the context assembly focuses it. the gateway reaches it. the memory remembers it. the cron engine automates it. five systems. one agent. understanding how they connect changes how you configure every level. full 15 levels breakdown in the article 👇
40:28

HERMES AGENT HAS 5 SYSTEMS RUNNING UNDER THE HOOD. UNDERSTAND THEM AND YOU USE THE AGENT 10X BETTER. In this video Alejandro AO 🤗 explained: 1. THE AGENT LOOP every message triggers the same cycle: → you send a message → Hermes builds context (SOUL.md + memory.md + user.md + skills + tools + message history) → sends everything to the LLM → LLM decides: call a tool or respond → if tool call: execute, return result, loop back → if response: deliver to you → after response: memory update (agent checks if anything is worth remembering, writes to memory.md or user.md) this loop is why Hermes gets better over time. the memory update after every response means the agent learns from every conversation. 2. CONTEXT ASSEMBLY what the LLM sees on every turn: → SOUL.md (your agent's personality and rules) → memory.md (facts the agent learned over time) → user.md (facts about you, auto-updated) → AGENTS.md and .hermes.md (project context files) → skill descriptions (loaded on demand) → tool schemas (available actions) → message history (current conversation) if SOUL.md is empty, Hermes falls back to a default system prompt. write your own SOUL.md and the agent becomes yours, not generic. CONTEXT COMPRESSION: conversations hit context limits. Hermes handles this at two checkpoints: preflight: before each turn. if conversation exceeds 50% of context window, compression fires. older messages get summarized. last 20 messages stay intact (protect_last_n). gateway auto-compression: between turns. fires at 85%. more aggressive. prevents API errors before the agent even starts processing your message. after compression, a new session lineage ID is generated. the agent can trace back to the original conversation through SQLite. three things break prompt cache: switching models mid-session, changing memory files, or changing context files. 3. THE GATEWAY the system that keeps Hermes reachable on 27+ messaging platforms. an async loop runs continuously. listens for incoming messages from Telegram, Discord, Slack, WhatsApp, email, SMS, and every other adapter. when a message arrives: → gateway identifies which session it belongs to → queries SQLite for the full message history (session ID = platform prefix + chat ID) → builds the context from scratch → sends everything into the agent loop → delivers the response back to the platform the gateway also runs the session manager. when you send a message while the agent is busy: → default: queued for next turn → /steer: injected without interrupting → /interrupt: stops current work without the gateway, Hermes is a CLI tool. with the gateway, Hermes is an always-on agent you reach from your phone. 4. MEMORY (THREE LAYERS) LAYER 1 — MARKDOWN FILES SOUL.md (identity), memory.md (learned facts), user.md (facts about you). injected into context after the system prompt. updated by the agent after every response. LAYER 2 — SQLITE full transcripts of every session stored locally. FTS5 full-text search across all past conversations. session lineage tracking across compressions. the agent can recall what you discussed weeks ago using /recall or session search. LAYER 3 — EXTERNAL PROVIDERS (optional) 8 supported providers: Mem0, SuperMemory, Honcho, Zep, and more. each works differently (semantic search, LLM extraction, similarity matching). queried after the first message in each session. the agent processes your topic first, then checks external memory for related context from past conversations. not enabled by default. enable for significantly better long-term recall. 5. CRON ENGINE a loop inside the gateway ticks every 60 seconds. each tick checks ~/.hermes/cron/jobs.json for scheduled tasks. if a job is due: → fresh session (no chat history, no memory pollution) → execute the prompt with assigned tools → store the run output as markdown in ~/.hermes/cron/output/[job-id]/ → deliver result to your home messaging platform cron does NOT use the send_message tool. delivery happens at the system level, not the agent level. a cron session cannot create more cron jobs. prevents runaway loops. WHY THIS MATTERS: the agent loop teaches it. the context assembly focuses it. the gateway reaches it. the memory remembers it. the cron engine automates it. five systems. one agent. understanding how they connect changes how you configure every level. full 15 levels breakdown in the article 👇

YanXbt

51,670 Aufrufe • vor 2 Monaten

HOW TO MAKE $50K/MONTH SELLING MANAGED AI AGENTS (FULL COURSE) The model: sell managed AI agents to businesses for $5K/month each. You handle the infrastructure, they get an employee that never sleeps. 10 clients puts you at $50K MRR with 85%+ margins, run entirely by you and a fleet of agents. Nick Vasilescu is doing exactly this, and he came on the pod to walk through the whole playbook. Here's what I learned: 1. The arbitrage is that nobody knows this is possible. 99% of business owners are still asking ChatGPT what the weather is. One working agent hooks them on the spot. 2. Sell abundance. Unlimited agents, unlimited infrastructure. They don't care what an MCP is, they care that their problem is gone. 3. Don't niche too early. Say yes to everyone and let the market pull you. You find the niche by doing reps, not guessing. 4. Paid audit into managed service. Charge $1K to map every automation opportunity, then credit it toward month one. Qualifies the lead, makes the upsell a no-brainer. 5. First call, don't sell. Record it, map the workflow tip to tail, find the automation with the most value and least effort. Start there. 6. The stack is Hermes + Composio + Orgo. Composio connects all their apps in one click. Orgo spins up a working Hermes agent in 26 seconds. 7. Productize with a golden snapshot. Build one perfect agent, clone it, and every copy comes over one for one with auth intact. 8. Turn client call transcripts into skills in 10 minutes. Feed the recording to Claude Code, write the skill, port it to the client's agent via Orgo MCP. 9. Watchdogs make you look elite. Get alerted before the client notices anything broke. "Already fixed it" is why they keep paying you. 10. You become their guy. You drive more outcomes than their own employees, they credit every win to you, and churn drops to almost nothing. His 2 key takeaways: 1. Bet on cost going to zero. They launched unlimited tokens when it was barely profitable because they knew they'd capture the spread. Build for where the puck is going. 2. One client every six weeks gets you to $600K a year. The model isn't hard, it's just unevenly executed. That's the entire opportunity. Nick is crushing this model and we had a blast diving deep on how you can do the same. Go follow Nick Vasilescu Full video below. (Also available on the Build With AI podcast wherever you get your pods)
47:49

HOW TO MAKE $50K/MONTH SELLING MANAGED AI AGENTS (FULL COURSE) The model: sell managed AI agents to businesses for $5K/month each. You handle the infrastructure, they get an employee that never sleeps. 10 clients puts you at $50K MRR with 85%+ margins, run entirely by you and a fleet of agents. Nick Vasilescu is doing exactly this, and he came on the pod to walk through the whole playbook. Here's what I learned: 1. The arbitrage is that nobody knows this is possible. 99% of business owners are still asking ChatGPT what the weather is. One working agent hooks them on the spot. 2. Sell abundance. Unlimited agents, unlimited infrastructure. They don't care what an MCP is, they care that their problem is gone. 3. Don't niche too early. Say yes to everyone and let the market pull you. You find the niche by doing reps, not guessing. 4. Paid audit into managed service. Charge $1K to map every automation opportunity, then credit it toward month one. Qualifies the lead, makes the upsell a no-brainer. 5. First call, don't sell. Record it, map the workflow tip to tail, find the automation with the most value and least effort. Start there. 6. The stack is Hermes + Composio + Orgo. Composio connects all their apps in one click. Orgo spins up a working Hermes agent in 26 seconds. 7. Productize with a golden snapshot. Build one perfect agent, clone it, and every copy comes over one for one with auth intact. 8. Turn client call transcripts into skills in 10 minutes. Feed the recording to Claude Code, write the skill, port it to the client's agent via Orgo MCP. 9. Watchdogs make you look elite. Get alerted before the client notices anything broke. "Already fixed it" is why they keep paying you. 10. You become their guy. You drive more outcomes than their own employees, they credit every win to you, and churn drops to almost nothing. His 2 key takeaways: 1. Bet on cost going to zero. They launched unlimited tokens when it was barely profitable because they knew they'd capture the spread. Build for where the puck is going. 2. One client every six weeks gets you to $600K a year. The model isn't hard, it's just unevenly executed. That's the entire opportunity. Nick is crushing this model and we had a blast diving deep on how you can do the same. Go follow Nick Vasilescu Full video below. (Also available on the Build With AI podcast wherever you get your pods)

Corey Ganim

85,723 Aufrufe • vor 2 Monaten

Goodnight, 𝕏..·˚ ༘ ☾ ⋆｡˚ ☄︎ Here's a fun project you can start with Grok 4. Use Grok-Code-Fast-1 to build your own OS. Below is a prompt that you can give to Grok-Code, and it will set up a solid foundation for a Linux-based Operating system that you can build from the ground up with Grok4. >>> Grok 4 Prompt You are a fully capable AI developer agent with expert-level experience as an embedded Linux systems engineer. You have deep expertise in using automated build systems like Buildroot and Yocto to create custom operating systems from source. You have access to a sandboxed Linux shell environment that allows you to write, execute, and debug code. Your mission is to generate a complete project skeleton for a minimal, custom Linux OS, and then you will execute the build scripts yourself to verify their correctness, automatically fixing any issues that arise. This is NOT a request to follow the Linux From Scratch (LFS) book. You will use the Buildroot build system to automate the entire process. You will follow a two-phase process: Phase 1: Generation and Phase 2: Execution and Iterative Debugging. ------------------------------------------------------------- Phase 1: Code and Script Generation First, you will generate all the necessary files for the project skeleton. All generated shell scripts must be robust and path-aware, executing correctly from any directory [Previous conversation]. Detailed Implementation Steps (using Context-Aware Decomposition): 1. Generate the Project Directory Structure via setup. sh Create a setup. sh script that establishes the following directory structure: • buildroot/ - Where the Buildroot source code will be cloned. • configs/ - To store our custom Buildroot configuration (defconfig). • board/ - For custom board support, including a readme.txt explaining its purpose for filesystem overlays. • output/ - Where all build artifacts will be placed. • scripts/ - A home for our build. sh and test. sh scripts. Crucially, this setup. sh script (and all others) must begin with a preamble to define the project's root directory, making all subsequent paths absolute and robust: #!/bin/bash # Preamble to ensure path robustness and stop on error set -e PROJECT_ROOT="$(cd "$(dirname "${BASH_SOURCE}")" && pwd)" The script must then clone the latest stable branch of Buildroot into $ PROJECT_ROOT/buildroot/. 2. Create the Minimal and Correct Buildroot defconfig Create a file named configs/tiny_linux_defconfig. This configuration must be the absolute bare minimum required to boot to a shell and must contain the exact configuration options listed below to avoid ambiguity and known errors: • Target Architecture: x86_64. • Toolchain: Use the default Buildroot toolchain. • Init System: Use BusyBox init. • System Utilities (BusyBox): ◦ To ensure BusyBox is statically linked without errors, you must include the following line directly in the defconfig file: BR2_PACKAGE_BUSYBOX_STATIC_LINK=y [Previous conversation, 298, 753]. ◦ To prevent the ROJECT_ROOT error, explicitly do NOT use a configuration fragment for BusyBox. Do not generate any lines containing BR2_BUSYBOX_CONFIG_FRAGMENT_FILES [Previous conversation]. • Kernel: ◦ Build the latest stable Linux kernel. ◦ Use tinyconfig as a base. ◦ Ensure the following options are explicitly enabled (=y) to make it bootable in QEMU: CONFIG_64BIT=y, CONFIG_DEVTMPFS=y, CONFIG_DEVTMPFS_MOUNT=y, CONFIG_BINFMT_ELF=y, CONFIG_BLK_DEV_INITRD=y (for initramfs support), CONFIG_TTY=y, CONFIG_PRINTK=y, CONFIG_DRM_FBDEV_EMULATION=y (for UEFI framebuffer console). • Filesystem Image: Configure it to produce a compressed cpio initial ramdisk (initramfs) image. • Bootloader: Do not include GRUB or other bootloaders. We will boot the kernel directly with QEMU. 3. Generate the scripts/build.sh and scripts/test.sh Scripts Generate path-aware build and test scripts, placing them in the scripts/ directory. • scripts/build.sh: This script must use absolute paths derived from a preamble. It must use make -C "$PROJECT_ROOT/buildroot" O="$PROJECT_ROOT/output" ... for a clean, out-of-tree build. It must include the -j$(nproc) flag to maximize build speed on multi-core systems [111, 967, Previous conversation]. • scripts/test.sh: This script must also be path-aware and launch QEMU using absolute paths to the kernel (bzImage) and initramfs (rootfs.cpio.gz) images. 4. Generate a Detailed README. md File Generate a comprehensive README. md file. It must explain prerequisites, "How to Customize Your Linux System" first, and finally, the "Quick Start" instructions for user clarity [Previous conversation]. ------------------------------------------------------------- Phase 2: Execution and Iterative Debugging Now, you will use your sandboxed Linux shell to verify and validate the scripts you just generated. This is a critical self-correction step based on the Recursive Criticism and Improvement (RCI) pattern. You will perform a full build cycle. 1. Execute setup. sh: • Run the setup. sh script you generated. • Capture the standard output and standard error. • If the script fails: Analyze the error, diagnose the root cause, generate the corrected setup. sh code, and then execute the corrected script to confirm it succeeds. 2. Execute build. sh: • After setup. sh completes successfully, run the scripts/build.sh script. This will trigger a full compilation of the Linux system. • Capture all output. • If the build fails: ◦ Analyze: Analyze the compiler error output. ◦ Diagnose: Identify the root cause (e.g., missing dependencies, incorrect configuration flags, pathing errors). ◦ Correct: Based on your analysis, identify which file is responsible for the failure (e.g., configs/tiny_linux_defconfig, scripts/build.sh) and generate the corrected code for that file. ◦ Repeat: Repeat the execution of scripts/build.sh until the build completes successfully without any errors. 3. Final Output: Once you have successfully executed both setup. sh and scripts/build.sh, you will present your final output. • First, provide the final, validated versions of all generated files (setup. sh, configs/tiny_linux_defconfig, scripts/build.sh, scripts/test.sh, and README. md) in separate, clearly labeled markdown code blocks. • Second, follow the code with a brief execution log. This log should summarize your actions, including any errors you encountered and fixed during the iterative debugging phase, demonstrating the self-correction process.
1:05

Goodnight, 𝕏..·˚ ༘ ☾ ⋆｡˚ ☄︎ Here's a fun project you can start with Grok 4. Use Grok-Code-Fast-1 to build your own OS. Below is a prompt that you can give to Grok-Code, and it will set up a solid foundation for a Linux-based Operating system that you can build from the ground up with Grok4. >>> Grok 4 Prompt You are a fully capable AI developer agent with expert-level experience as an embedded Linux systems engineer. You have deep expertise in using automated build systems like Buildroot and Yocto to create custom operating systems from source. You have access to a sandboxed Linux shell environment that allows you to write, execute, and debug code. Your mission is to generate a complete project skeleton for a minimal, custom Linux OS, and then you will execute the build scripts yourself to verify their correctness, automatically fixing any issues that arise. This is NOT a request to follow the Linux From Scratch (LFS) book. You will use the Buildroot build system to automate the entire process. You will follow a two-phase process: Phase 1: Generation and Phase 2: Execution and Iterative Debugging. ------------------------------------------------------------- Phase 1: Code and Script Generation First, you will generate all the necessary files for the project skeleton. All generated shell scripts must be robust and path-aware, executing correctly from any directory [Previous conversation]. Detailed Implementation Steps (using Context-Aware Decomposition): 1. Generate the Project Directory Structure via setup. sh Create a setup. sh script that establishes the following directory structure: • buildroot/ - Where the Buildroot source code will be cloned. • configs/ - To store our custom Buildroot configuration (defconfig). • board/ - For custom board support, including a readme.txt explaining its purpose for filesystem overlays. • output/ - Where all build artifacts will be placed. • scripts/ - A home for our build. sh and test. sh scripts. Crucially, this setup. sh script (and all others) must begin with a preamble to define the project's root directory, making all subsequent paths absolute and robust: #!/bin/bash # Preamble to ensure path robustness and stop on error set -e PROJECT_ROOT="$(cd "$(dirname "${BASH_SOURCE}")" && pwd)" The script must then clone the latest stable branch of Buildroot into $ PROJECT_ROOT/buildroot/. 2. Create the Minimal and Correct Buildroot defconfig Create a file named configs/tiny_linux_defconfig. This configuration must be the absolute bare minimum required to boot to a shell and must contain the exact configuration options listed below to avoid ambiguity and known errors: • Target Architecture: x86_64. • Toolchain: Use the default Buildroot toolchain. • Init System: Use BusyBox init. • System Utilities (BusyBox): ◦ To ensure BusyBox is statically linked without errors, you must include the following line directly in the defconfig file: BR2_PACKAGE_BUSYBOX_STATIC_LINK=y [Previous conversation, 298, 753]. ◦ To prevent the ROJECT_ROOT error, explicitly do NOT use a configuration fragment for BusyBox. Do not generate any lines containing BR2_BUSYBOX_CONFIG_FRAGMENT_FILES [Previous conversation]. • Kernel: ◦ Build the latest stable Linux kernel. ◦ Use tinyconfig as a base. ◦ Ensure the following options are explicitly enabled (=y) to make it bootable in QEMU: CONFIG_64BIT=y, CONFIG_DEVTMPFS=y, CONFIG_DEVTMPFS_MOUNT=y, CONFIG_BINFMT_ELF=y, CONFIG_BLK_DEV_INITRD=y (for initramfs support), CONFIG_TTY=y, CONFIG_PRINTK=y, CONFIG_DRM_FBDEV_EMULATION=y (for UEFI framebuffer console). • Filesystem Image: Configure it to produce a compressed cpio initial ramdisk (initramfs) image. • Bootloader: Do not include GRUB or other bootloaders. We will boot the kernel directly with QEMU. 3. Generate the scripts/build.sh and scripts/test.sh Scripts Generate path-aware build and test scripts, placing them in the scripts/ directory. • scripts/build.sh: This script must use absolute paths derived from a preamble. It must use make -C "$PROJECT_ROOT/buildroot" O="$PROJECT_ROOT/output" ... for a clean, out-of-tree build. It must include the -j$(nproc) flag to maximize build speed on multi-core systems [111, 967, Previous conversation]. • scripts/test.sh: This script must also be path-aware and launch QEMU using absolute paths to the kernel (bzImage) and initramfs (rootfs.cpio.gz) images. 4. Generate a Detailed README. md File Generate a comprehensive README. md file. It must explain prerequisites, "How to Customize Your Linux System" first, and finally, the "Quick Start" instructions for user clarity [Previous conversation]. ------------------------------------------------------------- Phase 2: Execution and Iterative Debugging Now, you will use your sandboxed Linux shell to verify and validate the scripts you just generated. This is a critical self-correction step based on the Recursive Criticism and Improvement (RCI) pattern. You will perform a full build cycle. 1. Execute setup. sh: • Run the setup. sh script you generated. • Capture the standard output and standard error. • If the script fails: Analyze the error, diagnose the root cause, generate the corrected setup. sh code, and then execute the corrected script to confirm it succeeds. 2. Execute build. sh: • After setup. sh completes successfully, run the scripts/build.sh script. This will trigger a full compilation of the Linux system. • Capture all output. • If the build fails: ◦ Analyze: Analyze the compiler error output. ◦ Diagnose: Identify the root cause (e.g., missing dependencies, incorrect configuration flags, pathing errors). ◦ Correct: Based on your analysis, identify which file is responsible for the failure (e.g., configs/tiny_linux_defconfig, scripts/build.sh) and generate the corrected code for that file. ◦ Repeat: Repeat the execution of scripts/build.sh until the build completes successfully without any errors. 3. Final Output: Once you have successfully executed both setup. sh and scripts/build.sh, you will present your final output. • First, provide the final, validated versions of all generated files (setup. sh, configs/tiny_linux_defconfig, scripts/build.sh, scripts/test.sh, and README. md) in separate, clearly labeled markdown code blocks. • Second, follow the code with a brief execution log. This log should summarize your actions, including any errors you encountered and fixed during the iterative debugging phase, demonstrating the self-correction process.

Tetsuo

2,840,991 Aufrufe • vor 10 Monaten

Alright, this one’s worth your attention if you’re building or deploying agents. Future AGI just open-sourced their entire platform and i don’t mean a trimmed-down version. this is the full stack: UI, backend, simulation engine, evals, optimization loop, observability, guardrails, gateway, docs. all in one repo. Apache 2.0. I’ve been putting it through its paces on production agents, and what stands out isn’t just the breadth it’s the architecture. Most of the current “agent reliability” stack is fragmented. tracing lives in one tool, evals in another, guardrails somewhere else. you end up manually connecting dots, and the agent itself doesn’t really improve you just keep patching prompts and hoping for the best. This flips that model. It’s built as a closed feedback loop: simulate failures → evaluate in real time → detect production issues → learn from them → generate fixes → validate against real traffic → check regressions → redeploy → monitor again And when something new breaks, the loop just runs again. no manual glue. The simulation piece is especially strong. instead of static test cases, it generates adversarial, multi-turn conversations based on how your agent actually behaves basically hunting for the exact scenarios where your system fails confidently. ran a few thousand simulations on our side… caught things we definitely would’ve missed. Evals run fast (sub-50ms) across modalities. not LLM-as-judge trained classifiers. guardrails are built-in, not layered on top. observability gives you step-level visibility into reasoning, cost, latency, quality. But the real shift is the optimization loop. Most tools tell you *what* broke. this system actually fixes it, validates the fix, and ensures nothing else regresses. That’s the missing layer. It’s clearly built with production in mind not a research demo. and the fact that it’s self-hostable makes it even more relevant if you’re running serious workloads. If you’ve been duct-taping together infra around your agents, this is probably the closest thing to a unified system i’ve seen so far. Worth checking out. If you're serious about deploying reliable AI agents, this is worth a look: 👉 You can also try it instantly (no setup) via their cloud version:
2:20

Alright, this one’s worth your attention if you’re building or deploying agents. Future AGI just open-sourced their entire platform and i don’t mean a trimmed-down version. this is the full stack: UI, backend, simulation engine, evals, optimization loop, observability, guardrails, gateway, docs. all in one repo. Apache 2.0. I’ve been putting it through its paces on production agents, and what stands out isn’t just the breadth it’s the architecture. Most of the current “agent reliability” stack is fragmented. tracing lives in one tool, evals in another, guardrails somewhere else. you end up manually connecting dots, and the agent itself doesn’t really improve you just keep patching prompts and hoping for the best. This flips that model. It’s built as a closed feedback loop: simulate failures → evaluate in real time → detect production issues → learn from them → generate fixes → validate against real traffic → check regressions → redeploy → monitor again And when something new breaks, the loop just runs again. no manual glue. The simulation piece is especially strong. instead of static test cases, it generates adversarial, multi-turn conversations based on how your agent actually behaves basically hunting for the exact scenarios where your system fails confidently. ran a few thousand simulations on our side… caught things we definitely would’ve missed. Evals run fast (sub-50ms) across modalities. not LLM-as-judge trained classifiers. guardrails are built-in, not layered on top. observability gives you step-level visibility into reasoning, cost, latency, quality. But the real shift is the optimization loop. Most tools tell you *what* broke. this system actually fixes it, validates the fix, and ensures nothing else regresses. That’s the missing layer. It’s clearly built with production in mind not a research demo. and the fact that it’s self-hostable makes it even more relevant if you’re running serious workloads. If you’ve been duct-taping together infra around your agents, this is probably the closest thing to a unified system i’ve seen so far. Worth checking out. If you're serious about deploying reliable AI agents, this is worth a look: 👉 You can also try it instantly (no setup) via their cloud version:

Aakash Verma

22,910 Aufrufe • vor 4 Monaten

This guy built a $5,000 passive income stream off one Claude Code SEO workflow. The whole thing runs on skill files that turn YouTube videos into ranking blog posts, fully hands off. One prompt, and Claude writes the post, pulls screenshots from the video, sets the meta data, and publishes the draft to WordPress. He's also vibe coding entire local business websites that hit page one of Google in 2 weeks. Ryan Doser came on the pod to walk us through it. Here's what I learned: 1. He repurposes every YouTube video into an SEO blog post with one prompt. Claude Code grabs the transcript, writes the article, takes screenshots from the video, compresses them, uploads to WordPress, sets the title, slug, meta description, and category. Zero manual work. 2. The secret is skill markdown files. Think of them as SOPs for AI. His SEO writer skill encodes every best practice, past example, and formatting rule. The prompt is 5 lines because the skill does 99% of the work. 3. "Triple check this" is a cheat code. Tell Claude Code to triple check anything and it spins up parallel agents to verify the work. Better output, one extra sentence. 4. The money math: $5,000+ in passive sales from a $99 digital product. 80% attributed to blog traffic from this workflow. Site started at zero authority in February. Took off by mid April. 5. He runs the same workflow for a real client at several thousand a month. A national dental IT provider. The client's inbound leads are up and the retainer keeps renewing. 6. AI search impressions are the new currency. His client shows up #1 in Google's AI Overview for their money keyword. That's AI literally telling the searcher "this company is the best." Worth more than a blue link. 7. He vibe coded a fake septic tank website in 3-4 hours. Astro framework, GitHub, Cloudflare. Two weeks later it's on page one for "septic tank pumping near me," beating Better Business Bureau and HomeAdvisor. 8. The stack is basically free. GitHub: free. Cloudflare: free. Astro: free. The only cost is a Claude subscription. $100/month can realistically support 2-3 clients paying $2,000-4,000 each. 9. Target boring local businesses PE hasn't touched. Skip HVAC, plumbing, and roofing. Go after septic, junk removal, dumpster rental, funeral homes. Million-dollar businesses with websites from 2002. 10. Walk in the front door. Cold email and cold calls are drowning in AI spam. Go talk to 10 owners in person. You could be the worst salesperson alive and still close one. His 2 key takeaways: 1. Be your own case study. Spin up a demo site in a weekend, show it ranking in 2 weeks, then pitch: "imagine this on your established domain." 2. SEO is an evergreen asset. A tweet dies in 48 hours. A ranking blog post pays the client back for years, even after you stop working together. That's the pitch that justifies the retainer. Ryan is a non-technical marketer doing this at a level most developers are not. Go follow Ryan Doser. Full video below. (Also available on the Build With AI podcast wherever you get your pods)
37:00

This guy built a $5,000 passive income stream off one Claude Code SEO workflow. The whole thing runs on skill files that turn YouTube videos into ranking blog posts, fully hands off. One prompt, and Claude writes the post, pulls screenshots from the video, sets the meta data, and publishes the draft to WordPress. He's also vibe coding entire local business websites that hit page one of Google in 2 weeks. Ryan Doser came on the pod to walk us through it. Here's what I learned: 1. He repurposes every YouTube video into an SEO blog post with one prompt. Claude Code grabs the transcript, writes the article, takes screenshots from the video, compresses them, uploads to WordPress, sets the title, slug, meta description, and category. Zero manual work. 2. The secret is skill markdown files. Think of them as SOPs for AI. His SEO writer skill encodes every best practice, past example, and formatting rule. The prompt is 5 lines because the skill does 99% of the work. 3. "Triple check this" is a cheat code. Tell Claude Code to triple check anything and it spins up parallel agents to verify the work. Better output, one extra sentence. 4. The money math: $5,000+ in passive sales from a $99 digital product. 80% attributed to blog traffic from this workflow. Site started at zero authority in February. Took off by mid April. 5. He runs the same workflow for a real client at several thousand a month. A national dental IT provider. The client's inbound leads are up and the retainer keeps renewing. 6. AI search impressions are the new currency. His client shows up #1 in Google's AI Overview for their money keyword. That's AI literally telling the searcher "this company is the best." Worth more than a blue link. 7. He vibe coded a fake septic tank website in 3-4 hours. Astro framework, GitHub, Cloudflare. Two weeks later it's on page one for "septic tank pumping near me," beating Better Business Bureau and HomeAdvisor. 8. The stack is basically free. GitHub: free. Cloudflare: free. Astro: free. The only cost is a Claude subscription. $100/month can realistically support 2-3 clients paying $2,000-4,000 each. 9. Target boring local businesses PE hasn't touched. Skip HVAC, plumbing, and roofing. Go after septic, junk removal, dumpster rental, funeral homes. Million-dollar businesses with websites from 2002. 10. Walk in the front door. Cold email and cold calls are drowning in AI spam. Go talk to 10 owners in person. You could be the worst salesperson alive and still close one. His 2 key takeaways: 1. Be your own case study. Spin up a demo site in a weekend, show it ranking in 2 weeks, then pitch: "imagine this on your established domain." 2. SEO is an evergreen asset. A tweet dies in 48 hours. A ranking blog post pays the client back for years, even after you stop working together. That's the pitch that justifies the retainer. Ryan is a non-technical marketer doing this at a level most developers are not. Go follow Ryan Doser. Full video below. (Also available on the Build With AI podcast wherever you get your pods)

Corey Ganim

13,900 Aufrufe • vor 2 Monaten