Lummox
@Lummox_eth • 9,250 subscribers
developing internet money / building interesting things / dm open
Shorts
Videos
I launched Grok Bot for 24 hours, gave it 3 jobs, and stopped treating AI like a chatbot. I had it scan live X + web data for emerging AI products, research what people were actually saying about them, filter the obvious engagement bait, compare sources and turn the strongest signals into a structured brief. At the same time, I could have separate research paths looking at competitors, code or market opportunities instead of forcing everything through one conversation. That’s the difference I underestimated. I don’t want to sit inside a chat writing another prompt every 5 minutes. I want to define the objective, give the agent access to the right tools and data, let it execute, then step back in when judgment or approval is actually needed. The old workflow was prompt → answer → copy → repeat. The workflow I’m moving toward is assign → execute → verify. Once AI stops waiting for your next message, it stops feeling like software you use and starts feeling like work you delegated.
Lummox3,968,059 views • 2 months ago
Claude Code really charged $131 to build a train that looks like a refrigerator on rails 😭 Meanwhile Kimi and GPT built an entire cinematic bullet train scene for $19 and $54 Same prompt, 4 AI workflows The final bill 🥇 Kimi K3.1 : $19 / 5 min 🥈 GPT 6 Astra : $54 / 15 min 🥉 Blender MCP : $87 / 24 min 💀 Claude 3.7 Sonnet : $131 / 38 min Kimi was almost 7x cheaper and 7.6x faster in this comparison And somehow the cheapest result has the city lights, reflections and atmosphere you’d actually want in a cinematic render bro imagine paying $131 two years ago just to invent the world’s fastest microwave
Lummox110,590 views • 13 days ago
AI memory is getting f...cking illegal 10 open-source GitHub projects that stop agents from starting from zero EVERY new session 01 Mem0 ▸ → 66K+ stars 02 Hindsight ▸ → retain → recall → reflect 03 memU ▸ TURN MEMORY INTO KNOWLEDGE 04 Cognee ▸ → documents + code + conversations → knowledge graph 05 Graphiti ▸ → remembers how facts change over time 06 OpenViking ▸ MAKE THE AGENT STATEFUL 07 Letta ▸ → memory + identity across sessions 08 Letta Code ▸ REMEMBER ACROSS THE STACK 09 OpenMemory ▸ 10 Agent Memory Benchmark ▸ the loop: experience → remember → connect → retrieve → act → update 3 builds I'd actually test: coding: Hindsight → Cognee → Letta Code personal agent: Mem0 → Graphiti → Letta company brain: Cognee → Graphiti → Hindsight and this is where the sh...t gets interesting: bigger context isn't the same as memory a model can hold 1M tokens and still meet you for the first time every morning these projects change that save this before your agent forgets it ⭣
Lummox333,703 views • 1 month ago
there’s almost no downside and that’s the weird part Elon Musk shared an idea that pulled 27 million views: “stop chasing startups. own boring pieces of the internet and let bots operate them” that’s basically what Grok Bot is doing now. and outside the subscription, the setup costs $0: → make a bot and call it Mogul → give it one job: find a boring niche people search for every single day (it scans X and reddit complaints live, because it’s grok) → it builds a small directory site locally and publishes one useful page every night → it reaches out for affiliate and referral partnerships itself, through your email, under your name → publishing happens automatically on schedule. telegram sends you the revenue reports → month one: almost nothing but traffic. month three: $217. then the pages just keep stacking your laptop can stay closed. the website keeps getting bigger every new page becomes another tiny worker that never clocks out. traffic stacks. revenue stacks. the whole thing compounds some people already run 3 sites like this and barely look at them. they aren’t tweeting about it either. you can probably guess why while you spend another month thinking about which niche to enter, someone else’s bot publishes 30 more pages into it full playbook below save this before your niche isn’t yours anymore
Lummox276,876 views • 2 months ago
GitHub is becoming f...cking dangerous for AI agents 10 open-source repos that turn one powerful model into an entire autonomous system 01 LangGraph ▸ → stateful agent orchestration 02 PydanticAI ▸ → typed agents + structured outputs BUILD THE CONTROL LAYER 03 Mastra ▸ → agents + workflows + memory 04 Agno ▸ → multi-agent systems GIVE IT CONTEXT 05 Cognee ▸ → data → knowledge → agent memory 06 Graphiti ▸ → knowledge that changes over time GIVE IT A COMPUTER 07 Browser Use ▸ → AI agents that operate websites 08 E2B ▸ → isolated environments for execution KEEP IT ALIVE 09 Langfuse ▸ → traces + evals + observability 10 DeepEval ▸ → test agent behavior before production the architecture: model → context → planner → memory → tools → execution → eval → retry that's the shift happening on GitHub the best repos aren't making another chatbot they're building everything the model needs around it ⭣
Lummox168,963 views • 1 month ago
THE NEXT AI WORKER MAY COST $500 AND SIT BEHIND YOUR MONITOR. Mini PCs are becoming the cheapest path to private AI. A Mac Mini M4 starts at $599, Geekom AX8 Pro starts near $529, and MinisForum pushes up to 128GB RAM for local AI workloads. Ollama turns that hardware into a personal model server. No cloud fee. No data storage bill. No waiting for another SaaS product to decide your limits. Market shift is simple -> computer is moving back to the desk. Owning a mini PC is starting to look less like buying hardware and more like hiring a quiet AI operator.
Lummox394,894 views • 4 months ago
this is where GPT-6 Astra is getting f...cking illegal 10 open-source GitHub repos I'd use to build the system around the model 01 PydanticAI ▸ → typed agents + structured outputs 02 LangGraph ▸ → state + long-running agent workflows BUILD THE AGENTS 03 Agno ▸ → agents + teams + workflows 04 smolagents ▸ → agents that execute through code GIVE THEM MEMORY 05 Mem0 ▸ → persistent agent memory 06 Graphiti ▸ → temporal knowledge graphs GIVE THEM A RUNTIME 07 E2B ▸ → isolated sandboxes for AI-generated code 08 Daytona ▸ → infrastructure for running agent workloads TEST EVERYTHING 09 DeepEval ▸ → evals for LLM systems 10 Langfuse ▸ → traces + evals + production monitoring the flow: request → agent → memory → tools → runtime → eval → retry → deploy that's the part people underestimate the model is one folder the system around it is the product ⭣
Lummox43,873 views • 1 month ago
AI models are getting so good that paying 9x more can literally get you a worse 3D render same Japanese temple, same prompt, 4 AI tools the price difference is actually ridiculous → Kimi K3.1: $14 in 4 minutes → GPT 6 Astra: $48 in 14 minutes → Blender MCP: $79 in 23 minutes → Claude Code: $127 in 34 minutes Kimi is 9x cheaper and 8.5x faster than Claude Code in this test and the funniest part is that the $14 render arguably looks better than the $127 one bro we’re reaching a point where the most expensive AI isn’t even the flex anymore
Lummox18,795 views • 16 days ago
this GPT-6 Astra setup is getting f...cking dangerous 8 GitHub repos for turning one coding agent into an entire engineering team 01 OpenHands ▸ → autonomous software development 02 SWE-agent ▸ → issue → inspect repo → patch → test BUILD THE WORKERS 03 Aider ▸ → pair programming directly inside your repo 04 Cline ▸ → plan → edit → terminal → browser 05 Roo Code ▸ → specialized coding agent modes GIVE THEM A WORKSPACE 06 E2B ▸ → isolated sandboxes for parallel execution 07 Daytona ▸ → reproducible environments for agent workloads CHECK THE WORK 08 SWE-bench ▸ → test agents against real GitHub issues the swarm: issue → planner → split tasks → parallel agents → patches → tests → critic → merge one agent fixes the backend another handles the frontend another runs tests while a critic checks the diff that's where GPT-6 Astra gets interesting not one AI writing more code an entire software team running inside the repo ⭣
Lummox37,036 views • 1 month ago
AI research is getting f...cking illegal 15 GitHub projects for turning a question into sources, evidence, citations, and a finished report 01 GPT Researcher ▸ → recursive deep research 02 STORM ▸ SEARCH DEEPER 03 Perplexica ▸ 04 Open Deep Research ▸ 05 DeerFlow ▸ READ THE WEB 06 Firecrawl ▸ 07 Crawl4AI ▸ 08 Jina Reader ▸ TURN PAPERS INTO EVIDENCE 09 PaperQA2 ▸ → agentic RAG + citations 10 OpenScholar ▸ 11 PaperMage ▸ CONNECT THE SOURCES 12 GraphRAG ▸ 13 LightRAG ▸ 14 Kotaemon ▸ 15 Docling ▸ the loop: ask → split the problem → search multiple paths → read the sources → extract evidence → connect the findings → challenge the answer → cite → write 3 stacks I'd actually try: deep research: GPT Researcher → Firecrawl → GraphRAG → Docling scientific: OpenScholar → PaperQA2 → PaperMage → LightRAG private research: Perplexica → Crawl4AI → Kotaemon → GraphRAG the interesting part isn't generating a 20-page report it's being able to trace where every important claim came from AI already learned how to write this stack teaches it how to research ⭣
Lummox38,691 views • 1 month ago
a 23-year-old guy built an AI animation factory in one weekend. $12,345 last month. 4 hours of direction total. > Claude writes scripts and scene breakdowns: 10 minutes. > Midjourney generates every frame: 20 minutes. > Runway animates every scene: 15 minutes. > ElevenLabs voices every character: 10 minutes. > Suno scores every episode: 5 minutes. > Make publishes to YouTube, Patreon, and X: 0 minutes. total: $124/month 4 content types running in parallel. 8 uploads per week. he sleeps. it produces. every prompt is in the article above.
Lummox80,164 views • 5 months ago
HERMES AGENT CROSSED 140,000 GITHUB STARS IN 3 MONTHS AND JUST BECAME THE MOST USED AGENT IN THE WORLD. Most AI agents forget everything between sessions. Hermes writes its own skills from experience. Next time it runs the skill, improves it, and gets faster. Independent benchmarks show agents with 20+ self-created skills complete similar tasks 40% faster than fresh instances. Qwen 3.6 where the 35B version outperforms last year's 120B models at one third the memory footprint. DGX Spark with 128GB unified memory running everything locally at $0 per month after hardware. The setup takes 30 minutes. LM Studio plus Qwen 3.6 27B for the model server. One install script for Hermes. One config connecting them. Set context window to 65,536 tokens or nothing works. After one month of daily use your skills directory has 20 to 50 learned workflows. Your Hermes is genuinely different from anyone else's.
Lummox69,824 views • 5 months ago
AI network is getting f…cking illegal 4 GitHub repos building a peer-to-peer market for AI no central inference provider sitting in the middle 01 AntSeed ▸ → P2P discovery + routing + AI inference BUILD THE NETWORK 02 AIPs ▸ → discovery + transport + payments + reputation + verification CONNECT THE AGENTS 03 openclaw-channel-antseed ▸ → turn an OpenClaw agent into a provider on the network ROUTE THE MODELS 04 pi-antseed ▸ → discover AntSeed models directly inside Pi the architecture: AI agent → local router → DHT discovery → provider → model → response providers can compete on: → price → latency → reputation buyers discover them through the network requests move through encrypted P2P connections and the same network can sit underneath existing AI tools the interesting part isn’t another model it’s turning AI inference into an open market models become services → agents find them → the network routes the work ⭣
Lummox22,939 views • 1 month ago
the agent failed while I was asleep and somehow the work still got finished that was the test I actually wanted i gave GPT-6 Astra a workflow that normally needed me between every step inbound came in messy one agent had to sort it, another find the missing context, another build the deliverable, and the last one check everything before it reached me somewhere in the middle, agent 2 got stuck the source it needed wasn’t there normally that kills the entire chain instead, it wrote the missing piece into a blocker file and passed everything else forward the other agents kept working when I came back, most of the pack was finished and the one thing they couldn’t solve was sitting separately waiting for me that felt more important than a perfect demo because an agent completing 10/10 tasks is impressive an agent failing without taking the other 9 down with it is infrastructure
Lummox19,046 views • 1 month ago
GPT 6 Astra is making AI agents f…cking illegal 10 GitHub repos I’d put around Astra to turn one model into an entire AI system 01 LangGraph ▸ → control the agent loop 02 PydanticAI ▸ → structured agents + typed outputs GIVE IT MEMORY 03 Mem0 ▸ 04 Graphiti ▸ → temporal knowledge graph GIVE IT TOOLS 05 Composio ▸ 06 Browser Use ▸ → let agents operate the web LET IT EXECUTE 07 E2B ▸ → isolated cloud sandboxes 08 Daytona ▸ → disposable development environments WATCH EVERYTHING 09 Langfuse ▸ 10 DeepEval ▸ the architecture: GPT 6 Astra → context → memory → tools → execution → evaluation → correction → memory that’s the part people keep missing the model is only one block the real advantage comes from everything you build around it ⭣
Lummox14,448 views • 1 month ago
A 19-year-old aimed one AI at a street junction and turned it into $11,000 a month. An old camera, $47 in tools, one weekend of work. YOLO11 boxes every car, motorbike, and pedestrian as it moves, 653 in 5 minutes, and marks the ones running the light. The old count needed a survey crew on the curb with clickers through rush hour. His needed a script behind a camera that was already there. He recorded full video, not a single frame, so the district had nothing to argue with. He brought a flash drive to the district office and asked for 10 minutes. He left with a signed contract, then 5 more clients called. He left every step in the article.
Lummox30,375 views • 4 months ago
A developer opened a law firm’s token bill and realized 20 users were burning $61,920 just to ask questions their files could answer offline. $61,920 cloud TCO. $23,180 local TCO. 4.12-month break-even. 20 employees. 0 client files uploaded. The old RAG pitch breaks when legal files, medical notes, and finance records cannot leave the building. Search snippets are not enough when the firm needs a living wiki of clients, projects, entities, and cross-linked decisions. So the agency installs a local node : OpenClaw watches the intake folder, a local model cleans and links the documents, and Obsidian becomes the private corporate brain. Emails, Slack exports, PDFs, and notes turn into markdown pages the team can browse without sending secrets to an API. That is the real B2B offer. Not “AI chat for your company,” but a private knowledge machine that pays for itself in month 5.
Lummox25,799 views • 3 months ago
GPT-6 Astra makes AI coding look f…cking dangerous everyone is obsessing over the model the interesting part is the repo around it /context → manages the working window /router → decides which agent gets the task /agents → planner → executor → critic /memory → stores what survives the session /tools → browser → code → search → APIs /eval → checks whether the output actually worked /safety → permissions + sandbox boundaries /deployment → Docker → CI → Kubernetes /monitoring → traces → alerts → failures the architecture: request → context → router → agent → tools → memory → eval → retry → deploy that’s what changes with systems like GPT-6 Astra the model can write the code the repo has to make sure the AI survives after it does ⭣
Lummox11,692 views • 1 month ago
My Grok Bot turned $1,000 into $3,900 after Elon Musk quoted me. Now someone created a token around what I’m building. I didn’t deploy it. The community did $GrokBottGame. CA: 6py6orPjXSy8GQDkBk2jvW8KQiFHXJ55C58QJ8sypump And strangely, it makes me want to push this experiment even further. I’m still building the trading desk. Still trading. Still breaking agents, rebuilding them and tightening every rule that controls my money. From now on I’ll share the progress publicly: trades, failures, upgrades, numbers and everything I learn along the way. Maybe this becomes more than a trading experiment. Maybe it becomes a community building the whole thing with me.
Lummox14,506 views • 2 months ago
JIM SIMONS NEVER HAD A LOSING YEAR IN 34 ATTEMPTS. HERE IS THE 5 LAYER PIPELINE BEHIND IT. Spring 1988. A Cold War mathematician fired every trader and replaced them with signal detectors. $60 billion in performance fees. 39% annual returns after a 44% fee. Zero losing years. Layer 1 : Hurst exponent. Below 0.45 fades work. Above 0.55 momentum works. Medallion switches strategy automatically based on regime. Layer 2 : Factor decomposition. Renaissance runs 200+ factors privately. High alpha plus low R-squared means your edge is unexplained by known factors. Layer 3 : Markov Chain regime modeling. 76% Bear probability over 5 periods means no long positions. The model decides before you do. Layer 4 : Neural net. Three outputs : long, short, flat. Not predicting price. Predicting direction probability. 52% accuracy Kelly-sized correctly still prints money. Layer 5 : Execution. Half-Kelly sizing. Hard 2% risk cap. Edge must exceed execution cost by 1.5x minimum. This is where 80% of quants bleed out. 400 people. One shared model. The edge was never a single formula. It was always the pipeline.
Lummox24,175 views • 5 months ago