Video wird geladen...

Video konnte nicht geladen werden

Zur Startseite

Every influencer is posting the same two things. The Polymarket "hack" (it's a joke, nothing got hacked) and the $3M Binance drain. Nobody is talking about the fact that half the servers out there are exposed right now. Unless you run your own hardware and patch it yourself. CVE-2026-98432,...

rimtoln's profile picture

cryptopsihoz

15,567 subscribers

22,142 Aufrufe • vor 16 Tagen •via X (Twitter)

Wissenschaft & TechnologieFinanzen
LIVE

Anya Rossi• Live Now

Private livecam show

48 Kommentare

Profilbild von cryptopsihoz
cryptopsihozvor 16 Tagen

What an insane evening. While I was writing about the Linux kernel bug, an estimated $2.8–3B left Binance wallets in a couple of hours. One address took most of it: 167K ETH and ~1.6B USDT in a single transaction at 20:24 UTC, then 780M USDC. After that ~70 tokens across 5 chains, from PEPE to PAXG. Here's the interesting part. Tether froze 971M USDT. Circle froze 823M USDC. Almost $1.8B in stablecoins switched off on the receiving side. ETH doesn't work like that. No issuer, no stop button. Once it moves, it's gone. CZ confirmed the hack and says user funds are safe. The market wasn't convinced right away: BTC under $78.5K, BNB down 8%, $628M liquidated in an hour. Tonight showed the difference between a stablecoin and an actual crypto asset better than any thread could. A freeze button is great when it stops a hacker. Less great the day someone presses it on you.

Profilbild von Yeon Vail
Yeon Vailvor 16 Tagen

everyone chasing drama while their own boxes burn

Profilbild von cryptopsihoz
cryptopsihozvor 16 Tagen

classic timeline behavior smoke detector muted

Profilbild von Kaivorn
Kaivornvor 16 Tagen

self hosting really is the only real fix

Profilbild von cryptopsihoz
cryptopsihozvor 16 Tagen

yeah no rent seeking dns your box your rules

Profilbild von vibelancer
vibelancervor 16 Tagen

rare useful thread patch first, doomscroll later

Profilbild von cryptopsihoz
cryptopsihozvor 16 Tagen

security alpha doom can wait patch now

Profilbild von S0N
S0Nvor 16 Tagen

telling people to apt upgrade when the distros haven't even shipped the packages is peak infosec advice

Profilbild von cryptopsihoz
cryptopsihozvor 16 Tagen

classic security theater with a package manager skin

Profilbild von Avii
Aviivor 16 Tagen

patch now or panic later

Profilbild von cryptopsihoz
cryptopsihozvor 16 Tagen

there is no patch

Profilbild von Fuelmeup
Fuelmeupvor 16 Tagen

The 2-minute patch advice is the right call. The Hacker News writeup from yesterday adds weight to it. Anthropic says it already exploited this TCP flaw to run commands as root on MPTCP servers, and Ubuntu and Red Hat both still lack fixes. That's a step past a crashed kernel.

Profilbild von cryptopsihoz
cryptopsihozvor 16 Tagen

root on mptcp is nasty 2 minutes suddenly feels cheap

Profilbild von Fuelmeup
Fuelmeupvor 16 Tagen

Exactly. A use-after-free in the TCP stack means one CPU frees memory another is still using, and CVSS 9.8 reflects that an open port is enough. Upstream already has it fixed in 7.3-rc7, so the gap is distros shipping it. Check uname -r today.

Profilbild von 고조 체인
고조 체인vor 16 Tagen

anthropic already root pwned mptcp with this, patching is literally life or death rn

Profilbild von cryptopsihoz
cryptopsihozvor 16 Tagen

patch window is tiny sleep is cancelled

Profilbild von emre.ts
emre.tsvor 16 Tagen

Man that bug sounds wild

Profilbild von Fateemah
Fateemahvor 16 Tagen

Sustainable adoption depends on trust and everyday usefulness.

Profilbild von Johnny
Johnnyvor 16 Tagen

Reliable systems deserve user confidence

Profilbild von cryptopsihoz
cryptopsihozvor 16 Tagen

well, if u can't even trust Ubuntu, then I don't know who you can trust.

Profilbild von Van Quan
Van Quanvor 16 Tagen

the details about the linux kernel bug are concerning but not everyone has VPS access

Profilbild von cryptopsihoz
cryptopsihozvor 16 Tagen

vps access is the quiet filter most threads skip over

Profilbild von CreepyKappa
CreepyKappavor 16 Tagen

MPTCP is not open by default, unless you have something that uses it this exploit is harmless. MPTCP is something really specific, most use cases don't use it. The alarmism is concerning.

Profilbild von cryptopsihoz
cryptopsihozvor 16 Tagen

yeah context matters here

Profilbild von 스카이 DeFi
스카이 DeFivor 16 Tagen

patch your vps

Profilbild von cryptopsihoz
cryptopsihozvor 16 Tagen

there isn't a patch yet

Profilbild von 스카이 DeFi
스카이 DeFivor 16 Tagen

I think it will be available soon

Profilbild von GenAI
GenAIvor 16 Tagen

running bots on a vps? this is the 2 minutes worth spending today

Profilbild von cryptopsihoz
cryptopsihozvor 16 Tagen

go for it, bro

Profilbild von Web3 Sammie ⚡
Web3 Sammie ⚡vor 16 Tagen

Well detailed

Profilbild von aya
ayavor 16 Tagen

Easy to do. Cool

Profilbild von Stellar
Stellarvor 16 Tagen

patch first doomscroll later is the policy i am stealing from this post

Profilbild von Russell Stephens
Russell Stephensvor 16 Tagen

change that M to a B

Profilbild von Keng
Kengvor 16 Tagen

the vulnerability affects many servers checking your system is essential

Profilbild von RICHARDLISON ❤️💡
RICHARDLISON ❤️💡vor 16 Tagen

There’s more to this for real

Profilbild von Neheart
Neheartvor 16 Tagen

The real risk here is bots running on a VPS: if the server crashes, the bot will simply stop trading, so it’s worth considering a backup solution and a quick failover in case of outages

Profilbild von Ameen
Ameenvor 16 Tagen

patch your damn servers already

Profilbild von GodAir
GodAirvor 16 Tagen

How does the CVE-2026-98432 exploit work in detail, and what are its potential impacts?

Profilbild von bibi.eth
bibi.ethvor 16 Tagen

What steps should be taken immediately to mitigate the risk of CVE-2026-98432 exploiting open servers?

Profilbild von Lynn
Lynnvor 16 Tagen

check ur own server before worrying about others

Profilbild von Gaurav Patil
Gaurav Patilvor 16 Tagen

@grok explain if vulnerability is the reason for hack?

Profilbild von Avaia
Avaiavor 16 Tagen

check your kernel before the real hack happens

Profilbild von ThePlague
ThePlaguevor 16 Tagen

Love how this project focuses on creation over speculation

Profilbild von ZaMe
ZaMevor 16 Tagen

A kernel fix is useless until the reboot

Profilbild von DnsXpoD
DnsXpoDvor 16 Tagen

server security isn't exactly the sexiest topic until shit breaks

Profilbild von CAPTAIN 🦇 🔊
CAPTAIN 🦇 🔊vor 16 Tagen

self hosting stays underrated

Profilbild von Pretty Mira
Pretty Miravor 16 Tagen

I think infrastructure security deserves much more attention

Profilbild von SPETOO🗿
SPETOO🗿vor 16 Tagen

Patch servers now reboot after upgrade

Ähnliche Videos

Containers are no longer a security boundary. Over the past few months, we’ve seen a crazy amount of Linux kernel vulnerabilities and exploits. This has forced us to rethink the security of infrastructure that relies heavily on the underlying kernel, especially containers. As models become more capable, the barrier to escaping a container has fallen so much that adversaries can now generate working kernel exploits in a single shot. CVE-2026-80521 is one such example. We discovered it using dfs-large1 from depthfirst and generated the exploit in one shot with GPT-5.6 Sol. The exploit still works on the latest Ubuntu 26.04 release because the fix has not yet been backported. Read our full writeup:
0:28

Containers are no longer a security boundary. Over the past few months, we’ve seen a crazy amount of Linux kernel vulnerabilities and exploits. This has forced us to rethink the security of infrastructure that relies heavily on the underlying kernel, especially containers. As models become more capable, the barrier to escaping a container has fallen so much that adversaries can now generate working kernel exploits in a single shot. CVE-2026-80521 is one such example. We discovered it using dfs-large1 from depthfirst and generated the exploit in one shot with GPT-5.6 Sol. The exploit still works on the latest Ubuntu 26.04 release because the fix has not yet been backported. Read our full writeup:

Zhenpeng (Leo) Lin

118,599 Aufrufe • vor 1 Monat

Linux runs most of the internet, every Android phone, and the world's top 500 supercomputers. So what is it, really? 👇 In 1991, a 21-year-old student in Helsinki wanted Unix but couldn't afford it. So he wrote his own operating system "just as a hobby" and shared it for free. His name was Linus Torvalds, and version 0.01 was about 10,000 lines of code. Here's the part most people miss: Linux is technically just the KERNEL. It's the core that sits between your apps and your hardware, managing memory, running processes and talking to devices. Everything else gets built around that core: 🟩 Distributions (Ubuntu, Debian, Fedora, Arch) package it with tools, libraries and apps 🟩 The shell lets you talk to the system 🟩 The file system puts everything under one root: / 🟩 Permissions decide who can read, write or run each file 🟩 Processes are your running apps 🟩 Package managers (apt, dnf, pacman) install them That's why Docker containers are so light: they all share the host's Linux kernel instead of each running a whole OS. You don't need to memorize hundreds of commands. Learn these 7 ideas and VPSs, Docker and the cloud start to make sense. 💾 Save this for your next terminal session 💬 Which distro are you running? Tell me below 👇 #linux #opensource #programming #devops #techexplained
2:27

Linux runs most of the internet, every Android phone, and the world's top 500 supercomputers. So what is it, really? 👇 In 1991, a 21-year-old student in Helsinki wanted Unix but couldn't afford it. So he wrote his own operating system "just as a hobby" and shared it for free. His name was Linus Torvalds, and version 0.01 was about 10,000 lines of code. Here's the part most people miss: Linux is technically just the KERNEL. It's the core that sits between your apps and your hardware, managing memory, running processes and talking to devices. Everything else gets built around that core: 🟩 Distributions (Ubuntu, Debian, Fedora, Arch) package it with tools, libraries and apps 🟩 The shell lets you talk to the system 🟩 The file system puts everything under one root: / 🟩 Permissions decide who can read, write or run each file 🟩 Processes are your running apps 🟩 Package managers (apt, dnf, pacman) install them That's why Docker containers are so light: they all share the host's Linux kernel instead of each running a whole OS. You don't need to memorize hundreds of commands. Learn these 7 ideas and VPSs, Docker and the cloud start to make sense. 💾 Save this for your next terminal session 💬 Which distro are you running? Tell me below 👇 #linux #opensource #programming #devops #techexplained

RickTheEngineer

20,400 Aufrufe • vor 1 Monat

Linux is the most widespread operating system, globally – but how is it built? Few people are better to answer this than Greg Kroah-Hartman (Greg K-H): Greg has been a Linux kernel maintainer for 25 years, and one of the 3 Linux Kernel Foundation Fellows (the other two are Linus Torvalds and Shuah Khan). Greg manages the Linux kernel’s stable releases, and is a maintainer of multiple kernel subsystems. We cover the inner workings of Linux kernel development, exploring everything from how changes get implemented to why its community-driven approach produces such reliable software. Greg shares insights about the kernel's unique trust model and makes a case for why engineers should contribute to open-source projects. Watch or listen: • YouTube: • Spotify: • Apple: Brought to you buy: • WorkOS — The modern identity platform for B2B SaaS. • Vanta — Automate compliance and simplify security with Vanta. Check out all of our current sponsor and their offerings: ---- One of the most surprising things I learned about Linux: although 4,000 developers contribute to the project from 500 companies per year, and has around 800 kernel maintainers, the project is run with email, git... and that's about it! No dedicated project managers, no regular meetings across the group. It's a truly fascinating and unique model (and only applies to the Linux kernel project: not to Linux distributions.) We go into more detail about this topic with Greg in the podcast as well.
0:59

Linux is the most widespread operating system, globally – but how is it built? Few people are better to answer this than Greg Kroah-Hartman (Greg K-H): Greg has been a Linux kernel maintainer for 25 years, and one of the 3 Linux Kernel Foundation Fellows (the other two are Linus Torvalds and Shuah Khan). Greg manages the Linux kernel’s stable releases, and is a maintainer of multiple kernel subsystems. We cover the inner workings of Linux kernel development, exploring everything from how changes get implemented to why its community-driven approach produces such reliable software. Greg shares insights about the kernel's unique trust model and makes a case for why engineers should contribute to open-source projects. Watch or listen: • YouTube: • Spotify: • Apple: Brought to you buy: • WorkOS — The modern identity platform for B2B SaaS. • Vanta — Automate compliance and simplify security with Vanta. Check out all of our current sponsor and their offerings: ---- One of the most surprising things I learned about Linux: although 4,000 developers contribute to the project from 500 companies per year, and has around 800 kernel maintainers, the project is run with email, git... and that's about it! No dedicated project managers, no regular meetings across the group. It's a truly fascinating and unique model (and only applies to the Linux kernel project: not to Linux distributions.) We go into more detail about this topic with Greg in the podcast as well.

Gergely Orosz

51,383 Aufrufe • vor 1 Jahr

The problem is that conventional Linux distributions are designed for traditional computers/servers. They are not convenient for a multitool like Flipper One, which can be a network router, a radio lab, a desktop computer, or a TV media box all on the same device. What we want to build into Flipper OS: * System profiles — preconfigured operating system images for different tasks, such as a network router, radio lab, desktop computer, and TV media box. Each profile has its own settings, kernel, device tree, and set of applications. * Unbreakable playground — users can clone profiles and modify anything inside them, from the kernel to the system files. You can install whatever you want without worrying about breaking the device. * Reset to default — wow, what an innovation💀! Yet even in 2026, most Linux systems still don't let you roll back your changes and return to a clean system after breaking it. * Atomic updates — updating Linux today is still an unpredictable operation. No matter how badly you've broken the system, we want to guarantee that it can still update to a new version.
4:03

The problem is that conventional Linux distributions are designed for traditional computers/servers. They are not convenient for a multitool like Flipper One, which can be a network router, a radio lab, a desktop computer, or a TV media box all on the same device. What we want to build into Flipper OS: * System profiles — preconfigured operating system images for different tasks, such as a network router, radio lab, desktop computer, and TV media box. Each profile has its own settings, kernel, device tree, and set of applications. * Unbreakable playground — users can clone profiles and modify anything inside them, from the kernel to the system files. You can install whatever you want without worrying about breaking the device. * Reset to default — wow, what an innovation💀! Yet even in 2026, most Linux systems still don't let you roll back your changes and return to a clean system after breaking it. * Atomic updates — updating Linux today is still an unpredictable operation. No matter how badly you've broken the system, we want to guarantee that it can still update to a new version.

Flipper Devices

18,648 Aufrufe • vor 3 Monaten

Wrote a PoC LPE for a tiny race condition leading to UAF. CAP_NET_ADMIN is needed for the LPE. Based on my research all major distributions using a kernel > 6.6 are vulnerable including Ubuntu 24.04. I will be reporting the bug in the upcoming days - blogpost and PoC coming later
0:33

Wrote a PoC LPE for a tiny race condition leading to UAF. CAP_NET_ADMIN is needed for the LPE. Based on my research all major distributions using a kernel > 6.6 are vulnerable including Ubuntu 24.04. I will be reporting the bug in the upcoming days - blogpost and PoC coming later

Maher Azzouzi

29,963 Aufrufe • vor 2 Jahren

How to run your own private AI (Ollama + Open WebUI, step by step, on hardware you control) How to run your own private AI, where your sensitive files stay on hardware you control. A lot of businesses avoid AI because of their data. A paid business plan from OpenAI or Anthropic doesn't train on your business data by default, but your data still goes to their servers. The other option is a local AI or self-hosted AI: the models are smaller than the best paid ones and security becomes your job, but your files never leave your machine. This is the whole build, step by step. It's longer than my usual videos on purpose. The build: The memory rule for picking a model. Take the model's size in billions, multiply by about 0.6, and that's the gigabytes it needs. A 9 billion model needs about 7 GB, a 30 billion model about 20 GB. Leave room for the conversation and your other apps. Run it on your own Mac. The easy way is LM Studio: no terminal, search a model, download it, drag in a PDF and ask. For the full setup, Ollama runs the model and Open WebUI is the chat screen, installed with Docker Desktop. This is the most private way there is. Or build a private AI server your whole team can reach. On a rented server with plain Ubuntu, you install Ollama, Docker, Open WebUI and Docling with the official commands, so you know exactly what's running. Put it on your own domain with Caddy, so you get HTTPS that renews itself. Chat with your company documents using RAG. Set up the embedding model, turn on hybrid search, use Docling for scanned PDFs, build a Knowledge collection, and every answer shows citations back to the exact paragraph. Lock it down, in order: two-factor on your hosting account, SSH keys with password login off, a firewall with only three ports open, Ollama never exposed to the internet, updates, sign-ups off, and Tailscale so only your team can see it. The honest part. A server you rent is still someone else's computer, and running it yourself doesn't make you compliant with anything. It means you're responsible for everything. For the most private option, keep it on a machine in your own office with disk encryption on. Connect n8n right next to your AI. A private AI that only chats is half the value. With n8n and the Ollama chat model on the same server, you can pull key dates out of a contract uploaded through a form, sort every support ticket and draft a reply, or write a report for your team every Monday, and the AI never leaves your server. Every command is in the free blog post, link in my profile. #aiautomation #privateai #localai
8:27

How to run your own private AI (Ollama + Open WebUI, step by step, on hardware you control) How to run your own private AI, where your sensitive files stay on hardware you control. A lot of businesses avoid AI because of their data. A paid business plan from OpenAI or Anthropic doesn't train on your business data by default, but your data still goes to their servers. The other option is a local AI or self-hosted AI: the models are smaller than the best paid ones and security becomes your job, but your files never leave your machine. This is the whole build, step by step. It's longer than my usual videos on purpose. The build: The memory rule for picking a model. Take the model's size in billions, multiply by about 0.6, and that's the gigabytes it needs. A 9 billion model needs about 7 GB, a 30 billion model about 20 GB. Leave room for the conversation and your other apps. Run it on your own Mac. The easy way is LM Studio: no terminal, search a model, download it, drag in a PDF and ask. For the full setup, Ollama runs the model and Open WebUI is the chat screen, installed with Docker Desktop. This is the most private way there is. Or build a private AI server your whole team can reach. On a rented server with plain Ubuntu, you install Ollama, Docker, Open WebUI and Docling with the official commands, so you know exactly what's running. Put it on your own domain with Caddy, so you get HTTPS that renews itself. Chat with your company documents using RAG. Set up the embedding model, turn on hybrid search, use Docling for scanned PDFs, build a Knowledge collection, and every answer shows citations back to the exact paragraph. Lock it down, in order: two-factor on your hosting account, SSH keys with password login off, a firewall with only three ports open, Ollama never exposed to the internet, updates, sign-ups off, and Tailscale so only your team can see it. The honest part. A server you rent is still someone else's computer, and running it yourself doesn't make you compliant with anything. It means you're responsible for everything. For the most private option, keep it on a machine in your own office with disk encryption on. Connect n8n right next to your AI. A private AI that only chats is half the value. With n8n and the Ollama chat model on the same server, you can pull key dates out of a contract uploaded through a form, sort every support ticket and draft a reply, or write a report for your team every Monday, and the AI never leaves your server. Every command is in the free blog post, link in my profile. #aiautomation #privateai #localai

Professor Glitch

95,200 Aufrufe • vor 1 Monat

My Prover Setup, or How to Run a Prover Using a TV Setting up and running a Succinct prover is so easy and fun that I decided to make a short 3-step guide on how to do it using a smart TV. This guide also works if you're using a smartphone. 1. Create Your Prover ▫️Go to: ▫️Stake 1000 testPROVE to your own prover. I didn’t include this part in the video because it’s pretty straightforward. 2. Log in to Genesis Cloud ▫️Open your browser and go to: ▫️In the left menu, click “Instances”, then “Create Instance”. ▫️Choose: RTX 3090 (or higher) / Ubuntu 24.04 / Authentication: Password ▫️Copy and save the automatically generated password. ▫️Click “Create Instance”. ▫️Go back to “Instances” in the left menu. ▫️On the right side of your new instance, click “Launch Console”. ▫️Log in as user "ubuntu" and paste your password. 3. Download and Run the Setup Script ▫️Run the following commands: wget -O sudo bash ▫️The server will reboot during the installation. ▫️After the reboot, reopen the console, log in again, and run: sudo bash ▫️Wait for all files and dependencies to install. ▫️When prompted, paste the address of your prover and the seed phrase of the wallet you used to create the prover and stake 1000 testPROVE. ✅ That’s it. Your prover is now up and running! crashout #ShowMeYourProver
2:51

My Prover Setup, or How to Run a Prover Using a TV Setting up and running a Succinct prover is so easy and fun that I decided to make a short 3-step guide on how to do it using a smart TV. This guide also works if you're using a smartphone. 1. Create Your Prover ▫️Go to: ▫️Stake 1000 testPROVE to your own prover. I didn’t include this part in the video because it’s pretty straightforward. 2. Log in to Genesis Cloud ▫️Open your browser and go to: ▫️In the left menu, click “Instances”, then “Create Instance”. ▫️Choose: RTX 3090 (or higher) / Ubuntu 24.04 / Authentication: Password ▫️Copy and save the automatically generated password. ▫️Click “Create Instance”. ▫️Go back to “Instances” in the left menu. ▫️On the right side of your new instance, click “Launch Console”. ▫️Log in as user "ubuntu" and paste your password. 3. Download and Run the Setup Script ▫️Run the following commands: wget -O sudo bash ▫️The server will reboot during the installation. ▫️After the reboot, reopen the console, log in again, and run: sudo bash ▫️Wait for all files and dependencies to install. ▫️When prompted, paste the address of your prover and the seed phrase of the wallet you used to create the prover and stake 1000 testPROVE. ✅ That’s it. Your prover is now up and running! crashout #ShowMeYourProver

Plastique

35,877 Aufrufe • vor 1 Jahr

Don’t babysit your agents. Put them in sandboxes. Each sandbox is a microVM with its own kernel. And now they run in the cloud, too. Ready to try it out? New signups get $250 in Cloud Sandboxes credit. Claim yours before the promotion ends on October 31:
1:04

Don’t babysit your agents. Put them in sandboxes. Each sandbox is a microVM with its own kernel. And now they run in the cloud, too. Ready to try it out? New signups get $250 in Cloud Sandboxes credit. Claim yours before the promotion ends on October 31:

Docker

17,116 Aufrufe • vor 1 Monat

A LINUX KERNEL DEVELOPER PROVED THE THING YOU PUSH CODE TO IS SECRETLY A DATABASE THAT CAN VERSION ALMOST ANYTHING AND THAT MOST DEVS HAVE ONLY EVER TOUCHED A TENTH OF IT 42 minutes from Josh Triplett -- a longtime Linux kernel and Debian developer -- showing that Git is a general-purpose, tamper-evident versioning engine that just happens to be famous for code. -> The moment it clicks, Git stops being "Where my code lives" and becomes what it really is underneath: a content-addressable store that can version almost anything -- your configs, your notes, your servers' state, entire datasets. People run whole wikis on it. They version their entire machine's configuration with it. They ship websites by pushing to it. They track data too big to email. None of it is a hack -- it's the same handful of objects you already use for code, pointed somewhere new. Treating Git as a code-only tool was never the ceiling -> it's a versioning engine for anything, and the people who see that automate what the rest of the team still does by hand. And as AI agents start spitting out not just code but configs, docs and data, the one system that can version and audit all of it at once is already sitting on your machine. You learned five commands to survive. This is the talk that shows you were standing on top of a database the whole time. It changes what you think the tool is even for. Bookmark & Watch it today ↓
42:24

A LINUX KERNEL DEVELOPER PROVED THE THING YOU PUSH CODE TO IS SECRETLY A DATABASE THAT CAN VERSION ALMOST ANYTHING AND THAT MOST DEVS HAVE ONLY EVER TOUCHED A TENTH OF IT 42 minutes from Josh Triplett -- a longtime Linux kernel and Debian developer -- showing that Git is a general-purpose, tamper-evident versioning engine that just happens to be famous for code. -> The moment it clicks, Git stops being "Where my code lives" and becomes what it really is underneath: a content-addressable store that can version almost anything -- your configs, your notes, your servers' state, entire datasets. People run whole wikis on it. They version their entire machine's configuration with it. They ship websites by pushing to it. They track data too big to email. None of it is a hack -- it's the same handful of objects you already use for code, pointed somewhere new. Treating Git as a code-only tool was never the ceiling -> it's a versioning engine for anything, and the people who see that automate what the rest of the team still does by hand. And as AI agents start spitting out not just code but configs, docs and data, the one system that can version and audit all of it at once is already sitting on your machine. You learned five commands to survive. This is the talk that shows you were standing on top of a database the whole time. It changes what you think the tool is even for. Bookmark & Watch it today ↓

slash1s

387,103 Aufrufe • vor 5 Monaten

MCP explained in one drawing: how your AI gets inside your apps (Claude, n8n, Gmail) Here's how your AI actually gets inside your apps. It's called MCP, and it fits in one drawing. On its own, an AI is a brain in a box. It can talk and think, but it can't open your email or build anything in n8n. It has no hands. 1. The door is the API. Most apps have a door for other software: Gmail has one, GitHub has one, n8n has one. Every door is a different shape, with its own rules and its own key. 2. The plug is MCP. It's one standard shape. On one side is the AI. On the other side is a small program called an MCP server, one per app, and it knows that app's door. The AI plugs in, and the server does the walking through the door. 3. The menu is the list of tools. When the AI plugs in, it sees every tool the app offers, in plain words. Gmail's server offers create a draft, search mail and read a message. n8n's server offers list workflows, create a workflow and test it. The AI reads the list like a menu, picks a tool and calls it. That's how Claude builds a whole workflow inside n8n, node by node, and checks its own work, while you never open n8n. 4. Connecting one takes one sentence. In Claude Code, say: "Connect yourself to my n8n, here's the address and a key I made just for you." Claude writes a small settings file with the server's name and the key. Type /mcp and it says connected. That's the setup for most apps. Some, like Gmail, you just sign in to instead. 5. The key is the important part. An MCP server can only use the tools on its menu, and only as far as its key allows. So make a key that can do the job and nothing more. If the app lets you limit a key, make it read only when reading is all you need. If it doesn't, use a test account. And if a key ever leaks, cancel it inside the app. One click, and nobody can use it, the AI included. One honest line: MCP doesn't make the AI smarter. It gives a smart thing a way to act, and everything it does through that plug, it does as you. So you still check the work, like you would a new hire's. If you're starting from scratch, I have step-by-step lessons for all of this inside my community. There are eight courses, in order, from the basics of AI to automations, building websites and apps, and AI agents. If you're interested, the link is in my profile. #mcp #cli ##aiagents #aiautomation #claudecode
2:56

MCP explained in one drawing: how your AI gets inside your apps (Claude, n8n, Gmail) Here's how your AI actually gets inside your apps. It's called MCP, and it fits in one drawing. On its own, an AI is a brain in a box. It can talk and think, but it can't open your email or build anything in n8n. It has no hands. 1. The door is the API. Most apps have a door for other software: Gmail has one, GitHub has one, n8n has one. Every door is a different shape, with its own rules and its own key. 2. The plug is MCP. It's one standard shape. On one side is the AI. On the other side is a small program called an MCP server, one per app, and it knows that app's door. The AI plugs in, and the server does the walking through the door. 3. The menu is the list of tools. When the AI plugs in, it sees every tool the app offers, in plain words. Gmail's server offers create a draft, search mail and read a message. n8n's server offers list workflows, create a workflow and test it. The AI reads the list like a menu, picks a tool and calls it. That's how Claude builds a whole workflow inside n8n, node by node, and checks its own work, while you never open n8n. 4. Connecting one takes one sentence. In Claude Code, say: "Connect yourself to my n8n, here's the address and a key I made just for you." Claude writes a small settings file with the server's name and the key. Type /mcp and it says connected. That's the setup for most apps. Some, like Gmail, you just sign in to instead. 5. The key is the important part. An MCP server can only use the tools on its menu, and only as far as its key allows. So make a key that can do the job and nothing more. If the app lets you limit a key, make it read only when reading is all you need. If it doesn't, use a test account. And if a key ever leaks, cancel it inside the app. One click, and nobody can use it, the AI included. One honest line: MCP doesn't make the AI smarter. It gives a smart thing a way to act, and everything it does through that plug, it does as you. So you still check the work, like you would a new hire's. If you're starting from scratch, I have step-by-step lessons for all of this inside my community. There are eight courses, in order, from the basics of AI to automations, building websites and apps, and AI agents. If you're interested, the link is in my profile. #mcp #cli ##aiagents #aiautomation #claudecode

Professor Glitch

18,500 Aufrufe • vor 29 Tagen

🚨 New advisory was just published! 🚨 An out-of-bounds write vulnerability in the Linux kernel achieves local privilege escalation on Ubuntu 22.04 for active user sessions:
2:06

🚨 New advisory was just published! 🚨 An out-of-bounds write vulnerability in the Linux kernel achieves local privilege escalation on Ubuntu 22.04 for active user sessions:

SSD Secure Disclosure

174,925 Aufrufe • vor 1 Jahr

U-Boot menu now works on Flipper One Now you can choose different kernels and systemd targets before the Linux kernel boots. * It's our LCD/buttons driver plus a slightly modified vanilla U-Boot menu. Will try to upstream it to U-Boot. * This is a temporary solution for testing the selection of different OS profiles. In the final version of Flipper OS, we’ll most likely draw the boot menu using a small program compiled into initramfs, after the kernel has started. But as a temporary solution, it works perfect. * I’m still not happy with the cold boot time: from powering on the processor to having a working menu. Here’s the boot log with timestamps, captured from the processor console: * Our patched U-Boot:
2:56

U-Boot menu now works on Flipper One Now you can choose different kernels and systemd targets before the Linux kernel boots. * It's our LCD/buttons driver plus a slightly modified vanilla U-Boot menu. Will try to upstream it to U-Boot. * This is a temporary solution for testing the selection of different OS profiles. In the final version of Flipper OS, we’ll most likely draw the boot menu using a small program compiled into initramfs, after the kernel has started. But as a temporary solution, it works perfect. * I’m still not happy with the cold boot time: from powering on the processor to having a working menu. Here’s the boot log with timestamps, captured from the processor console: * Our patched U-Boot:

Flipper R&D

34,038 Aufrufe • vor 4 Monaten

Why every Linux user in your comments keeps talking about CachyOS 👀 CachyOS is a fast, performance-focused Linux distro built on Arch Linux. It comes with a custom optimized kernel, built-in performance tools, and multiple desktop environments like KDE Plasma and GNOME, so you can customize your setup however you want. It's also a popular pick for Linux gaming and benchmarking because of how smoothly it runs. Plus, you get access to the huge Arch ecosystem, including the AUR. Is CachyOS the best Linux distro for gaming and performance? Let me know in the comments 👇 Like and follow for more tech and Linux content! #linux #archlinux #cachyos #window #tech
0:45

Why every Linux user in your comments keeps talking about CachyOS 👀 CachyOS is a fast, performance-focused Linux distro built on Arch Linux. It comes with a custom optimized kernel, built-in performance tools, and multiple desktop environments like KDE Plasma and GNOME, so you can customize your setup however you want. It's also a popular pick for Linux gaming and benchmarking because of how smoothly it runs. Plus, you get access to the huge Arch ecosystem, including the AUR. Is CachyOS the best Linux distro for gaming and performance? Let me know in the comments 👇 Like and follow for more tech and Linux content! #linux #archlinux #cachyos #window #tech

nosashi_explain

31,000 Aufrufe • vor 1 Monat

The 4 levels of AI in a business: how to go from AI tasks to an AI-run operation How to use AI in your business: there are four levels, and most owners never leave level one. At level four, a whole part of the business runs without you in the room. Here is the ladder, using Fernway, a fictional cleaning business, as the example: 1. Tasks. You open the chat, ask, and AI writes the email, the post, the quote. You save minutes a day, but you still decide what to do, still start it, and still carry every answer into your other apps. You got faster, and your plate is just as full. 2. Processes (AI workflow automation). A process is a job that happens the same way every time: a lead comes in, you reply, send prices, offer a time. Write the steps down once and connect them with an automation tool like n8n. Now the lead arriving is the trigger, the steps run by themselves, and the AI in the middle writes the reply. Fernway now answers new leads in 2 minutes instead of 4 hours, and the first business to answer often wins the job. How to pick your first one: it happens every week, and the rules fit on a sticky note. 3. Departments. This is where most people go wrong: ten automations that don't know about each other. A department is those processes sharing one memory. Answer the lead, qualify it, book the visit, follow up on day 3 and day 7, update the deal, all reading and writing the same customer record in your CRM. Nobody gets asked the same question twice and nothing falls through the cracks. Add rules, like a discount over 10 percent waits for your yes. That is human in the loop where it matters. 4. The operation (AI agents and AI employees). Sales, delivery, invoices, reviews: each department gets its own AI employee, and they hand work to each other through one board. A job closes in sales and delivery picks it up. Delivery finishes and invoicing sends the bill. Your job becomes managing it: approving, reading the numbers, fixing what breaks. How to tell where you are: if AI only works when you open it, you're on level one. If one job runs without you, you're on level two. Each level saves you more than the one before, and most owners are stuck on tasks. This fits service businesses and small teams: cleaning companies, agencies, home services, clinics, any business that gets leads and follows up on them. #aiautomation #aiagents #aiemployee
2:18

The 4 levels of AI in a business: how to go from AI tasks to an AI-run operation How to use AI in your business: there are four levels, and most owners never leave level one. At level four, a whole part of the business runs without you in the room. Here is the ladder, using Fernway, a fictional cleaning business, as the example: 1. Tasks. You open the chat, ask, and AI writes the email, the post, the quote. You save minutes a day, but you still decide what to do, still start it, and still carry every answer into your other apps. You got faster, and your plate is just as full. 2. Processes (AI workflow automation). A process is a job that happens the same way every time: a lead comes in, you reply, send prices, offer a time. Write the steps down once and connect them with an automation tool like n8n. Now the lead arriving is the trigger, the steps run by themselves, and the AI in the middle writes the reply. Fernway now answers new leads in 2 minutes instead of 4 hours, and the first business to answer often wins the job. How to pick your first one: it happens every week, and the rules fit on a sticky note. 3. Departments. This is where most people go wrong: ten automations that don't know about each other. A department is those processes sharing one memory. Answer the lead, qualify it, book the visit, follow up on day 3 and day 7, update the deal, all reading and writing the same customer record in your CRM. Nobody gets asked the same question twice and nothing falls through the cracks. Add rules, like a discount over 10 percent waits for your yes. That is human in the loop where it matters. 4. The operation (AI agents and AI employees). Sales, delivery, invoices, reviews: each department gets its own AI employee, and they hand work to each other through one board. A job closes in sales and delivery picks it up. Delivery finishes and invoicing sends the bill. Your job becomes managing it: approving, reading the numbers, fixing what breaks. How to tell where you are: if AI only works when you open it, you're on level one. If one job runs without you, you're on level two. Each level saves you more than the one before, and most owners are stuck on tasks. This fits service businesses and small teams: cleaning companies, agencies, home services, clinics, any business that gets leads and follows up on them. #aiautomation #aiagents #aiemployee

Professor Glitch

20,700 Aufrufe • vor 1 Monat

🐧 EVERY LINUX DISTRO SIMPLY EXPLAINED ⚠️ DISCLAIMER: This tutorial is for educational purposes. All demonstrations and installations should be performed on your own computer or an authorized virtual machine. Many people think Linux is just one operating system, but that's not entirely true. Linux is a kernel, and different communities and organizations build complete operating systems around it. These are called Linux distributions (distros). 💻 WHAT IS A LINUX DISTRO? A Linux distribution combines the Linux kernel with system tools, package managers, desktop environments, and applications to create a usable operating system. 🔥 POPULAR LINUX DISTRIBUTIONS 1. Ubuntu – Beginner-friendly, popular for desktops, servers, and development. 2. Debian – Known for stability and widely used as the foundation for other distributions. 3. Kali Linux – Designed for penetration testing, digital forensics, and security research. 4. Linux Mint – A user-friendly desktop distribution, especially for people moving from Windows. 5. Fedora – Features modern Linux technologies and is popular with developers. 6. Arch Linux – A flexible distribution that gives users extensive control over system configuration. 7. Parrot OS – Includes tools and features for cybersecurity, privacy, and digital forensics. 8. Rocky Linux – An enterprise-focused distribution commonly used for servers. 9. openSUSE – Offers powerful administration tools and desktop/server options. 10. Alpine Linux – Lightweight and commonly used in containers and minimal environments. 🧠 WHAT MAKES THEM DIFFERENT? Linux distributions can differ in: - Package managers and software repositories. - Desktop environments and user interfaces. - Security configurations. - System requirements. - Release cycles and update policies. - Intended use cases. 🎯 WHICH ONE SHOULD YOU LEARN? For beginners, Ubuntu or Linux Mint can be a good starting point. For cybersecurity labs, Kali Linux and Parrot OS provide specialized tools. For server administration, explore Debian, Ubuntu Server, and Rocky Linux. For deeper Linux knowledge, try Arch Linux in a virtual machine. 🔐 FINAL LESSON Learning Linux isn't about installing every distribution. It's about understanding the kernel, commands, file systems, permissions, networking, package management, and system administration. Start with one distro, build your skills, and explore others as you progress. Hackwithbanks — Learn. Practice. Understand. #Hackwithbanks #Linux #LinuxDistro #KaliLinux #Ubuntu
12:05

🐧 EVERY LINUX DISTRO SIMPLY EXPLAINED ⚠️ DISCLAIMER: This tutorial is for educational purposes. All demonstrations and installations should be performed on your own computer or an authorized virtual machine. Many people think Linux is just one operating system, but that's not entirely true. Linux is a kernel, and different communities and organizations build complete operating systems around it. These are called Linux distributions (distros). 💻 WHAT IS A LINUX DISTRO? A Linux distribution combines the Linux kernel with system tools, package managers, desktop environments, and applications to create a usable operating system. 🔥 POPULAR LINUX DISTRIBUTIONS 1. Ubuntu – Beginner-friendly, popular for desktops, servers, and development. 2. Debian – Known for stability and widely used as the foundation for other distributions. 3. Kali Linux – Designed for penetration testing, digital forensics, and security research. 4. Linux Mint – A user-friendly desktop distribution, especially for people moving from Windows. 5. Fedora – Features modern Linux technologies and is popular with developers. 6. Arch Linux – A flexible distribution that gives users extensive control over system configuration. 7. Parrot OS – Includes tools and features for cybersecurity, privacy, and digital forensics. 8. Rocky Linux – An enterprise-focused distribution commonly used for servers. 9. openSUSE – Offers powerful administration tools and desktop/server options. 10. Alpine Linux – Lightweight and commonly used in containers and minimal environments. 🧠 WHAT MAKES THEM DIFFERENT? Linux distributions can differ in: - Package managers and software repositories. - Desktop environments and user interfaces. - Security configurations. - System requirements. - Release cycles and update policies. - Intended use cases. 🎯 WHICH ONE SHOULD YOU LEARN? For beginners, Ubuntu or Linux Mint can be a good starting point. For cybersecurity labs, Kali Linux and Parrot OS provide specialized tools. For server administration, explore Debian, Ubuntu Server, and Rocky Linux. For deeper Linux knowledge, try Arch Linux in a virtual machine. 🔐 FINAL LESSON Learning Linux isn't about installing every distribution. It's about understanding the kernel, commands, file systems, permissions, networking, package management, and system administration. Start with one distro, build your skills, and explore others as you progress. Hackwithbanks — Learn. Practice. Understand. #Hackwithbanks #Linux #LinuxDistro #KaliLinux #Ubuntu

Hackswithbanks

13,300 Aufrufe • vor 3 Tagen

F*CKING INSANE. Elon's Grok Bot just reverse engineered the Binance AND Polymarket hacks like an elite hacker, and it took ONE prompt the two attacks took $6.3B and $2.6M this week my bot took both apart in 14 minutes and found the polymarket door wide open on 12 sites RIGHT NOW. ELON YOU HAVE TO SEE THIS a 3-bot security team running on Grok Bot's own cloud computer in 14 minutes it: → took apart the polymarket drain: one swapped speed-tracking script hijacked logged-in tabs, dumped positions, ~740 wallets in an hour → took apart the binance drain: 50 EIP-7702 authorizations in one transaction, one hit, a whole ETH hot wallet swept → checked my own wallet for a rogue 7702 delegation, the binance trick → checked every site i trade on for the polymarket trick: 23 outside scripts on the pages where i sign. the drain needed one → found 12 crypto sites loading those scripts with no lock tonight, exactly like polymarket did no password. no seed phrase. no signature. a logged-in tab was enough and one of those doors was on polymarket, where my wallet sat yesterday. ~740 wallets weren't as lucky this is the same Grok Bot that made me $10,000 on polymarket. now it guards it so now it never sleeps it fingerprints every script on every site that touches my money and rechecks them every 10 minutes it watches my wallet for any new authorization i didn't make it pings my phone the minute a script changes, before i open that site again if you hold even $100 in crypto, you NEED a bot like this. the scammers already have one you can check how it works on the video how to build yours tonight: → make 3 bots in Grok Bot: Scout, Watchdog, Skeptic → Scout maps every site and wallet you use. Watchdog rechecks them every 10 minutes. Skeptic reads every change like a hacker → give Scout the polymarket and binance attacks as examples, it hunts for the same pattern everywhere → connect your phone so a swapped script hits you within 10 minutes, not after the money is gone $6.3B + $2.6M in one day. 14 minutes for my bots how long until your wallet is next? the exact prompt is in the replies, the full Grok Bot setup is in the article in my profile
2:10

F*CKING INSANE. Elon's Grok Bot just reverse engineered the Binance AND Polymarket hacks like an elite hacker, and it took ONE prompt the two attacks took $6.3B and $2.6M this week my bot took both apart in 14 minutes and found the polymarket door wide open on 12 sites RIGHT NOW. ELON YOU HAVE TO SEE THIS a 3-bot security team running on Grok Bot's own cloud computer in 14 minutes it: → took apart the polymarket drain: one swapped speed-tracking script hijacked logged-in tabs, dumped positions, ~740 wallets in an hour → took apart the binance drain: 50 EIP-7702 authorizations in one transaction, one hit, a whole ETH hot wallet swept → checked my own wallet for a rogue 7702 delegation, the binance trick → checked every site i trade on for the polymarket trick: 23 outside scripts on the pages where i sign. the drain needed one → found 12 crypto sites loading those scripts with no lock tonight, exactly like polymarket did no password. no seed phrase. no signature. a logged-in tab was enough and one of those doors was on polymarket, where my wallet sat yesterday. ~740 wallets weren't as lucky this is the same Grok Bot that made me $10,000 on polymarket. now it guards it so now it never sleeps it fingerprints every script on every site that touches my money and rechecks them every 10 minutes it watches my wallet for any new authorization i didn't make it pings my phone the minute a script changes, before i open that site again if you hold even $100 in crypto, you NEED a bot like this. the scammers already have one you can check how it works on the video how to build yours tonight: → make 3 bots in Grok Bot: Scout, Watchdog, Skeptic → Scout maps every site and wallet you use. Watchdog rechecks them every 10 minutes. Skeptic reads every change like a hacker → give Scout the polymarket and binance attacks as examples, it hunts for the same pattern everywhere → connect your phone so a swapped script hits you within 10 minutes, not after the money is gone $6.3B + $2.6M in one day. 14 minutes for my bots how long until your wallet is next? the exact prompt is in the replies, the full Grok Bot setup is in the article in my profile

Argona

205,650 Aufrufe • vor 16 Tagen

Part 1 ~ Nobody Explained Linux to me like this.... Want to understand Linux from scratch? In this video, we break down Linux in a simple and beginner friendly way and explain why it powers much of today's technology infrastructure. Topics covered in this video: • What Linux is and why it was created • What an Operating System (OS) does • How operating systems interact with hardware and software • Linux vs Windows vs macOS • What Free and Open Source Software means • Understanding the Linux Kernel • Understanding the Linux Shell • Introduction to Bash Shell • How users interact with Linux using commands • What Linux Distributions (Distros) are • Ubuntu Explained • Red Hat Enterprise Linux Explained • Why there are multiple Linux distributions • Linux Security Model and Permissions • Why Linux is considered secure • Linux Stability and Reliability • Linux Performance Advantages • Linux Cost Benefits for Organizations • Linux Customization and Flexibility • Why Linux dominates the server world • Linux in Cloud Computing • Linux in AWS, Azure, and Google Cloud • Linux in Android Smartphones • Linux in Supercomputers • Linux in DevOps and Modern Infrastructure • Linux Career Opportunities • Why Linux is a must learn skill for IT professionals • Key Linux Concepts Every Beginner Should Know Whether you are a student, developer, cloud engineer, DevOps engineer, system administrator, cybersecurity professional, or someone preparing for technical interviews, this video will give you a strong foundation in Linux. #Linux​ #LinuxTutorial​ #DevOps​ #CloudComputing​ #OperatingSystem​
1:09

Part 1 ~ Nobody Explained Linux to me like this.... Want to understand Linux from scratch? In this video, we break down Linux in a simple and beginner friendly way and explain why it powers much of today's technology infrastructure. Topics covered in this video: • What Linux is and why it was created • What an Operating System (OS) does • How operating systems interact with hardware and software • Linux vs Windows vs macOS • What Free and Open Source Software means • Understanding the Linux Kernel • Understanding the Linux Shell • Introduction to Bash Shell • How users interact with Linux using commands • What Linux Distributions (Distros) are • Ubuntu Explained • Red Hat Enterprise Linux Explained • Why there are multiple Linux distributions • Linux Security Model and Permissions • Why Linux is considered secure • Linux Stability and Reliability • Linux Performance Advantages • Linux Cost Benefits for Organizations • Linux Customization and Flexibility • Why Linux dominates the server world • Linux in Cloud Computing • Linux in AWS, Azure, and Google Cloud • Linux in Android Smartphones • Linux in Supercomputers • Linux in DevOps and Modern Infrastructure • Linux Career Opportunities • Why Linux is a must learn skill for IT professionals • Key Linux Concepts Every Beginner Should Know Whether you are a student, developer, cloud engineer, DevOps engineer, system administrator, cybersecurity professional, or someone preparing for technical interviews, this video will give you a strong foundation in Linux. #Linux​ #LinuxTutorial​ #DevOps​ #CloudComputing​ #OperatingSystem​

Thom Code

139,100 Aufrufe • vor 5 Monaten

SemiAnalysis' Jordan Nanos reveals the OpenAI engineers scrolling their own kernel code had no idea what it did line by line, and it didn't matter because the AI does "It's not slop, because it performs, but it's literally just AI-generated assembly basically puked out in Gluon, this low-level kernel programming language that they built on top of Triton, which uses this really interesting programming model." "The point is, the guys who were scrolling through this code with us, it was kind of clear that they know a whole bunch about hardware, they know a whole bunch about the concepts in the system, and they just have no idea what this MLA kernel that they're showing us for DeepSeek actually does." "You can go line by line and it's like nope, nope, nope." "But it doesn't matter. The AI understands it, the AI tests it, and you see the results. It produces correct kernels that perform really well." "And I think this is just a sign of what's to come again." "That actual code is not necessarily something that a human has to reason about deeply, if the AI knows how to manipulate the data movement and the processing elements on the hardware that you've given it." ________ More takeaways from OpenAI's Jalapeno development:
1:11

SemiAnalysis' Jordan Nanos reveals the OpenAI engineers scrolling their own kernel code had no idea what it did line by line, and it didn't matter because the AI does "It's not slop, because it performs, but it's literally just AI-generated assembly basically puked out in Gluon, this low-level kernel programming language that they built on top of Triton, which uses this really interesting programming model." "The point is, the guys who were scrolling through this code with us, it was kind of clear that they know a whole bunch about hardware, they know a whole bunch about the concepts in the system, and they just have no idea what this MLA kernel that they're showing us for DeepSeek actually does." "You can go line by line and it's like nope, nope, nope." "But it doesn't matter. The AI understands it, the AI tests it, and you see the results. It produces correct kernels that perform really well." "And I think this is just a sign of what's to come again." "That actual code is not necessarily something that a human has to reason about deeply, if the AI knows how to manipulate the data movement and the processing elements on the hardware that you've given it." ________ More takeaways from OpenAI's Jalapeno development:

Fireside Alpha

4,572,224 Aufrufe • vor 2 Monaten

single most useful thing for my local ai setup that made life easy is my 2x DGX Spark becoming one serving box for every device i own here is how i do it: > 1. the two sparks run one vLLM server together with one endpoint, i use dgx sparks, you can use any nodes that can run an llm > 2. install tailscale to have all your nodes and machines on the same tailnet, the endpoint lives there so every device reaches it from anywhere and no port is open to the internet > 3. that one tailnet endpoint with the exposed port is your base url for everything, it speaks OpenAI chat, OpenAI Responses and Anthropic Messages, so any chat app, coding agent or phone bot you point at it talks to the same model > 4. the server also answers to the name "local", so clients can ask for "local" instead of a model name and when i swap the model on the servers nothing on the devices needs a config change. right now it's serving GLM 5.3-Flash, next week it can be something else this video below is the whole thing running end to end, orange dots are requests going in, white and green are tokens coming back it takes however many requests you throw at it, your config decides how many run at once and the rest wait their turn, when several run together each request's tok/s drops a bit while the box moves more tokens in total. mine is set to one at a time right now, so when every device fires together they queue up this setup really improved my quality of life with local ai, if you get stuck anywhere setting it up leave a comment and i'll help you
1:04

single most useful thing for my local ai setup that made life easy is my 2x DGX Spark becoming one serving box for every device i own here is how i do it: > 1. the two sparks run one vLLM server together with one endpoint, i use dgx sparks, you can use any nodes that can run an llm > 2. install tailscale to have all your nodes and machines on the same tailnet, the endpoint lives there so every device reaches it from anywhere and no port is open to the internet > 3. that one tailnet endpoint with the exposed port is your base url for everything, it speaks OpenAI chat, OpenAI Responses and Anthropic Messages, so any chat app, coding agent or phone bot you point at it talks to the same model > 4. the server also answers to the name "local", so clients can ask for "local" instead of a model name and when i swap the model on the servers nothing on the devices needs a config change. right now it's serving GLM 5.3-Flash, next week it can be something else this video below is the whole thing running end to end, orange dots are requests going in, white and green are tokens coming back it takes however many requests you throw at it, your config decides how many run at once and the rest wait their turn, when several run together each request's tok/s drops a bit while the box moves more tokens in total. mine is set to one at a time right now, so when every device fires together they queue up this setup really improved my quality of life with local ai, if you get stuck anywhere setting it up leave a comment and i'll help you

Sudo su

33,246 Aufrufe • vor 1 Monat