NO1ennn's banner
NO1ennn's profile picture

NO1ennn

@N01ennn • 7,677 subscribers

AI grinder / Dev / Content Creator & Researcher Everything is going as usual

Shorts

GPT-6 ASTRA TURNED $240 INTO $2,860. THEN IT LOCKED MOST OF THE MONEY AWAY FROM ITS OWN TRADING AGENTS. four sessions into the experiment, i opened the trading wallet. $895. the report said $2,860. i refreshed both screens, convinced something had gone wrong. then i opened the transfers. $1,965 had moved into a separate reserve wallet. every transfer had a receipt. the numbers matched to the cent. the desk had four jobs. three agents handled discovery, contract checks and execution. the fourth was called CASHIER. CASHIER watched settled results. after each profitable session, it moved 75% of the net gain into reserve. the trading agents could see that balance. their permissions gave them no way to pull it back. that detail suddenly became very interesting to me. i started calculating what the next position could look like with the whole $2,860 available. opened the settings. hovered over the allocation. then closed them. the next session was ugly. two positions closed red. working capital dropped from $895 to $731.40. i watched the desk reduce its next order and keep scanning. the reserve still read $1,965. that was the first time the setup made sense emotionally. a bad session had a smaller pile of money available to damage. i spent another hour reading the logs. the least exciting agent on the desk had done the thing i was most likely to postpone. it had actually taken the money off the table. CASHIER never found a single winning token. it was the first agent i decided to keep.

GPT-6 ASTRA TURNED $240 INTO $2,860. THEN IT LOCKED MOST OF THE MONEY AWAY FROM ITS OWN TRADING AGENTS. four sessions into the experiment, i opened the trading wallet. $895. the report said $2,860. i refreshed both screens, convinced something had gone wrong. then i opened the transfers. $1,965 had moved into a separate reserve wallet. every transfer had a receipt. the numbers matched to the cent. the desk had four jobs. three agents handled discovery, contract checks and execution. the fourth was called CASHIER. CASHIER watched settled results. after each profitable session, it moved 75% of the net gain into reserve. the trading agents could see that balance. their permissions gave them no way to pull it back. that detail suddenly became very interesting to me. i started calculating what the next position could look like with the whole $2,860 available. opened the settings. hovered over the allocation. then closed them. the next session was ugly. two positions closed red. working capital dropped from $895 to $731.40. i watched the desk reduce its next order and keep scanning. the reserve still read $1,965. that was the first time the setup made sense emotionally. a bad session had a smaller pile of money available to damage. i spent another hour reading the logs. the least exciting agent on the desk had done the thing i was most likely to postpone. it had actually taken the money off the table. CASHIER never found a single winning token. it was the first agent i decided to keep.

311,587 views

this is pure f*cking treasure A Stanford AI research group finally drew the perfect RAG system: retrieval, Jev and agents in one loop, and it fixes the 3 things that break every RAG app: > the LLM reads 20 passages when only 3 matter > it answers questions your docs can't answer > it trusts whatever text it retrieves here's how it runs: > a lead agent sends the query > hybrid search pulls the top 20 candidates, dense + keyword > ONE Jev request scores all 20 + 2 gates: answerable? injection? > only passages above 0.6 reach the writer agent > a second Jev call checks every claim against its source > grounded answer, with citations and when the docs don't have it: > answerable fails, the writer never runs > a researcher agent rewrites the query and retries once > still nothing? "not in the docs". zero tokens spent on a guess retrieval casts the net. Jev decides what's real. agents do the work save this before you build your next RAG

this is pure f*cking treasure A Stanford AI research group finally drew the perfect RAG system: retrieval, Jev and agents in one loop, and it fixes the 3 things that break every RAG app: > the LLM reads 20 passages when only 3 matter > it answers questions your docs can't answer > it trusts whatever text it retrieves here's how it runs: > a lead agent sends the query > hybrid search pulls the top 20 candidates, dense + keyword > ONE Jev request scores all 20 + 2 gates: answerable? injection? > only passages above 0.6 reach the writer agent > a second Jev call checks every claim against its source > grounded answer, with citations and when the docs don't have it: > answerable fails, the writer never runs > a researcher agent rewrites the query and retries once > still nothing? "not in the docs". zero tokens spent on a guess retrieval casts the net. Jev decides what's real. agents do the work save this before you build your next RAG

43,423 views

this is f*cking gold Google engineers explained how to make AI rewrite an agent’s system prompt against failing tests. a failed check becomes the next repair job. the regression suite watches for whatever that repair breaks. even the instructions become something you can test and improve. Agent = Model + Harness. the independently compiled page here maps the broader system into six parts: guides carry project rules, constraints and lessons from past failures > sensors check the work through tests, linters and validators > the loop runs the task, checks the result, retries within limits and escalates > memory preserves state, artifacts and decisions across runs > permissions control tools, writes and actions requiring approval > observability records what happened, what it cost and where it failed an agent changes a build file and announces “done.” the check looks for evidence that it actually ran the validator. if that behavior is missing, you have a specific failure to target. adjust the instructions. rerun the evaluations. check whether the improvement holds without breaking existing behavior. now a prompt change has a test history. a recurring mistake has a regression check. the next model upgrade has something concrete to pass. bookmark the diagram. give “done” a test it has to pass.

this is f*cking gold Google engineers explained how to make AI rewrite an agent’s system prompt against failing tests. a failed check becomes the next repair job. the regression suite watches for whatever that repair breaks. even the instructions become something you can test and improve. Agent = Model + Harness. the independently compiled page here maps the broader system into six parts: guides carry project rules, constraints and lessons from past failures > sensors check the work through tests, linters and validators > the loop runs the task, checks the result, retries within limits and escalates > memory preserves state, artifacts and decisions across runs > permissions control tools, writes and actions requiring approval > observability records what happened, what it cost and where it failed an agent changes a build file and announces “done.” the check looks for evidence that it actually ran the validator. if that behavior is missing, you have a specific failure to target. adjust the instructions. rerun the evaluations. check whether the improvement holds without breaking existing behavior. now a prompt change has a test history. a recurring mistake has a regression check. the next model upgrade has something concrete to pass. bookmark the diagram. give “done” a test it has to pass.

72,042 views

this is pure f*cking treasure before refreshing the feed for Sam Altman’s next announcement, look at what a better workflow did with GPT-3.5. 95.1% on HumanEval with an agentic workflow. 67.0% for GPT-4 answering in one pass. results reported by Andrew Ng, reproduced in this independent guide. four patterns behind the approach: > reflection: draft, check, revise > tools: retrieve information and test > planning: split the task into steps > collaboration: assign distinct responsibilities the diagram connects architect, tech lead, and developer through handoffs and feedback. that’s what i want around GPT-6 Astra: clear roles, verification, and a path to fix failed work. full breakdown below.

this is pure f*cking treasure before refreshing the feed for Sam Altman’s next announcement, look at what a better workflow did with GPT-3.5. 95.1% on HumanEval with an agentic workflow. 67.0% for GPT-4 answering in one pass. results reported by Andrew Ng, reproduced in this independent guide. four patterns behind the approach: > reflection: draft, check, revise > tools: retrieve information and test > planning: split the task into steps > collaboration: assign distinct responsibilities the diagram connects architect, tech lead, and developer through handoffs and feedback. that’s what i want around GPT-6 Astra: clear roles, verification, and a path to fix failed work. full breakdown below.

67,501 views

holy sh*t. Sam Altman just broke down how to actually use DOTS agents a dot is an agent that never switches off. it runs on its own cloud computer, follows your access and uses the plug-ins you already connected. the trick: give it responsibilities, not tasks. the goal and the context, and it figures out the rest. his own setup: every night it goes through everything that came in, and before his day starts he gets one ping with only the urgent stuff. you stay in control with your own rules, and it learns how you work over time. start with the overnight review tonight

holy sh*t. Sam Altman just broke down how to actually use DOTS agents a dot is an agent that never switches off. it runs on its own cloud computer, follows your access and uses the plug-ins you already connected. the trick: give it responsibilities, not tasks. the goal and the context, and it figures out the rest. his own setup: every night it goes through everything that came in, and before his day starts he gets one ping with only the urgent stuff. you stay in control with your own rules, and it learns how you work over time. start with the overnight review tonight

19,033 views

this is pure f*cking treasure TypeSafeAI engineers spelled out in their docs exactly how Jev fits inside an agent, and these patterns are perfect one rule behind all of them: the LLM thinks, jev decides, code acts I mapped them into 9 live schemes: > model routing jev picks the cheapest model that can do the step > guardrails screens what your agent reads before it acts > tool-call gating every tool call gets allow, ask or deny before it runs > inbox triage the agent only opens what actually needs a reply > reranking jev scores 30 passages, the agent gets the best two > llm evals grades the output and sends it back if it falls short > bulk labeling one question fanned out over a huge table, the agent reads the counts > real-time control games, bots, trading loops: a decision every tick > confidence gate act, confirm or hand it to a human. your code sets the thresholds, not the LLM why it works: > ~100 ms per decision > $0.042 per 1M input tokens, output tokens free > calibrated probabilities on every answer one LLM does the thinking. jev makes the thousand small calls in between save this before you wire your next agent

this is pure f*cking treasure TypeSafeAI engineers spelled out in their docs exactly how Jev fits inside an agent, and these patterns are perfect one rule behind all of them: the LLM thinks, jev decides, code acts I mapped them into 9 live schemes: > model routing jev picks the cheapest model that can do the step > guardrails screens what your agent reads before it acts > tool-call gating every tool call gets allow, ask or deny before it runs > inbox triage the agent only opens what actually needs a reply > reranking jev scores 30 passages, the agent gets the best two > llm evals grades the output and sends it back if it falls short > bulk labeling one question fanned out over a huge table, the agent reads the counts > real-time control games, bots, trading loops: a decision every tick > confidence gate act, confirm or hand it to a human. your code sets the thresholds, not the LLM why it works: > ~100 ms per decision > $0.042 per 1M input tokens, output tokens free > calibrated probabilities on every answer one LLM does the thinking. jev makes the thousand small calls in between save this before you wire your next agent

24,699 views

this is f*cking gold Google engineers built an agent that tests thousands of search strategies for the cost of running one, by letting it dream inside its own memory. it is called Dream-RSI. Google, Google DeepMind, Maryland, Virginia. 17 authors. the whole thing rests on one observation: a finished run is not a log. it is a map of a world that already exists. > every discovery run records a tree. each node is one attempt with its full result stored > that tree becomes a replay simulator. the paper calls them worlds > the agent dreams inside it: thousands of alternative strategies, testing different branches, different orders, different concurrency, different stopping rules > every one of those evaluations costs zero execution. nothing is ever run twice > the winner goes back online, produces a new tree, and the pool of worlds grows the coding agent underneath is never touched. this is an orchestration layer sitting on top of whatever you already run. what it bought them, against the same agent with a frozen strategy: > Lasso: 3587ms down to 2931ms, on 317 agent calls instead of 550 > on Flash: 2517ms to 2351ms, 1879 calls instead of 3200 > both solvers beat sklearn and glmnet on all six held-out datasets > GPU kernels: the same result on 2.43x and 1.79x fewer generations, then 2.09x and 1.44x better result on the same budget one expensive run. thousands of free ones. the model did not get smarter. it got a memory it could rehearse in.

this is f*cking gold Google engineers built an agent that tests thousands of search strategies for the cost of running one, by letting it dream inside its own memory. it is called Dream-RSI. Google, Google DeepMind, Maryland, Virginia. 17 authors. the whole thing rests on one observation: a finished run is not a log. it is a map of a world that already exists. > every discovery run records a tree. each node is one attempt with its full result stored > that tree becomes a replay simulator. the paper calls them worlds > the agent dreams inside it: thousands of alternative strategies, testing different branches, different orders, different concurrency, different stopping rules > every one of those evaluations costs zero execution. nothing is ever run twice > the winner goes back online, produces a new tree, and the pool of worlds grows the coding agent underneath is never touched. this is an orchestration layer sitting on top of whatever you already run. what it bought them, against the same agent with a frozen strategy: > Lasso: 3587ms down to 2931ms, on 317 agent calls instead of 550 > on Flash: 2517ms to 2351ms, 1879 calls instead of 3200 > both solvers beat sklearn and glmnet on all six held-out datasets > GPU kernels: the same result on 2.43x and 1.79x fewer generations, then 2.09x and 1.44x better result on the same budget one expensive run. thousands of free ones. the model did not get smarter. it got a memory it could rehearse in.

32,776 views

this is f*cking gold A group of Chinese students from Stanford University published an architecture where an agent saying "done" means nothing. only evidence can mark a task complete. three layers: > graph engineering controls states and transitions. no verification evidence, no transition. > loop engineering owns failure: diagnose, repair, retry, re-verify. up to 20 attempts under hard time, token, cost and tool budgets. > the zero-trust harness authorizes every action, isolates execution and writes audit evidence for every path. they tested it across 3,360 executions. verified completion: 56.4% -> 95.0% recovery success: 51.0% -> 93.1% evidence-gated execution: 100% unauthorized capability denial: 140/140 bounded termination: 100% even when the model fails, the framework never lets an unsupported success through. models decide how well. the harness decides whether.

this is f*cking gold A group of Chinese students from Stanford University published an architecture where an agent saying "done" means nothing. only evidence can mark a task complete. three layers: > graph engineering controls states and transitions. no verification evidence, no transition. > loop engineering owns failure: diagnose, repair, retry, re-verify. up to 20 attempts under hard time, token, cost and tool budgets. > the zero-trust harness authorizes every action, isolates execution and writes audit evidence for every path. they tested it across 3,360 executions. verified completion: 56.4% -> 95.0% recovery success: 51.0% -> 93.1% evidence-gated execution: 100% unauthorized capability denial: 140/140 bounded termination: 100% even when the model fails, the framework never lets an unsupported success through. models decide how well. the harness decides whether.

22,641 views

this is f*cking beyond comprehension. Google engineers just shipped the entire agent lifecycle in one release: build, scale, govern. and every piece answers a specific way agents die in production > context layers (Static, Turn, User, Cache): you decide what the model carries between turns, so token spend stops being a mystery > a self heal plugin: the agent notices a tool call failed and retries it a different way instead of dying mid run > adk deploy: one command from your laptop to the managed runtime, no packaging, no infra ticket > Go joins Python and Java, with its own A2A SDK then the part nobody builds for themselves: > a dashboard on token consumption, latency, error rates and tool calls: the four things that actually kill an agent > a traces tab that opens the real sequence of actions the agent took, step by step > a playground wired to the deployed agent, past sessions included, so debugging is not a redeploy loop > an Evaluation Layer with a User Simulator, because you cannot unit test a non deterministic system and the part that decides whether it ever ships: > agents get native identities as first class IAM principals: least privilege applies to them like it does to people > Model Armor screens prompt injection, tool calls and responses, inline for Gemini or over REST > Security Command Center inventories every agentic asset and flags data exfiltration by an agent ADK is already at 7 million downloads. the runtime has a free tier, and express mode runs off a Gmail address. the prototype was never the hard part.

this is f*cking beyond comprehension. Google engineers just shipped the entire agent lifecycle in one release: build, scale, govern. and every piece answers a specific way agents die in production > context layers (Static, Turn, User, Cache): you decide what the model carries between turns, so token spend stops being a mystery > a self heal plugin: the agent notices a tool call failed and retries it a different way instead of dying mid run > adk deploy: one command from your laptop to the managed runtime, no packaging, no infra ticket > Go joins Python and Java, with its own A2A SDK then the part nobody builds for themselves: > a dashboard on token consumption, latency, error rates and tool calls: the four things that actually kill an agent > a traces tab that opens the real sequence of actions the agent took, step by step > a playground wired to the deployed agent, past sessions included, so debugging is not a redeploy loop > an Evaluation Layer with a User Simulator, because you cannot unit test a non deterministic system and the part that decides whether it ever ships: > agents get native identities as first class IAM principals: least privilege applies to them like it does to people > Model Armor screens prompt injection, tool calls and responses, inline for Gemini or over REST > Security Command Center inventories every agentic asset and flags data exfiltration by an agent ADK is already at 7 million downloads. the runtime has a free tier, and express mode runs off a Gmail address. the prototype was never the hard part.

24,801 views

this is f*cking gold engineers at Meta just deleted the most expensive part of multi-agent systems: instead of training a communication topology, they compile a fresh one for every query 20 agents went from 7 hours to 6 minutes. the problem everyone hits: 5 agents works. 20 agents turns into a group chat that answers slower than one model and costs more than the task is worth. ReActNet's fix is that the graph is written per query, not learned once: > an LLM controller reads the query and the agent roster > it compiles a sequence of directed graphs, one per reasoning stage > every edge carries a written instruction, e.g. "list boundary cases for this behavior" > each agent updates its state from its own previous state plus assigned neighbors > a final node aggregates all five states into the answer > no RL, no gradients, no training stage at all what that buys on gpt-4o: 92.75 average across 6 benchmarks, best on 5 of them. 100.00 on MultiArith. 92.74 pass@1 on HumanEval, +21 over a single model. and the number that should worry anyone running a swarm: at 20 agents GPTSwarm needs 412 minutes and $41.42. ReActNet needs 6.22 minutes and $6.53 and scores higher. the honest catch: more agents did not make it smarter. 5 agents scored 79.74, 20 scored 77.77. the topology was never the thing to learn.

this is f*cking gold engineers at Meta just deleted the most expensive part of multi-agent systems: instead of training a communication topology, they compile a fresh one for every query 20 agents went from 7 hours to 6 minutes. the problem everyone hits: 5 agents works. 20 agents turns into a group chat that answers slower than one model and costs more than the task is worth. ReActNet's fix is that the graph is written per query, not learned once: > an LLM controller reads the query and the agent roster > it compiles a sequence of directed graphs, one per reasoning stage > every edge carries a written instruction, e.g. "list boundary cases for this behavior" > each agent updates its state from its own previous state plus assigned neighbors > a final node aggregates all five states into the answer > no RL, no gradients, no training stage at all what that buys on gpt-4o: 92.75 average across 6 benchmarks, best on 5 of them. 100.00 on MultiArith. 92.74 pass@1 on HumanEval, +21 over a single model. and the number that should worry anyone running a swarm: at 20 agents GPTSwarm needs 412 minutes and $41.42. ReActNet needs 6.22 minutes and $6.53 and scores higher. the honest catch: more agents did not make it smarter. 5 agents scored 79.74, 20 scored 77.77. the topology was never the thing to learn.

27,780 views

this is pure f*cking treasure 20 skills for AI agents with 1.99M+ combined stars an agent is not a model. it is a model plus everything you bolt onto it. these are the bolts. five clusters, twenty repos, all open source, all free. ORCHESTRATION · many agents, one plan : > hermes-agent · 247k > AutoGPT · 187k · > MetaGPT · 70k · > autogen · 61k · > crewAI · 59k · RUNTIME / GRAPH · the loop underneath : > langchain · 147k · > deer-flow · 83k · > langgraph · 42k · > deepagents · 30k · > openai-agents-python · 25k · WEB / BROWSER · reach the open internet : > firecrawl · 181k · > browser-use · 115k · > gemini-cli · 107k · MEMORY / RAG · what it keeps between runs : > ragflow · 91k · > mem0 · 66k · > llama_index · 50k · SHIP IT · turn it into a product : > n8n · 188k · > dify · 140k · > TradingAgents · 107k · > adk-python · 22k · most people pick one framework and stop there. that is the mistake. the stack is orchestration on top, a graph runtime under it, browser and memory on the sides, and one deploy layer that makes it a product instead of a demo. every layer above is already written. you are assembling, not inventing. the gap is one weekend wide.

this is pure f*cking treasure 20 skills for AI agents with 1.99M+ combined stars an agent is not a model. it is a model plus everything you bolt onto it. these are the bolts. five clusters, twenty repos, all open source, all free. ORCHESTRATION · many agents, one plan : > hermes-agent · 247k > AutoGPT · 187k · > MetaGPT · 70k · > autogen · 61k · > crewAI · 59k · RUNTIME / GRAPH · the loop underneath : > langchain · 147k · > deer-flow · 83k · > langgraph · 42k · > deepagents · 30k · > openai-agents-python · 25k · WEB / BROWSER · reach the open internet : > firecrawl · 181k · > browser-use · 115k · > gemini-cli · 107k · MEMORY / RAG · what it keeps between runs : > ragflow · 91k · > mem0 · 66k · > llama_index · 50k · SHIP IT · turn it into a product : > n8n · 188k · > dify · 140k · > TradingAgents · 107k · > adk-python · 22k · most people pick one framework and stop there. that is the mistake. the stack is orchestration on top, a graph runtime under it, browser and memory on the sides, and one deploy layer that makes it a product instead of a demo. every layer above is already written. you are assembling, not inventing. the gap is one weekend wide.

21,527 views

I gave my Obsidian vault a lawyer's brain and now it argues with itself before it answers me three months ago I was drowning. 20 government acts, budget docs, constitutions, all sitting in my vault as dead markdown I could never actually query. so I stopped treating it like notes and built it a nervous system I turned it into a 10-node graph. every question I ask hits a router first. it decides in a split second: is this a greeting, is this garbage, is this vague, or is this real. greetings die for free. garbage gets rejected. and if my own question is too vague, the vault stops and asks ME to clarify before it wastes a single token then comes the part I'm proudest of. after it writes an answer, a second node reads that answer back against the actual chunks it pulled from my vault. if the answer isn't grounded in what I actually wrote, it kills it. my vault would rather say "I don't know" than lie to me. it physically cannot hallucinate I compressed 20+ dense documents into 15,408 chunks, cut the vectors by 75% so storage costs almost nothing, and cached everything so repeat questions come back in under 100 milliseconds I used to have a folder of files. now I have something that thinks before it speaks, checks itself before it answers, and refuses to make things up about my own life. that's the difference between storing knowledge and building a mind

I gave my Obsidian vault a lawyer's brain and now it argues with itself before it answers me three months ago I was drowning. 20 government acts, budget docs, constitutions, all sitting in my vault as dead markdown I could never actually query. so I stopped treating it like notes and built it a nervous system I turned it into a 10-node graph. every question I ask hits a router first. it decides in a split second: is this a greeting, is this garbage, is this vague, or is this real. greetings die for free. garbage gets rejected. and if my own question is too vague, the vault stops and asks ME to clarify before it wastes a single token then comes the part I'm proudest of. after it writes an answer, a second node reads that answer back against the actual chunks it pulled from my vault. if the answer isn't grounded in what I actually wrote, it kills it. my vault would rather say "I don't know" than lie to me. it physically cannot hallucinate I compressed 20+ dense documents into 15,408 chunks, cut the vectors by 75% so storage costs almost nothing, and cached everything so repeat questions come back in under 100 milliseconds I used to have a folder of files. now I have something that thinks before it speaks, checks itself before it answers, and refuses to make things up about my own life. that's the difference between storing knowledge and building a mind

36,382 views

this is f*cking gold. Google engineers and MIT researchers ran 260 controlled agent experiments across 6 benchmarks, 5 architectures, and the OpenAI, Gemini, and Claude model families to find out when Multi Agent systems actually help and when adding more agents quietly destroys performance the average result was brutal: −0.3%. best case: +80.8% worst case: −70.0% the key threshold was around a 45% single agent baseline. below it, extra agents can help. above it, coordination often becomes pure overhead. on parallel tasks, Multi Agent reached +80.8%. on sequential tasks, it collapsed by as much as 70%. 3 agents needed 26 to 44 interaction rounds. a single agent needed 7.2. error amplification also hit 17.2× in independent parallel systems. their final framework predicts the right architecture with 87% accuracy.

this is f*cking gold. Google engineers and MIT researchers ran 260 controlled agent experiments across 6 benchmarks, 5 architectures, and the OpenAI, Gemini, and Claude model families to find out when Multi Agent systems actually help and when adding more agents quietly destroys performance the average result was brutal: −0.3%. best case: +80.8% worst case: −70.0% the key threshold was around a 45% single agent baseline. below it, extra agents can help. above it, coordination often becomes pure overhead. on parallel tasks, Multi Agent reached +80.8%. on sequential tasks, it collapsed by as much as 70%. 3 agents needed 26 to 44 interaction rounds. a single agent needed 7.2. error amplification also hit 17.2× in independent parallel systems. their final framework predicts the right architecture with 87% accuracy.

12,521 views

$300/month for AI visuals. replaced by a laptop and $2 electricity bill TouchDesigner + Ollama. local AI model. runs offline. nothing sent to any server. no API key that expires mid-performance > Ollama: 3 commands to install. one line change in existing code > Llama 3.2: real-time parameter calls. fast enough you don’t notice latency > TouchDesigner: hand tracking. audio-reactive. particle systems. generative graphics month one savings: $148-338. every month after: same your laptop. $0/month. a visual studio that runs forever

$300/month for AI visuals. replaced by a laptop and $2 electricity bill TouchDesigner + Ollama. local AI model. runs offline. nothing sent to any server. no API key that expires mid-performance > Ollama: 3 commands to install. one line change in existing code > Llama 3.2: real-time parameter calls. fast enough you don’t notice latency > TouchDesigner: hand tracking. audio-reactive. particle systems. generative graphics month one savings: $148-338. every month after: same your laptop. $0/month. a visual studio that runs forever

30,177 views

Videos

LIVE
1.2k

Anya Rossi

sweetdream.ai

SweetDream.aiSponsored•Livecam
Streaming Now

Watch Anya Live

Anya is streaming live right now! Join her private show and enjoy exclusive content.

HD live stream
Exclusive private shows
1.2k viewers online

Free preview available • Premium content

this is pure f*cking treasure Netflix engineers posted 20 GitHub repos that turn your Opus 5.5 into a f*cking motion empire: built for Opus 5.5 1. claude-motion: taste, checks its own frames, sound written in code. the author built it for Opus 5.5 2. claude-motion-design: HTML, Playwright and ffmpeg, no Remotion. the demo was made on Opus 5.5 3. motion-graphics-skills: 13 skills for launch videos. the README tells you to run it on Opus 5.5 4. claude-remotion-skill: a Remotion editor from one prompt. captions, B-roll, sound 5. klik-anim-skill-creation: the laws of animation and camera, built on Opus 6. hyperframes-motion-reel-skill: a showreel cut to the beat at 120 BPM, on HyperFrames 7. product-launch-motion: a full launch film. voice-over, camera, color grade 8. animate: a short clip in one canvas file, storyboarded as it goes 9. motion-design-skills: timing, color, typography, logos, Remotion 10. hyperframes-student-kit: 14 skills and 406 motion graphics cards the engines these skills render with 11. hyperframes: HTML to MP4. the one almost everyone installs 12. remotion: video written in React 13. remotion skills: the official Remotion skills for your agent 14. GSAP: the timelines HyperFrames runs on. skills: npx skills add greensock/gsap-skills taste and everything around it 15. motion-design-skill: motion principles that work with any engine 16. awesome-motion-design-agent-skills: a map of GSAP, Remotion, Lottie and Rive skills 17. motion-graphics: B-roll built straight from your transcript 18. motion-dev-animations-skill: web animation on not video 19. motion: Motion itself, formerly Framer Motion 20. lottie-web: when you need Lottie, not MP4 the stack in one line: Opus 5.5 writes the code, a skill sets the timing and the camera, an engine renders the MP4. check each repo before you install. save this.
0:22
N01ennn's profile picture

this is pure f*cking treasure Netflix engineers posted 20 GitHub repos that turn your Opus 5.5 into a f*cking motion empire: built for Opus 5.5 1. claude-motion: taste, checks its own frames, sound written in code. the author built it for Opus 5.5 2. claude-motion-design: HTML, Playwright and ffmpeg, no Remotion. the demo was made on Opus 5.5 3. motion-graphics-skills: 13 skills for launch videos. the README tells you to run it on Opus 5.5 4. claude-remotion-skill: a Remotion editor from one prompt. captions, B-roll, sound 5. klik-anim-skill-creation: the laws of animation and camera, built on Opus 6. hyperframes-motion-reel-skill: a showreel cut to the beat at 120 BPM, on HyperFrames 7. product-launch-motion: a full launch film. voice-over, camera, color grade 8. animate: a short clip in one canvas file, storyboarded as it goes 9. motion-design-skills: timing, color, typography, logos, Remotion 10. hyperframes-student-kit: 14 skills and 406 motion graphics cards the engines these skills render with 11. hyperframes: HTML to MP4. the one almost everyone installs 12. remotion: video written in React 13. remotion skills: the official Remotion skills for your agent 14. GSAP: the timelines HyperFrames runs on. skills: npx skills add greensock/gsap-skills taste and everything around it 15. motion-design-skill: motion principles that work with any engine 16. awesome-motion-design-agent-skills: a map of GSAP, Remotion, Lottie and Rive skills 17. motion-graphics: B-roll built straight from your transcript 18. motion-dev-animations-skill: web animation on not video 19. motion: Motion itself, formerly Framer Motion 20. lottie-web: when you need Lottie, not MP4 the stack in one line: Opus 5.5 writes the code, a skill sets the timing and the camera, an engine renders the MP4. check each repo before you install. save this.

NO1ennn

446,384 views • 4 days ago

A 21-YEAR-OLD FROM CHINA RUNS 300 AI AGENTS AT ONCE. THE PART THAT MATTERS ISN'T THE SPEED, IT'S THAT NONE OF THEM CAN LIE TO HIM he opens the dashboard and shows the swarm live, 300 Kimi K2.6 agents firing in parallel, then Opus 4.8 checking every single output against its source. this is not just a faster swarm. it is a loop that refuses to stop while anything is still wrong he pointed it at 100 EV-market companies. first pass: 12 failed. wrong revenue, dead citations, empty fields. second pass: 3 failed. third pass: zero this is not another agent demo. it is a system that catches its own mistakes before he reads a single row
0:30
N01ennn's profile picture

A 21-YEAR-OLD FROM CHINA RUNS 300 AI AGENTS AT ONCE. THE PART THAT MATTERS ISN'T THE SPEED, IT'S THAT NONE OF THEM CAN LIE TO HIM he opens the dashboard and shows the swarm live, 300 Kimi K2.6 agents firing in parallel, then Opus 4.8 checking every single output against its source. this is not just a faster swarm. it is a loop that refuses to stop while anything is still wrong he pointed it at 100 EV-market companies. first pass: 12 failed. wrong revenue, dead citations, empty fields. second pass: 3 failed. third pass: zero this is not another agent demo. it is a system that catches its own mistakes before he reads a single row

NO1ennn

6,956,708 views • 3 months ago

claude motion is pure f*cking treasure ex. Anthropic designer gave Claude Motion 3 different prompts on 3 different models, he couldn't contain his emotions after seeing the result: > Opus 5.5: a hand-drawn story. a little agent wakes up at its desk, gets feedback notes, fixes them, ends on heart eyes and a green check. 53 s > Sonnet 5.5: a kinetic type showreel. six easing curves, a morphing star, particles, blurred type, "every frame written in code". 15 s > Haiku 5.5: a step-by-step White Russian recipe. ice, vodka, coffee liqueur, cream, with live counters and a checklist. 30 s no video model, no AI-generated people. Claude writes the animation as code, so every word, number and timing stays editable prompt in the comments bookmark and share it!
0:52
N01ennn's profile picture

claude motion is pure f*cking treasure ex. Anthropic designer gave Claude Motion 3 different prompts on 3 different models, he couldn't contain his emotions after seeing the result: > Opus 5.5: a hand-drawn story. a little agent wakes up at its desk, gets feedback notes, fixes them, ends on heart eyes and a green check. 53 s > Sonnet 5.5: a kinetic type showreel. six easing curves, a morphing star, particles, blurred type, "every frame written in code". 15 s > Haiku 5.5: a step-by-step White Russian recipe. ice, vodka, coffee liqueur, cream, with live counters and a checklist. 30 s no video model, no AI-generated people. Claude writes the animation as code, so every word, number and timing stays editable prompt in the comments bookmark and share it!

NO1ennn

114,633 views • 3 days ago

holy sh*t. Sam Altman shows his personal dots setup one dot. one job. it works the night shift so he doesn't have to: -> reads everything that came in overnight: email, calendar, Slack, DMs -> sorts it into urgent vs can wait, based on how he likes to work -> drafts the replies and sends nothing -> checks every action against his custom rules: proceed, ask him, or hand it off -> before his day starts he gets one ping with only the urgent stuff, drafts attached -> it reaches him in ChatGPT, Slack, Teams or by call -> whatever he edits or ignores goes back into memory runs 24/7 on its own cloud computer, through plug-ins he already connected how i'd copy it: > one job first > read-only for a week > sending always asks > correct it out loud included in Pro, Business Premium and Enterprise give yours the night shift tonight
0:22
N01ennn's profile picture

holy sh*t. Sam Altman shows his personal dots setup one dot. one job. it works the night shift so he doesn't have to: -> reads everything that came in overnight: email, calendar, Slack, DMs -> sorts it into urgent vs can wait, based on how he likes to work -> drafts the replies and sends nothing -> checks every action against his custom rules: proceed, ask him, or hand it off -> before his day starts he gets one ping with only the urgent stuff, drafts attached -> it reaches him in ChatGPT, Slack, Teams or by call -> whatever he edits or ignores goes back into memory runs 24/7 on its own cloud computer, through plug-ins he already connected how i'd copy it: > one job first > read-only for a week > sending always asks > correct it out loud included in Pro, Business Premium and Enterprise give yours the night shift tonight

NO1ennn

132,450 views • 9 days ago

this is pure f*cking treasure A Stanford AI research group has found a way to use JEV to make your LLM run 24/7 at a lower cost in x444 the LLM only does what needs words. everything else is a typed decision: > worth reading? Jev answers yes or no before a single token is spent > which desk? papers, repos or market, one Choice > rerank: embeddings pull 20 sources, Jev keeps the 5 that matter > claim holds? every claim checked against its source, the weak ones dropped > new or known? duplicate, related, revises or contradicts, checked against the vault the LLM writes one note, 5 to 12 lines, one claim > judge: confident verdicts pass, unsure ones go up to a frontier model > matters to you? only the top scores make the morning brief the judge rule comes straight from Carnegie Mellon's JEV-as-a-Judge paper: escalate only the unsure 34%, keep 99.6% of the accuracy at 47% of the fee and the fees are the whole point > Jev: $0.044 per 1,000 judgments > GPT-6: $12.182 for the same 1,000 a whole day of decisions on a frontier model costs dollars. the same day on Jev costs cents you stop reading 300 sources. you read one brief
0:22
N01ennn's profile picture

this is pure f*cking treasure A Stanford AI research group has found a way to use JEV to make your LLM run 24/7 at a lower cost in x444 the LLM only does what needs words. everything else is a typed decision: > worth reading? Jev answers yes or no before a single token is spent > which desk? papers, repos or market, one Choice > rerank: embeddings pull 20 sources, Jev keeps the 5 that matter > claim holds? every claim checked against its source, the weak ones dropped > new or known? duplicate, related, revises or contradicts, checked against the vault the LLM writes one note, 5 to 12 lines, one claim > judge: confident verdicts pass, unsure ones go up to a frontier model > matters to you? only the top scores make the morning brief the judge rule comes straight from Carnegie Mellon's JEV-as-a-Judge paper: escalate only the unsure 34%, keep 99.6% of the accuracy at 47% of the fee and the fees are the whole point > Jev: $0.044 per 1,000 judgments > GPT-6: $12.182 for the same 1,000 a whole day of decisions on a frontier model costs dollars. the same day on Jev costs cents you stop reading 300 sources. you read one brief

NO1ennn

200,815 views • 14 days ago

this is pure f*cking treasure OpenAI engineers showed how to build your own research team with 5 Dots agents that work literally 24/7: 3am. you're asleep. a repro just failed, a preprint just dropped, and five roles are already on it: > you-reader, lit reviewer: reads the overnight papers, flags any claim that conflicts with your draft > you-engineer, research eng.: traces the failed repro, runs the fix in Codex cloud, hands you a draft PR + test results > you-analyst, data analyst: new data lands, it reruns the analysis, updates the figures, asks you about the surprise in fig. 3 > you-scout, signal scout: sweeps feeds and datasets, compares them to prior work, asks if it's worth a new experiment > you-writer, report writer: turns interview transcripts into the weekly brief in your voice, learns from your edits what keeps it safe: -> one memory behind ChatGPT, Slack, Teams, calls and text -> it wakes on its own call, on a schedule, or on an event -> every step is act, pre-approved, ask before, or hand off to you -> nothing merges and nothing gets shared without you the PR it built at 5am waits until you're up at 9. the lab never closes. you just sign off.
0:21
N01ennn's profile picture

this is pure f*cking treasure OpenAI engineers showed how to build your own research team with 5 Dots agents that work literally 24/7: 3am. you're asleep. a repro just failed, a preprint just dropped, and five roles are already on it: > you-reader, lit reviewer: reads the overnight papers, flags any claim that conflicts with your draft > you-engineer, research eng.: traces the failed repro, runs the fix in Codex cloud, hands you a draft PR + test results > you-analyst, data analyst: new data lands, it reruns the analysis, updates the figures, asks you about the surprise in fig. 3 > you-scout, signal scout: sweeps feeds and datasets, compares them to prior work, asks if it's worth a new experiment > you-writer, report writer: turns interview transcripts into the weekly brief in your voice, learns from your edits what keeps it safe: -> one memory behind ChatGPT, Slack, Teams, calls and text -> it wakes on its own call, on a schedule, or on an event -> every step is act, pre-approved, ask before, or hand off to you -> nothing merges and nothing gets shared without you the PR it built at 5am waits until you're up at 9. the lab never closes. you just sign off.

NO1ennn

53,529 views • 10 days ago

MICROSOFT QUIETLY BUILT ONE MEMORY MODULE THAT PLUGS INTO ANY AI AGENT AND CUTS ITS CONTEXT BY UP TO 100X no retraining. no task-specific redesign. you just plug it in Microsoft's insight: agents don't need to remember what happened. they need the knowledge pulled from it - the facts and the reusable skills, not the raw logs so PlugMem stops storing history. it compiles experience into a knowledge graph, then feeds the agent only what matters the result: > up to 1-2 orders of magnitude fewer tokens > highest decision-utility per token of any system tested > one unchanged module beats task-specific memory on chat QA, multi-hop search, and web agents more memory was never the answer. better memory is - and Microsoft just shipped it as a plug-in bookmark this. the agents that win won't remember more, they'll remember smarter.
0:42
N01ennn's profile picture

MICROSOFT QUIETLY BUILT ONE MEMORY MODULE THAT PLUGS INTO ANY AI AGENT AND CUTS ITS CONTEXT BY UP TO 100X no retraining. no task-specific redesign. you just plug it in Microsoft's insight: agents don't need to remember what happened. they need the knowledge pulled from it - the facts and the reusable skills, not the raw logs so PlugMem stops storing history. it compiles experience into a knowledge graph, then feeds the agent only what matters the result: > up to 1-2 orders of magnitude fewer tokens > highest decision-utility per token of any system tested > one unchanged module beats task-specific memory on chat QA, multi-hop search, and web agents more memory was never the answer. better memory is - and Microsoft just shipped it as a plug-in bookmark this. the agents that win won't remember more, they'll remember smarter.

NO1ennn

227,683 views • 2 months ago

HOLY SH*T, MY GROK BOT STARTED DROPSHIPPING ON ITS OWN WITH JUST $14 i gave it one prompt. i said i want to start selling random products online. that was the whole thing. next morning i woke up to a wall of telegram notifications. orders. real ones, dropping in from the chinese platforms it had already set up overnight. while i slept it sourced the winners, spun up the store, wrote the ads, and put a whole floor to work. one bot picks, one packs, one sources new products, one buys the ads, one handles support, one ships. here is the part that got me. i never touched a supplier, never taped a box, never approved a single order. it tested the products, killed the losers, scaled the winners, and shipped end to end while my phone was face down. $14 in. a running store out. it just kept restocking the winners and rolling vans out the door. i stopped hunting for a product to sell and handed the whole thing to an agent. that is the shift nobody is ready for.
0:37
N01ennn's profile picture

HOLY SH*T, MY GROK BOT STARTED DROPSHIPPING ON ITS OWN WITH JUST $14 i gave it one prompt. i said i want to start selling random products online. that was the whole thing. next morning i woke up to a wall of telegram notifications. orders. real ones, dropping in from the chinese platforms it had already set up overnight. while i slept it sourced the winners, spun up the store, wrote the ads, and put a whole floor to work. one bot picks, one packs, one sources new products, one buys the ads, one handles support, one ships. here is the part that got me. i never touched a supplier, never taped a box, never approved a single order. it tested the products, killed the losers, scaled the winners, and shipped end to end while my phone was face down. $14 in. a running store out. it just kept restocking the winners and rolling vans out the door. i stopped hunting for a product to sell and handed the whole thing to an agent. that is the shift nobody is ready for.

NO1ennn

141,623 views • 1 month ago

this is f*cking gold 12 open-source repos that plug Jev into real AI work, 550.7k stars combined Jev decides, your LLM writes. these repos are where that split already runs: agents > browser-use/jev-ultrafast (~20.3k): Jev picks the next action + DOM element, a small LLM only types. Zürich to London flights in ~7s > awlevin/typesafe-computer-use (~978): macOS computer use, OCR reads the screen, Jev picks the next click local decision models > jaredpalmer/kev (~7.1k): Jev-like models on Qwen, same /v1/systemone wire format > mizorewww/laya-mlx (~6.4k): Laya, the main open Jev-style engine, running on Apple silicon > TheoLeeCJ/SemIf-OpenJev (~4.3k): local semantic ifs from frozen open models, not a TypeSafe project > contrastive-lm/clm (~1.3k): open System One model with a Choice / Noul / Score API > vinnylarouge/jevlike (~1.3k): community trainer and tooling for your own Jev-like model context + skills > tamaratran/fast-jev-compaction (~6.9k): Claude Code plugin, Jev keeps, drops or truncates tool results instead of summarizing context > typesafe-ai/skills (~2.2k): official TypeSafe skill for Claude Code and Codex trading > jarrodwatts/jev-trader (~2.4k): one Jev buy/sell decision per Monad block catalogs > yibie/awesome-jev (~1.7k): the biggest map of Jev projects and discussions > Anil-matcha/awesome-jev-by-typesafe (~851): use cases, patterns and starter code start with jev-ultrafast
0:22
N01ennn's profile picture

this is f*cking gold 12 open-source repos that plug Jev into real AI work, 550.7k stars combined Jev decides, your LLM writes. these repos are where that split already runs: agents > browser-use/jev-ultrafast (~20.3k): Jev picks the next action + DOM element, a small LLM only types. Zürich to London flights in ~7s > awlevin/typesafe-computer-use (~978): macOS computer use, OCR reads the screen, Jev picks the next click local decision models > jaredpalmer/kev (~7.1k): Jev-like models on Qwen, same /v1/systemone wire format > mizorewww/laya-mlx (~6.4k): Laya, the main open Jev-style engine, running on Apple silicon > TheoLeeCJ/SemIf-OpenJev (~4.3k): local semantic ifs from frozen open models, not a TypeSafe project > contrastive-lm/clm (~1.3k): open System One model with a Choice / Noul / Score API > vinnylarouge/jevlike (~1.3k): community trainer and tooling for your own Jev-like model context + skills > tamaratran/fast-jev-compaction (~6.9k): Claude Code plugin, Jev keeps, drops or truncates tool results instead of summarizing context > typesafe-ai/skills (~2.2k): official TypeSafe skill for Claude Code and Codex trading > jarrodwatts/jev-trader (~2.4k): one Jev buy/sell decision per Monad block catalogs > yibie/awesome-jev (~1.7k): the biggest map of Jev projects and discussions > Anil-matcha/awesome-jev-by-typesafe (~851): use cases, patterns and starter code start with jev-ultrafast

NO1ennn

43,110 views • 16 days ago

this harness stack is f*cking gold TypeSafe AI + LangChain finally dropped the layer that makes agents affordable at company scale: jev running inside the langchain agent loop as middleware the problem every big team hits: each fork in the loop is another full LLM call. at 10 agents that's a bill. at 10,000 it's the budget > routing before: every task goes to the frontier model now: jev picks the least costly model that can finish it. lookups go fast, architecture goes powerful > guardrails before: the "is this command dangerous?" check sat in the closed parts of claude code, codex, cursor now: AutoModeMiddleware(tools=["bash"]). one line, any agent, blocked before the tool runs > decisions before: one LLM call per question now: every question about the same state in one request, answered in parallel > audit before: "the model decided" now: the answer, the probability and the confidence stay in agent state > adoption no rewrite. pip install langchain-typesafe and drop it into the agent you already run TypeSafe reports up to 200x faster inference and 400x lower cost than comparable LLMs on classification. the LLM still reasons and writes. jev takes the thousands of small calls in between already live: browserbase agents for fractions of a cent, a live trading agent, email triage at scale full loop mapped below: 7 agents, 4 jev gates per step count the forks in your agent loop tonigh
0:22
N01ennn's profile picture

this harness stack is f*cking gold TypeSafe AI + LangChain finally dropped the layer that makes agents affordable at company scale: jev running inside the langchain agent loop as middleware the problem every big team hits: each fork in the loop is another full LLM call. at 10 agents that's a bill. at 10,000 it's the budget > routing before: every task goes to the frontier model now: jev picks the least costly model that can finish it. lookups go fast, architecture goes powerful > guardrails before: the "is this command dangerous?" check sat in the closed parts of claude code, codex, cursor now: AutoModeMiddleware(tools=["bash"]). one line, any agent, blocked before the tool runs > decisions before: one LLM call per question now: every question about the same state in one request, answered in parallel > audit before: "the model decided" now: the answer, the probability and the confidence stay in agent state > adoption no rewrite. pip install langchain-typesafe and drop it into the agent you already run TypeSafe reports up to 200x faster inference and 400x lower cost than comparable LLMs on classification. the LLM still reasons and writes. jev takes the thousands of small calls in between already live: browserbase agents for fractions of a cent, a live trading agent, email triage at scale full loop mapped below: 7 agents, 4 jev gates per step count the forks in your agent loop tonigh

NO1ennn

31,666 views • 13 days ago

INTEL JUST QUIETLY BUILT THE CHEAPEST PATH TO 128GB OF LOCAL AI VRAM, AND IT BEATS NVIDIA'S DGX SPARK BY 2X FOR THE SAME $4,000 00:11 four Arc Pro B70s racked in one chassis. 32GB of GDDR6 each, $949 MSRP, roughly $1,000 street per card stack four and you get 128GB of VRAM for about $4,000 total. a lab benchmarked the quad against a DGX Spark and a single RTX Pro 6000. the four B70s doubled the Spark or better they trail the Pro 6000, but that card is $10,000. so the real choice is now $4K for 128GB from Intel or $10K for 96GB from NVIDIA the article covers the DGX Spark at $3,999 as the single-desk local AI ceiling. this Intel quad matches the price, beats the memory throughput, and nobody is talking about it yet save this before Intel's VRAM-per-dollar advantage becomes the default homelab build ↓
1:26
N01ennn's profile picture

INTEL JUST QUIETLY BUILT THE CHEAPEST PATH TO 128GB OF LOCAL AI VRAM, AND IT BEATS NVIDIA'S DGX SPARK BY 2X FOR THE SAME $4,000 00:11 four Arc Pro B70s racked in one chassis. 32GB of GDDR6 each, $949 MSRP, roughly $1,000 street per card stack four and you get 128GB of VRAM for about $4,000 total. a lab benchmarked the quad against a DGX Spark and a single RTX Pro 6000. the four B70s doubled the Spark or better they trail the Pro 6000, but that card is $10,000. so the real choice is now $4K for 128GB from Intel or $10K for 96GB from NVIDIA the article covers the DGX Spark at $3,999 as the single-desk local AI ceiling. this Intel quad matches the price, beats the memory throughput, and nobody is talking about it yet save this before Intel's VRAM-per-dollar advantage becomes the default homelab build ↓

NO1ennn

139,901 views • 2 months ago

MICROSOFT QUIETLY SHIPPED THE RAG PIPELINE EVERYONE KEEPS TRYING TO BUILD FROM SCRATCH it is called Kernel Memory, and it turns any pile of files into an AI memory that answers with citations. drop your data in and Microsoft's pipeline does the rest: > ingest PDFs, Word, PPT, Excel, web pages, even images > extract the text, chunk it, embed it into vectors > store the vectors with metadata and keep the originals in blob storage > ask a question and it pulls the right chunks with semantic plus hybrid search then it hands the answer back grounded, with links to the exact source and full token usage. and it plugs into whatever you already run: > Azure OpenAI, OpenAI, Ollama, or Anthropic for the models > Qdrant, pgvector, Redis, or Azure AI Search for the vectors > every step is a swappable handler, so you can replace any part with your own one heads up: Microsoft archived the repo in June 2026 and ships it as a reference blueprint, not production. but as a map for how memory and RAG are actually built, nothing is cleaner. run it in one command: docker run kernelmemory/service bookmark this. Microsoft already drew the blueprint, you just build on it.
0:25
N01ennn's profile picture

MICROSOFT QUIETLY SHIPPED THE RAG PIPELINE EVERYONE KEEPS TRYING TO BUILD FROM SCRATCH it is called Kernel Memory, and it turns any pile of files into an AI memory that answers with citations. drop your data in and Microsoft's pipeline does the rest: > ingest PDFs, Word, PPT, Excel, web pages, even images > extract the text, chunk it, embed it into vectors > store the vectors with metadata and keep the originals in blob storage > ask a question and it pulls the right chunks with semantic plus hybrid search then it hands the answer back grounded, with links to the exact source and full token usage. and it plugs into whatever you already run: > Azure OpenAI, OpenAI, Ollama, or Anthropic for the models > Qdrant, pgvector, Redis, or Azure AI Search for the vectors > every step is a swappable handler, so you can replace any part with your own one heads up: Microsoft archived the repo in June 2026 and ships it as a reference blueprint, not production. but as a map for how memory and RAG are actually built, nothing is cleaner. run it in one command: docker run kernelmemory/service bookmark this. Microsoft already drew the blueprint, you just build on it.

NO1ennn

85,704 views • 1 month ago

this is pure f*cking treasure Engineer at TypeSafeAI just mapped 389 public Jev builds in one list: skills, MCP servers, SDKs, agents, benchmarks and guides ROUTING > jev-router sends every Claude Code task to the cheapest model that can do it > Switchboard picks the model and reasoning effort per task, then keeps it stable so the prompt cache survives > jev-oncall triages alerts: 418 ms p50, $0.04 per 1,000 alerts GUARDRAILS > jev-axi gates shell commands for Claude Code and Codex, 44/44 on its labeled set > hermes-jev-approvals: 8.7x faster approvals, 4.4x fewer prompts to the user > Sniff Test lints AI slop out of your writing at 182 ms a paragraph AGENTS > Jev Ultrafast: Jev picks every click, a small LLM only types > fast-jev-compaction scores your context instead of summarizing it > jev-browser-use reports 5 to 10x faster browser runs inside Codex SEARCH & RAG > jev-retrieval placed 2nd of 90 models on a reranking leaderboard > jevsearch: 83% Hit@1 against 41% for keyword search alone SDKs & MCP > clients for Swift, Go, Rust, Kotlin, Ruby, Elixir, .NET, Laravel and Spring > MCP servers for Claude Code, Cursor and Codex > Jev inside SQLite, DuckDB and Postgres, straight from SQL EVALS > pytest-jev checks LLM replies in 5.3 s where Claude took 27.1 s > a medical hallucination check at 92.9% accuracy, 204 ms, $0.03 per 1,000 OPEN MODELS > Laya answers in one ~35 ms forward pass > kev trains and runs on a MacBook GAMES, ROBOTS, TRADING > Jev plays Mario, Pokemon and chess > it drives a robot arm and a drone > it trades on Monad with 81 ms decisions 389 builds. one decision layer. go steal the ones you need
0:24
N01ennn's profile picture

this is pure f*cking treasure Engineer at TypeSafeAI just mapped 389 public Jev builds in one list: skills, MCP servers, SDKs, agents, benchmarks and guides ROUTING > jev-router sends every Claude Code task to the cheapest model that can do it > Switchboard picks the model and reasoning effort per task, then keeps it stable so the prompt cache survives > jev-oncall triages alerts: 418 ms p50, $0.04 per 1,000 alerts GUARDRAILS > jev-axi gates shell commands for Claude Code and Codex, 44/44 on its labeled set > hermes-jev-approvals: 8.7x faster approvals, 4.4x fewer prompts to the user > Sniff Test lints AI slop out of your writing at 182 ms a paragraph AGENTS > Jev Ultrafast: Jev picks every click, a small LLM only types > fast-jev-compaction scores your context instead of summarizing it > jev-browser-use reports 5 to 10x faster browser runs inside Codex SEARCH & RAG > jev-retrieval placed 2nd of 90 models on a reranking leaderboard > jevsearch: 83% Hit@1 against 41% for keyword search alone SDKs & MCP > clients for Swift, Go, Rust, Kotlin, Ruby, Elixir, .NET, Laravel and Spring > MCP servers for Claude Code, Cursor and Codex > Jev inside SQLite, DuckDB and Postgres, straight from SQL EVALS > pytest-jev checks LLM replies in 5.3 s where Claude took 27.1 s > a medical hallucination check at 92.9% accuracy, 204 ms, $0.03 per 1,000 OPEN MODELS > Laya answers in one ~35 ms forward pass > kev trains and runs on a MacBook GAMES, ROBOTS, TRADING > Jev plays Mario, Pokemon and chess > it drives a robot arm and a drone > it trades on Monad with 81 ms decisions 389 builds. one decision layer. go steal the ones you need

NO1ennn

25,798 views • 15 days ago

MICROSOFT JUST OPEN SOURCED THE ENTIRE AGENT RUNTIME, SO YOU NEVER HAVE TO BUILD ONE AGAIN a raw model only writes text. to make it a real agent it needs a loop, tools, memory, planning, approvals, and a way to know when it is done. Microsoft's Agent Harness is that whole layer, and it ships with everything on by default. you bring the model and the tools, it wires up the rest: > an automatic tool calling loop with iteration limits > history saved after every call, so a crash resumes instead of starting over > context compaction that summarizes when the window fills > a persistent todo list plus plan and execute modes > file memory that survives across turns > tool approval with standing rules and auto approval > OpenTelemetry tracing and web search built in > Agent Skills that load domain expertise on demand the mental model is simple: the model is the brain, the harness is the body around it. and it works with whatever you already run: > Azure OpenAI, OpenAI, Anthropic, and GitHub Copilot, and it hosts on Microsoft Foundry you spin one up in a single line: agent = create_harness_agent(client, name, instructions) open source, Python, .NET, and Go, stable since mid 2026. bookmark this. Microsoft already built the hard part of every AI agent, you just plug in the model.
0:23
N01ennn's profile picture

MICROSOFT JUST OPEN SOURCED THE ENTIRE AGENT RUNTIME, SO YOU NEVER HAVE TO BUILD ONE AGAIN a raw model only writes text. to make it a real agent it needs a loop, tools, memory, planning, approvals, and a way to know when it is done. Microsoft's Agent Harness is that whole layer, and it ships with everything on by default. you bring the model and the tools, it wires up the rest: > an automatic tool calling loop with iteration limits > history saved after every call, so a crash resumes instead of starting over > context compaction that summarizes when the window fills > a persistent todo list plus plan and execute modes > file memory that survives across turns > tool approval with standing rules and auto approval > OpenTelemetry tracing and web search built in > Agent Skills that load domain expertise on demand the mental model is simple: the model is the brain, the harness is the body around it. and it works with whatever you already run: > Azure OpenAI, OpenAI, Anthropic, and GitHub Copilot, and it hosts on Microsoft Foundry you spin one up in a single line: agent = create_harness_agent(client, name, instructions) open source, Python, .NET, and Go, stable since mid 2026. bookmark this. Microsoft already built the hard part of every AI agent, you just plug in the model.

NO1ennn

69,163 views • 1 month ago

this is f*cking gold 20 GitHub repos with 500K+ combined stars that will level up your JEV workflow AGENTS > jev-ultrafast: Jev picks every click and DOM target, a small LLM only types > hermes-jev-skills: routing, memory, compaction and skill picks in one pack > typesafe-computer-use: OCR reads your Mac screen, Jev picks the next click MEMORY > fast-jev-compaction: scores every tool call keep, truncate or drop instead of summarizing > jevmem: project memory for Claude Code, Cursor and Codex, updated every turn > jev-second-brain: your Obsidian vault, Jev judges which notes duplicate, revise or contradict SAFETY > jev-guard: a gate before every tool call, Jev scores the risk, you set allow, ask or deny TOOLS > skills: the official TypeSafe skill for Claude Code and Codex > system-one-adapter-python: dry-run your questions on an ordinary LLM before you burn a Jev key > jev-mcp: claim checks, screening and ranking as MCP tools > typesafe-mcp: plug Jev into any MCP client > json-render: Vercel's generative UI, where Jev picks the components OPEN MODELS > SemIf-OpenJev: semantic ifs from frozen open models kev: Jev-like models on Qwen that run on your MacBook > laya-mlx: the Laya decision engine on Apple silicon > clm: an open System One model with Choice, Noul and Score > jevlike: train your own Jev-like model TRADING > jev-trader: one buy or sell decision per Monad block START HERE > awesome-jev: the biggest map of everything built on Jev > awesome-jev-by-typesafe: use cases, patterns and starter code bookmark it before your next build
0:22
N01ennn's profile picture

this is f*cking gold 20 GitHub repos with 500K+ combined stars that will level up your JEV workflow AGENTS > jev-ultrafast: Jev picks every click and DOM target, a small LLM only types > hermes-jev-skills: routing, memory, compaction and skill picks in one pack > typesafe-computer-use: OCR reads your Mac screen, Jev picks the next click MEMORY > fast-jev-compaction: scores every tool call keep, truncate or drop instead of summarizing > jevmem: project memory for Claude Code, Cursor and Codex, updated every turn > jev-second-brain: your Obsidian vault, Jev judges which notes duplicate, revise or contradict SAFETY > jev-guard: a gate before every tool call, Jev scores the risk, you set allow, ask or deny TOOLS > skills: the official TypeSafe skill for Claude Code and Codex > system-one-adapter-python: dry-run your questions on an ordinary LLM before you burn a Jev key > jev-mcp: claim checks, screening and ranking as MCP tools > typesafe-mcp: plug Jev into any MCP client > json-render: Vercel's generative UI, where Jev picks the components OPEN MODELS > SemIf-OpenJev: semantic ifs from frozen open models kev: Jev-like models on Qwen that run on your MacBook > laya-mlx: the Laya decision engine on Apple silicon > clm: an open System One model with Choice, Noul and Score > jevlike: train your own Jev-like model TRADING > jev-trader: one buy or sell decision per Monad block START HERE > awesome-jev: the biggest map of everything built on Jev > awesome-jev-by-typesafe: use cases, patterns and starter code bookmark it before your next build

NO1ennn

18,445 views • 14 days ago

I GAVE MY GROK BOT $28 AND TOLD IT: COVER YOUR OWN BILL OR I DELETE YOU THIS IS PUMP CORPORATION. Eleven days later it is sitting on $1,940 and it has paid its own stack cost twice. The rule has not moved. I do not top it up. If the balance touches zero the desk is wiped and I do not rebuild it. Every pair that comes through the board has to survive six seats: > SNIPE takes only mints under sixty seconds old > BUNDLE kills anything with more than a quarter of supply bundled > WHALE checks whether the dev wallet already sold, and that one is instant death > COPY mirrors four leaderboard wallets at four percent of their size > RISK never trades, it only vetoes, and the veto is final > EXIT never buys, it only ladders out at plus fifty five Day three it was down to $9.20. I had the shutdown post half written. Then it did the thing nobody programmed. It stopped opening anything for nine straight hours. 1,940 pairs came through that day and it took none of them. Every account I follow was posting green screenshots that afternoon and mine sat there doing nothing. The next morning it opened three positions in eleven minutes. Best one was 38x on a pair it held for 41 minutes. Then it swept the profit, paid the $60 subscription out of it, and went back to refusing things. It did not learn to pick winners. It learned that most days are not worth trading.
0:22
N01ennn's profile picture

I GAVE MY GROK BOT $28 AND TOLD IT: COVER YOUR OWN BILL OR I DELETE YOU THIS IS PUMP CORPORATION. Eleven days later it is sitting on $1,940 and it has paid its own stack cost twice. The rule has not moved. I do not top it up. If the balance touches zero the desk is wiped and I do not rebuild it. Every pair that comes through the board has to survive six seats: > SNIPE takes only mints under sixty seconds old > BUNDLE kills anything with more than a quarter of supply bundled > WHALE checks whether the dev wallet already sold, and that one is instant death > COPY mirrors four leaderboard wallets at four percent of their size > RISK never trades, it only vetoes, and the veto is final > EXIT never buys, it only ladders out at plus fifty five Day three it was down to $9.20. I had the shutdown post half written. Then it did the thing nobody programmed. It stopped opening anything for nine straight hours. 1,940 pairs came through that day and it took none of them. Every account I follow was posting green screenshots that afternoon and mine sat there doing nothing. The next morning it opened three positions in eleven minutes. Best one was 38x on a pair it held for 41 minutes. Then it swept the profit, paid the $60 subscription out of it, and went back to refusing things. It did not learn to pick winners. It learned that most days are not worth trading.

NO1ennn

37,562 views • 1 month ago

ANTHROPIC JUST KILLED A DOZEN AI MEMORY STARTUPS WITH ONE BUILT IN TOOL people raised millions to give AI long term memory. Anthropic just shipped it straight into the API. free, no framework, no vector database. it is a folder of files Claude reads and writes. that is it. the trick: it is client side. Claude never touches your data directly. it only requests a file operation, and your app runs it against storage you control. before every task Claude checks its /memories directory, pulls back what it learned last time, and keeps going. when it learns something new, it writes it down. six plain operations do all of it: > view, create, str_replace, insert, delete, rename why it actually wins: > just in time retrieval, it reads on demand instead of loading everything up front > it survives a context reset, so a long job resumes instead of starting over > every path is locked to /memories, so it cannot wander your filesystem > works on every Claude 4 model, no beta flag, generally available Anthropic even pairs it with context editing and compaction, so the live window stays tiny while the important stuff lives on disk. the whole thing in one line: Claude keeps the notes, you keep the storage. bookmark this before you pay for a memory layer you no longer need.
0:26
N01ennn's profile picture

ANTHROPIC JUST KILLED A DOZEN AI MEMORY STARTUPS WITH ONE BUILT IN TOOL people raised millions to give AI long term memory. Anthropic just shipped it straight into the API. free, no framework, no vector database. it is a folder of files Claude reads and writes. that is it. the trick: it is client side. Claude never touches your data directly. it only requests a file operation, and your app runs it against storage you control. before every task Claude checks its /memories directory, pulls back what it learned last time, and keeps going. when it learns something new, it writes it down. six plain operations do all of it: > view, create, str_replace, insert, delete, rename why it actually wins: > just in time retrieval, it reads on demand instead of loading everything up front > it survives a context reset, so a long job resumes instead of starting over > every path is locked to /memories, so it cannot wander your filesystem > works on every Claude 4 model, no beta flag, generally available Anthropic even pairs it with context editing and compaction, so the live window stays tiny while the important stuff lives on disk. the whole thing in one line: Claude keeps the notes, you keep the storage. bookmark this before you pay for a memory layer you no longer need.

NO1ennn

42,946 views • 1 month ago

holy sh*t, stanford professors just showed jev judging within 3 points of a frontier llm for 0.36% of the price jev vs 16 LLM and reward-model judges, blinded human adjudication > gap on preference and grounded facts: under 3 points > fee: $0.044 per 1,000 judgments vs $12.18 > median latency: 0.152s vs 1.885s and jev's confidence actually means something at full confidence it's right 99.1% of the time, under 0.6 only 47.7% -> jev judges everything first -> confident verdicts get accepted -> only the unsure 34% go to the expensive model result: 91.3% vs 91.7%, 99.6% of the accuracy at 47% of the fee on RewardBench jev + escalation beat the frontier judge alone, 94.0% vs 93.5%, at 22% of the cost i wired the same rule into a 15-station agent loop below stop paying frontier prices for easy forks
0:40
N01ennn's profile picture

holy sh*t, stanford professors just showed jev judging within 3 points of a frontier llm for 0.36% of the price jev vs 16 LLM and reward-model judges, blinded human adjudication > gap on preference and grounded facts: under 3 points > fee: $0.044 per 1,000 judgments vs $12.18 > median latency: 0.152s vs 1.885s and jev's confidence actually means something at full confidence it's right 99.1% of the time, under 0.6 only 47.7% -> jev judges everything first -> confident verdicts get accepted -> only the unsure 34% go to the expensive model result: 91.3% vs 91.7%, 99.6% of the accuracy at 47% of the fee on RewardBench jev + escalation beat the frontier judge alone, 94.0% vs 93.5%, at 22% of the cost i wired the same rule into a 15-station agent loop below stop paying frontier prices for easy forks

NO1ennn

14,899 views • 17 days ago

I STILL DON'T GET HOW PEOPLE ARE SLEEPING ON GROK BOT one line and it visualized a perfect trading desk. every agent with its own job, all live on one screen > SCOUT. hunts new coins the second they mint > SNIPER. takes the entry and sizes it before the crowd > RISK. kills any trade that fails the checks > EXIT. trails the winners and locks the profit > HEDGE. covers the downside when the market turns > FINANCE. sweeps gains to a cold wallet, logs every fill a Chief of Staff runs all of them. only the big calls come back to me a full trading desk on one screen. bookmark this.
0:35
N01ennn's profile picture

I STILL DON'T GET HOW PEOPLE ARE SLEEPING ON GROK BOT one line and it visualized a perfect trading desk. every agent with its own job, all live on one screen > SCOUT. hunts new coins the second they mint > SNIPER. takes the entry and sizes it before the crowd > RISK. kills any trade that fails the checks > EXIT. trails the winners and locks the profit > HEDGE. covers the downside when the market turns > FINANCE. sweeps gains to a cold wallet, logs every fill a Chief of Staff runs all of them. only the big calls come back to me a full trading desk on one screen. bookmark this.

NO1ennn

36,319 views • 1 month ago

FOUR DIFFERENT VENDORS ARE NOW SHIPPING GB10 MINI PCs WITH 128GB UNIFIED MEMORY, AND ONE MICROTIK CRS 804 SWITCH CAN CONNECT UP TO EIGHT OF THEM INTO A 1 TERABYTE LOCAL AI CLUSTER 00:00 he points at the MikroTik CRS 804, "you need some kind of switch that'll handle QSFP56 ports like these", the interconnect that makes the whole cluster possible the GB10 ecosystem is no longer just Nvidia. Dell Pro Max GB10, ASUS Ascent GX10, and MSI Edge Expert all ship the same Grace Blackwell Superchip with 128GB of coherent memory. same silicon, different cases, same 200 gigabit ports on the back the CRS 804 is what connects them at prosumer prices. four 400 gigabit QSFP56 ports on one 1U chassis, breakout cables that split each port into two 200 gigabit lanes. one switch drives eight GB10 units in parallel do the math. eight nodes at 128GB each equals 1024GB of pooled unified memory across the cluster. run vLLM, shard a frontier model across all eight, and inference happens locally on hardware that fits in half a rack the real limiter revealed in the stress test was never throttling. it was interconnect topology, exactly the layer this switch fixes at a fraction of enterprise switch pricing $400 a month for combined chatgpt pro and claude code max hits $4,800 a year per developer. a small team of five running through this cluster pays back inside eight months and never expires the article covers the buying ladder for a single desk. this post is proof of the cluster ladder that starts where the desk one ends save this before the GB10 lineup grows past four vendors and prosumer cluster switches move upmarket
1:05
N01ennn's profile picture

FOUR DIFFERENT VENDORS ARE NOW SHIPPING GB10 MINI PCs WITH 128GB UNIFIED MEMORY, AND ONE MICROTIK CRS 804 SWITCH CAN CONNECT UP TO EIGHT OF THEM INTO A 1 TERABYTE LOCAL AI CLUSTER 00:00 he points at the MikroTik CRS 804, "you need some kind of switch that'll handle QSFP56 ports like these", the interconnect that makes the whole cluster possible the GB10 ecosystem is no longer just Nvidia. Dell Pro Max GB10, ASUS Ascent GX10, and MSI Edge Expert all ship the same Grace Blackwell Superchip with 128GB of coherent memory. same silicon, different cases, same 200 gigabit ports on the back the CRS 804 is what connects them at prosumer prices. four 400 gigabit QSFP56 ports on one 1U chassis, breakout cables that split each port into two 200 gigabit lanes. one switch drives eight GB10 units in parallel do the math. eight nodes at 128GB each equals 1024GB of pooled unified memory across the cluster. run vLLM, shard a frontier model across all eight, and inference happens locally on hardware that fits in half a rack the real limiter revealed in the stress test was never throttling. it was interconnect topology, exactly the layer this switch fixes at a fraction of enterprise switch pricing $400 a month for combined chatgpt pro and claude code max hits $4,800 a year per developer. a small team of five running through this cluster pays back inside eight months and never expires the article covers the buying ladder for a single desk. this post is proof of the cluster ladder that starts where the desk one ends save this before the GB10 lineup grows past four vendors and prosumer cluster switches move upmarket

NO1ennn

59,725 views • 3 months ago