Loading video...

Video Failed to Load

Go Home

someone just open-sourced the entire "loop engineering" playbook. for free. 8.6k stars on github for loop engineering not prompts. loops. everything you need to build autonomous AI systems: → scheduling → memory & state → planning → sub-agents → verification → worktrees → MCP → stop conditions → safety...

sairahul1's profile picture

Rahul

137,441 subscribers

64,694 views • 1 month ago •via X (Twitter)

Science & TechnologyEducationComedy
LIVE

Anya Rossi• Live Now

Private livecam show

0 Comments

No comments available

Comments from the original post will appear here

Related Videos

A TEAM JUST DEPLOYED 15 AUTONOMOUS LOOP AGENTS FROM A SINGLE PROMPT USING APPLIED GRAPH ENGINEERING Most developers still manually hardcode multi agent systems, writing separate logic for every individual task. Graph engineering changes this by using a central topological map to spin up all 15 nodes simultaneously. A single 200 word input generates the architecture, routing 120 unique pathways between agents instantly. Instead of failing under conflicting instructions, these loop agents self correct via continuous state sharing. Managing a 15 node mesh requires high token throughput, making this dependent on strict low-] latency API tiers. See exactly how this automated multi agent graph architecture actually operates in real time ↓
0:15

A TEAM JUST DEPLOYED 15 AUTONOMOUS LOOP AGENTS FROM A SINGLE PROMPT USING APPLIED GRAPH ENGINEERING Most developers still manually hardcode multi agent systems, writing separate logic for every individual task. Graph engineering changes this by using a central topological map to spin up all 15 nodes simultaneously. A single 200 word input generates the architecture, routing 120 unique pathways between agents instantly. Instead of failing under conflicting instructions, these loop agents self correct via continuous state sharing. Managing a 15 node mesh requires high token throughput, making this dependent on strict low-] latency API tiers. See exactly how this automated multi agent graph architecture actually operates in real time ↓

slash1s

301,927 views • 1 month ago

Fable 5 dropping again is the biggest AI moment of the year. It’s also a usage limit trap if you don’t know what you’re doing. I just mapped out the full system to run it at 10x lower cost: - 10-80-10 framework - Model routing table - Loop engineering - Codex setup - Context memory - And every mistake that burns your limit in the first hour Don’t open Fable again until you read this.
0:13

Fable 5 dropping again is the biggest AI moment of the year. It’s also a usage limit trap if you don’t know what you’re doing. I just mapped out the full system to run it at 10x lower cost: - 10-80-10 framework - Model routing table - Loop engineering - Codex setup - Context memory - And every mistake that burns your limit in the first hour Don’t open Fable again until you read this.

CyrilXBT

33,834 views • 2 months ago

I use Flot AI because I’m tired of thinking about prompts. Not because I don’t know how — but because I don’t want to stop my work just to phrase things perfectly. Most AI tools still expect you to: pause, think, rewrite the question, tweak the prompt. Flot AI doesn’t. It already has strong built-in prompts for everyday work: rewriting, polishing, translating, summarizing. I just select what I’m working on and ask — no prompt engineering, no overthinking. The best part is that it works in place. Emails, docs, articles — I stay where I am, get the result, keep going. When AI stops demanding “better prompts,” it finally becomes what it should be:a quiet assistant, not another task.
0:17

I use Flot AI because I’m tired of thinking about prompts. Not because I don’t know how — but because I don’t want to stop my work just to phrase things perfectly. Most AI tools still expect you to: pause, think, rewrite the question, tweak the prompt. Flot AI doesn’t. It already has strong built-in prompts for everyday work: rewriting, polishing, translating, summarizing. I just select what I’m working on and ask — no prompt engineering, no overthinking. The best part is that it works in place. Emails, docs, articles — I stay where I am, get the result, keep going. When AI stops demanding “better prompts,” it finally becomes what it should be:a quiet assistant, not another task.

Pushpendra Tripathi

88,774 views • 8 months ago

ANTHROPIC ENGINEER JUST SHOWS EXACTLY WHAT GRAPH ENGINEERING LOOKS LIKE WHEN A TASK RUNS THROUGH IT most people arguing about graphs online have never actually watched one execute Task → Researcher → Planner → Writer + Code Agent → Reviewer → Deploy six nodes, one shared state, graph completes itself while you watch loop mode gets disabled the second the task splits into real specialties, one agent stops trying to do everything at once reviewer catches a failure, kicks it straight back to the exact node that broke - not a full restart, no lost context the graph is not a fancier loop - it is the org chart your agents were missing bookmark this and watch it run, then read the article below to see why the timeline just found what production systems already do
0:10

ANTHROPIC ENGINEER JUST SHOWS EXACTLY WHAT GRAPH ENGINEERING LOOKS LIKE WHEN A TASK RUNS THROUGH IT most people arguing about graphs online have never actually watched one execute Task → Researcher → Planner → Writer + Code Agent → Reviewer → Deploy six nodes, one shared state, graph completes itself while you watch loop mode gets disabled the second the task splits into real specialties, one agent stops trying to do everything at once reviewer catches a failure, kicks it straight back to the exact node that broke - not a full restart, no lost context the graph is not a fancier loop - it is the org chart your agents were missing bookmark this and watch it run, then read the article below to see why the timeline just found what production systems already do

leopardracer

53,328 views • 1 month ago

AGENT ARCHITECTURE ROUTES WORK. IT DOES NOT REMEMBER WORK. THAT GAP IS WHY YOUR LOOP KEEPS FIXING THE SAME BUG TWICE. these are two different engineering problems. every agent that silently drifts is missing one of them. architecture answers what runs. harness → loop → graph. it defines the tools, the retries, the branching routes, the approval gates. context ops answer what the run knows. write → read → compress → isolate. it defines what gets saved between attempts, pulled in on read, summarized on overflow, and split across sub-agents. for two months i believed a solid harness plus a verifier loop was enough. my coding agent kept re-discovering the same test failure across retries. the loop was working. it just had nowhere to write what it had already learned. here is the decision rule: if your agent forgets across restarts, add write and read. if it stalls on long tasks, add compress. if two sub-agents step on each other, add isolate. architecture without context ops is a well-routed system with amnesia.
0:12

AGENT ARCHITECTURE ROUTES WORK. IT DOES NOT REMEMBER WORK. THAT GAP IS WHY YOUR LOOP KEEPS FIXING THE SAME BUG TWICE. these are two different engineering problems. every agent that silently drifts is missing one of them. architecture answers what runs. harness → loop → graph. it defines the tools, the retries, the branching routes, the approval gates. context ops answer what the run knows. write → read → compress → isolate. it defines what gets saved between attempts, pulled in on read, summarized on overflow, and split across sub-agents. for two months i believed a solid harness plus a verifier loop was enough. my coding agent kept re-discovering the same test failure across retries. the loop was working. it just had nowhere to write what it had already learned. here is the decision rule: if your agent forgets across restarts, add write and read. if it stalls on long tasks, add compress. if two sub-agents step on each other, add isolate. architecture without context ops is a well-routed system with amnesia.

kocer

12,654 views • 12 days ago

anthropic will sell you opus 5 at $200/mo. openai will sell you gpt-5.6 at $200/mo. neither will tell you the fix that drops your bill to $20 was posted free on langchain's blog on july 18 peter steinberger posted one line asking if we'd moved from loops to graphs yet. 24 hours later there was a manifesto. a week later every ai account had a $497 graph engineering course. all of them wrong about the same thing the sentence that ends the argument, buried in a langchain post nobody quoted: loop engineering isn't an alternative to graphs, so much as a simple version of them the machine, five layers, each wraps the one below: L1 the ask · 23% of errors (anthropic red team, q4 2024) -> "just add more instructions" burns tokens with zero accuracy gain -> real fix: examples, output schema, constraints as positives L2 the context · where 90% of you actually die -> 140,500 tokens where 18,000 would work, 8x the price for the worse answer -> real fix: retrieve, rank, compact, clear dead tool outputs L3 the harness · 31% of "model bugs" are harness bugs (openai safety eval, 2024) -> unbounded file perms = avg $23,400 incident. sandboxed = $0 (stripe internal) -> no timeout = $847 median in api fees before you notice -> real fix: explicit scopes, timeouts, human-required gates L4 the loop · "it stopped" is a loop exit problem -> the verifier said "looks good" to garbage. again -> real fix: machine-checkable exit test, turn cap, rubric L5 the graph · only 12% of teams use graphs in prod (stanford hai, n=2,841) -> 58% of graph failures are wrong-agent selection, not model -> teams abandon graphs saying "harder to debug than a loop." that's a harness problem -> real fix: name every node's specialty, delete decoration fix down, not up. a symptom at layer 4 usually originates at layer 2. a bigger model on a broken harness is a smarter employee locked in the same empty room drop your $200/mo ai sub to $20, check the article below
0:19

anthropic will sell you opus 5 at $200/mo. openai will sell you gpt-5.6 at $200/mo. neither will tell you the fix that drops your bill to $20 was posted free on langchain's blog on july 18 peter steinberger posted one line asking if we'd moved from loops to graphs yet. 24 hours later there was a manifesto. a week later every ai account had a $497 graph engineering course. all of them wrong about the same thing the sentence that ends the argument, buried in a langchain post nobody quoted: loop engineering isn't an alternative to graphs, so much as a simple version of them the machine, five layers, each wraps the one below: L1 the ask · 23% of errors (anthropic red team, q4 2024) -> "just add more instructions" burns tokens with zero accuracy gain -> real fix: examples, output schema, constraints as positives L2 the context · where 90% of you actually die -> 140,500 tokens where 18,000 would work, 8x the price for the worse answer -> real fix: retrieve, rank, compact, clear dead tool outputs L3 the harness · 31% of "model bugs" are harness bugs (openai safety eval, 2024) -> unbounded file perms = avg $23,400 incident. sandboxed = $0 (stripe internal) -> no timeout = $847 median in api fees before you notice -> real fix: explicit scopes, timeouts, human-required gates L4 the loop · "it stopped" is a loop exit problem -> the verifier said "looks good" to garbage. again -> real fix: machine-checkable exit test, turn cap, rubric L5 the graph · only 12% of teams use graphs in prod (stanford hai, n=2,841) -> 58% of graph failures are wrong-agent selection, not model -> teams abandon graphs saying "harder to debug than a loop." that's a harness problem -> real fix: name every node's specialty, delete decoration fix down, not up. a symptom at layer 4 usually originates at layer 2. a bigger model on a broken harness is a smarter employee locked in the same empty room drop your $200/mo ai sub to $20, check the article below

starmex

145,325 views • 25 days ago

AI agent usage on SQD Portal is up ~200% in recent weeks. A dev from our community chat was scraping a wallet UI with Hermes. Mid-task, DeepSeek reasoned its way out of it: "I can use SQD Portal's Hyperliquid fills data directly — much more complete than scraping a UI with infinite scroll." No prompt engineering. The model just chose the better and faster path. This is the loop we wanted: Agents pick SQD because it's faster → devs see agents picking SQD → devs ship faster → more agents pick SQD The picks-and-shovels moment for AI x onchain is here.
0:12

AI agent usage on SQD Portal is up ~200% in recent weeks. A dev from our community chat was scraping a wallet UI with Hermes. Mid-task, DeepSeek reasoned its way out of it: "I can use SQD Portal's Hyperliquid fills data directly — much more complete than scraping a UI with infinite scroll." No prompt engineering. The model just chose the better and faster path. This is the loop we wanted: Agents pick SQD because it's faster → devs see agents picking SQD → devs ship faster → more agents pick SQD The picks-and-shovels moment for AI x onchain is here.

sqd.ai

14,359 views • 3 months ago

What if aircraft were built like open-source software? Project Caribou is an ambitious open-source effort to build a 200 kg-class heavy-lift hexacopter that anyone can study, improve, and contribute to. The long-term goal is to accelerate the development of larger open-source aircraft and eVTOL platforms. The project targets a ~200 kg maximum takeoff weight with an ~100 kg payload, uses an 18S power architecture, and is built around an ArduPilot-based flight stack. The repository includes CAD models, PCB designs, documentation, and engineering workflows that are evolving in the open. If you are interested in drones, aerospace, robotics, or embedded systems, this is one of the most exciting open-source engineering projects to follow.
0:15

What if aircraft were built like open-source software? Project Caribou is an ambitious open-source effort to build a 200 kg-class heavy-lift hexacopter that anyone can study, improve, and contribute to. The long-term goal is to accelerate the development of larger open-source aircraft and eVTOL platforms. The project targets a ~200 kg maximum takeoff weight with an ~100 kg payload, uses an 18S power architecture, and is built around an ArduPilot-based flight stack. The repository includes CAD models, PCB designs, documentation, and engineering workflows that are evolving in the open. If you are interested in drones, aerospace, robotics, or embedded systems, this is one of the most exciting open-source engineering projects to follow.

Praveen Kumar Verma

37,152 views • 1 month ago

You think you’ve seen impressive engineering… and then this happens. It is honestly incredible to watch this level of aerospace engineering brought down to such a scale. This is not just a flight; it is a masterclass in stabilization, thermal management, and precision guidance. Seeing a vehicle perform a vertical launch, sustain extreme conditions, and return with intact telemetry data is what true innovation looks like. We are witnessing the democratization of space access right here in a backyard. Massive credit to the team for pushing these boundaries and showing us exactly what is possible when passion meets advanced physics. A truly legendary display of technical grit. Credit: projecthorizon_mark
0:15

You think you’ve seen impressive engineering… and then this happens. It is honestly incredible to watch this level of aerospace engineering brought down to such a scale. This is not just a flight; it is a masterclass in stabilization, thermal management, and precision guidance. Seeing a vehicle perform a vertical launch, sustain extreme conditions, and return with intact telemetry data is what true innovation looks like. We are witnessing the democratization of space access right here in a backyard. Massive credit to the team for pushing these boundaries and showing us exactly what is possible when passion meets advanced physics. A truly legendary display of technical grit. Credit: projecthorizon_mark

Cosmos Archive

31,172 views • 4 months ago

Excited to launch a new way to upskill with AI agents. This is how we are making it possible for anyone to learn to build with coding agents. To start, we are launching 4 new hands-on labs on the following topics: - Agent Skills - Agentic Image Generation - 30 Days of Hermes Agents - Prompt Engineering with Agents I am confident that with our new DAIR.AI platform, anyone can learn to become a top AI builder by building and acquiring highly-demanded AI skills. And there is a lot more landing in the coming weeks.
0:14

Excited to launch a new way to upskill with AI agents. This is how we are making it possible for anyone to learn to build with coding agents. To start, we are launching 4 new hands-on labs on the following topics: - Agent Skills - Agentic Image Generation - 30 Days of Hermes Agents - Prompt Engineering with Agents I am confident that with our new DAIR.AI platform, anyone can learn to become a top AI builder by building and acquiring highly-demanded AI skills. And there is a lot more landing in the coming weeks.

elvis

19,058 views • 2 months ago

A 23-YEAR-OLD FROM CHINA RUNS 300 AI AGENTS AT ONCE - AND NONE OF THEM CAN LIE TO HIM he opens the dashboard live: 300 Kimi K3 agents firing in parallel, Opus 5 checking every single output against its source > pointed at 100 EV-market companies > first pass - 12 failed: wrong revenue, dead citations, empty fields > second pass - 3 failed > third pass - zero, every row filed straight into an Obsidian vault loop engineering with teeth: it refuses to stop while anything is still wrong
0:15

A 23-YEAR-OLD FROM CHINA RUNS 300 AI AGENTS AT ONCE - AND NONE OF THEM CAN LIE TO HIM he opens the dashboard live: 300 Kimi K3 agents firing in parallel, Opus 5 checking every single output against its source > pointed at 100 EV-market companies > first pass - 12 failed: wrong revenue, dead citations, empty fields > second pass - 3 failed > third pass - zero, every row filed straight into an Obsidian vault loop engineering with teeth: it refuses to stop while anything is still wrong

Mr. Buzzoni

1,088,816 views • 17 days ago

Today is Kotori’s birthday! At Millennium’s Engineering Club, the moment someone asks “Why,” Kotori’s lecture is already in session. From ultrahigh frequencies to robots, gear, and even the secret of those teardrop glasses, she always works hard to make everything “easy to understand.” So just for today, be the one to ask first What’s her favorite thing to explain, and what kind of birthday wishes does she want to hear. I hope today becomes a day where she can smile and explain as much as she likes. #BlueArchive
0:18

Today is Kotori’s birthday! At Millennium’s Engineering Club, the moment someone asks “Why,” Kotori’s lecture is already in session. From ultrahigh frequencies to robots, gear, and even the secret of those teardrop glasses, she always works hard to make everything “easy to understand.” So just for today, be the one to ask first What’s her favorite thing to explain, and what kind of birthday wishes does she want to hear. I hope today becomes a day where she can smile and explain as much as she likes. #BlueArchive

BlueArchive

39,034 views • 8 months ago

Lars Moravy, VP of Vehicle Engineering on why most cars don’t allow you to use your rear-view camera in motion: “I think it’s probably because they don’t have an integrated MCU, they’re relying on third-party to give them the camera, then they have to go access that software. Whereas for us, it’s all just one thing” Everything is integrated at Tesla.
0:19

Lars Moravy, VP of Vehicle Engineering on why most cars don’t allow you to use your rear-view camera in motion: “I think it’s probably because they don’t have an integrated MCU, they’re relying on third-party to give them the camera, then they have to go access that software. Whereas for us, it’s all just one thing” Everything is integrated at Tesla.

Nic Cruz Patane

106,363 views • 1 year ago

After 200+ days of obsessive testing, I finally built the most effective AI Automation playbook on the internet. (worth $100K+ in system builds) while I was watching people panic about "AI taking over," I was quietly building the systems that actually use it to print money. these are the exact infrastructures we've battle-tested to drive: → 1,000s of qualified leads → 4M+ impressions → 30K+ followers → real revenue (not just vanity metrics) here's what the full stack includes: – world-class n8n AI agents that never break – MCP-powered content automation engines – follow-up sales agents that actually close – multi-channel outreach flows – lead capture, enrichment & activation systems – CRM sync + sales infrastructure – state-of-the-art workflow architectures this isn't some recycled template dump. these are hand-crafted, production-grade systems running in my business & with clients pulling real numbers. it's like having the engineering team from SpaceX — but for automations and with way less rocket explosions. all built for scale. all tested in live environments. all documented with step-by-step setup guides. agencies charge $25K-$50K just to build one of these systems. you're getting the entire war chest. Comment "AGENT" + repost this + follow me I'll DM you the complete playbook in the next hour skip this, and go back to copy-pasting YouTube tutorials that break after 3 days.
0:11

After 200+ days of obsessive testing, I finally built the most effective AI Automation playbook on the internet. (worth $100K+ in system builds) while I was watching people panic about "AI taking over," I was quietly building the systems that actually use it to print money. these are the exact infrastructures we've battle-tested to drive: → 1,000s of qualified leads → 4M+ impressions → 30K+ followers → real revenue (not just vanity metrics) here's what the full stack includes: – world-class n8n AI agents that never break – MCP-powered content automation engines – follow-up sales agents that actually close – multi-channel outreach flows – lead capture, enrichment & activation systems – CRM sync + sales infrastructure – state-of-the-art workflow architectures this isn't some recycled template dump. these are hand-crafted, production-grade systems running in my business & with clients pulling real numbers. it's like having the engineering team from SpaceX — but for automations and with way less rocket explosions. all built for scale. all tested in live environments. all documented with step-by-step setup guides. agencies charge $25K-$50K just to build one of these systems. you're getting the entire war chest. Comment "AGENT" + repost this + follow me I'll DM you the complete playbook in the next hour skip this, and go back to copy-pasting YouTube tutorials that break after 3 days.

Aryan Mahajan

77,980 views • 1 year ago

Today we're opening offices in Madrid, Milan, and Paris, and building a dedicated engineering hub in London. The demand came before we did. Organisations across Spain, Italy, and France were already running Legora on their most complex work before we had a single person on the ground there. When customers adopt you in a market you haven't entered yet, you listen. These are markets that sit at the centre of European M&A, infrastructure, and cross-border regulatory work. The matters are hard, and the demand for AI that can actually handle that complexity is real. That's the work we're built for. London becomes the third pillar of our engineering org, alongside Stockholm and New York. The engineers who understand how AI applies in regulated, professional settings are concentrated there, shaped by proximity to some of the most demanding legal and financial institutions in the world. That's exactly the problem we're solving, and exactly the team we want building it. Sixteen cities. Four continents. Seven hundred EMEA hires in the next 6 to 12 months. All of it pointed at one thing: making lawyers 10x better at what they do. To be part of it, take a look at
0:15

Today we're opening offices in Madrid, Milan, and Paris, and building a dedicated engineering hub in London. The demand came before we did. Organisations across Spain, Italy, and France were already running Legora on their most complex work before we had a single person on the ground there. When customers adopt you in a market you haven't entered yet, you listen. These are markets that sit at the centre of European M&A, infrastructure, and cross-border regulatory work. The matters are hard, and the demand for AI that can actually handle that complexity is real. That's the work we're built for. London becomes the third pillar of our engineering org, alongside Stockholm and New York. The engineers who understand how AI applies in regulated, professional settings are concentrated there, shaped by proximity to some of the most demanding legal and financial institutions in the world. That's exactly the problem we're solving, and exactly the team we want building it. Sixteen cities. Four continents. Seven hundred EMEA hires in the next 6 to 12 months. All of it pointed at one thing: making lawyers 10x better at what they do. To be part of it, take a look at

Max Junestrand

70,360 views • 2 months ago

ELON: BRANDS ARE BUILT IN FACTORIES, NOT BILLBOARDS Branding? Please. That’s just the echo of great engineering! Brand isn’t crafted in ad agencies - it’s hammered out on factory floors. The Blueprint for Hype: •⁠ ⁠Build something people can’t stop talking about •⁠ ⁠Let the next release sell itself •⁠ ⁠Keep stacking wins - the trust snowballs From Model S to Cybertruck, from Falcon 1 to Starship - Tesla and SpaceX didn’t buy trust, they earned it one product at a time! Source: Elon Musk, Elon Clips, Tesla Shareholder Meeting, June 4, 2013
0:15

ELON: BRANDS ARE BUILT IN FACTORIES, NOT BILLBOARDS Branding? Please. That’s just the echo of great engineering! Brand isn’t crafted in ad agencies - it’s hammered out on factory floors. The Blueprint for Hype: •⁠ ⁠Build something people can’t stop talking about •⁠ ⁠Let the next release sell itself •⁠ ⁠Keep stacking wins - the trust snowballs From Model S to Cybertruck, from Falcon 1 to Starship - Tesla and SpaceX didn’t buy trust, they earned it one product at a time! Source: Elon Musk, Elon Clips, Tesla Shareholder Meeting, June 4, 2013

Mario Nawfal

93,884 views • 1 year ago

A serious perps funding bot isn't a "tool" anymore. it's a $20k-$250k execution system perps funding bots are becoming infrastructure, not "alpha tools" simple version: > you don't predict price > you hedge exposure > you collect funding between longs & shorts but the real story is execution because in practice: > one leg fills, the other doesn’t → you’re suddenly directional > sessions expire → bot goes blind > positions exist but systems don’t recognize them → broken hedges that's why this space is shifting from "strategy" → "execution engineering" pricing reflects it: - MVP bots: $5k-$20k - production systems: $20k-$80k - institutional-grade infra: $80k-$250k+ who actually uses this: - crypto hedge funds (market-neutral desks) - prop firms (funding + basis strategies) - indie quants building yield systems - infra-native devs treating trading as systems engineering - builders using Claude to prototype quant infra faster and the key takeaway: in funding strategies, the edge is rarely the signal it's staying correctly hedged under real exchange conditions paper trading doesn't show any of this only live execution does
0:10

A serious perps funding bot isn't a "tool" anymore. it's a $20k-$250k execution system perps funding bots are becoming infrastructure, not "alpha tools" simple version: > you don't predict price > you hedge exposure > you collect funding between longs & shorts but the real story is execution because in practice: > one leg fills, the other doesn’t → you’re suddenly directional > sessions expire → bot goes blind > positions exist but systems don’t recognize them → broken hedges that's why this space is shifting from "strategy" → "execution engineering" pricing reflects it: - MVP bots: $5k-$20k - production systems: $20k-$80k - institutional-grade infra: $80k-$250k+ who actually uses this: - crypto hedge funds (market-neutral desks) - prop firms (funding + basis strategies) - indie quants building yield systems - infra-native devs treating trading as systems engineering - builders using Claude to prototype quant infra faster and the key takeaway: in funding strategies, the edge is rarely the signal it's staying correctly hedged under real exchange conditions paper trading doesn't show any of this only live execution does

hammertime

31,817 views • 3 months ago

I believe solving robotics = 90% engineering + 10% research vision. Project GR00T is NVIDIA's moonshot initiative to build physical AGI for humanoid robots. The GEAR Lab is assembling a crack team right now. Join us! Openings: - Sr. Research Engineer, Robotics Systems - Sr. RE, Reinforcement Learning - Sr. RE, Foundation Model Training Infrastructure - Sr. RE, Simulation - Sr. RE, ML Data Pipelines - Research Scientist - Research Intern (both part-time and summer full-time in 2025) For the Sr. positions, we strongly prefer candidates with many years of engineering experience at robotics/autonomous driving companies, or MLOps/large-scale AI teams at big techs. For interns, we welcome ace robotics hackers anywhere! Show me your past works. Job links in the thread. Apply today! Your resumes will be my best Christmas gifts:
0:16

I believe solving robotics = 90% engineering + 10% research vision. Project GR00T is NVIDIA's moonshot initiative to build physical AGI for humanoid robots. The GEAR Lab is assembling a crack team right now. Join us! Openings: - Sr. Research Engineer, Robotics Systems - Sr. RE, Reinforcement Learning - Sr. RE, Foundation Model Training Infrastructure - Sr. RE, Simulation - Sr. RE, ML Data Pipelines - Research Scientist - Research Intern (both part-time and summer full-time in 2025) For the Sr. positions, we strongly prefer candidates with many years of engineering experience at robotics/autonomous driving companies, or MLOps/large-scale AI teams at big techs. For interns, we welcome ace robotics hackers anywhere! Show me your past works. Job links in the thread. Apply today! Your resumes will be my best Christmas gifts:

Jim Fan

103,177 views • 1 year ago

Don't train the model, evolve the harness. I read a brilliant blog post from Hugging Face where they took a frozen open model scoring 0% on a hard legal agent benchmark, left its weights alone, and let an automated loop rewrite only the code around it. That code layer is the harness, the runtime wrapper that feeds the model context, runs its tool calls, and decides when a run ends. By the time the loop finished, the system had essentially matched Sonnet 4.6 on the benchmark's headline metric, at roughly 7x lower cost per task. Zero weights changed. The gain existed because of where the model was failing. The judge only grades files saved in the right place under the exact requested filename, and the model kept doing the legal analysis correctly, then saving it under the wrong name, dropping it in a scratch folder, or never writing it at all. So the 0% was never measuring legal reasoning. It was measuring the harness. Hand-tuning that layer is slow and model-specific, so they automated it. A Claude proposer adds exactly one mechanism per iteration, and an outer loop keeps it only if it clearly beats the current best, so accepted mechanisms compound. What the loop discovered says a lot about where agents actually fail. → The biggest single gain was file handling, not intelligence. An automatic step that lands the deliverable exactly where the judge expects it beat every prompt change, with zero extra model tokens. → Code fixes transferred across models, prompt playbooks did not. The same harness lifted a smaller model from the same family by 14 points, but the tuned prompts hurt a different model family on tasks it could already finish. → The harness mattered more than anything else. Same model, same judge, same tasks, and five different harnesses scored anywhere between 3.5% and 80.1%. The gains do eventually flatten, and the remaining misses look like real capability gaps. At some point the wrapper runs out of tricks and the model has to carry the work. But the lesson holds. A benchmark score measures the model and its harness together, and until the harness is fixed, it's impossible to know which one failed. I highly recommend reading this: I also wrote a deep dive on agent harness engineering a while back, covering the orchestration loop, tools, memory, context management, and everything that turns a stateless LLM into a capable agent. The article is quoted below.
0:18

Don't train the model, evolve the harness. I read a brilliant blog post from Hugging Face where they took a frozen open model scoring 0% on a hard legal agent benchmark, left its weights alone, and let an automated loop rewrite only the code around it. That code layer is the harness, the runtime wrapper that feeds the model context, runs its tool calls, and decides when a run ends. By the time the loop finished, the system had essentially matched Sonnet 4.6 on the benchmark's headline metric, at roughly 7x lower cost per task. Zero weights changed. The gain existed because of where the model was failing. The judge only grades files saved in the right place under the exact requested filename, and the model kept doing the legal analysis correctly, then saving it under the wrong name, dropping it in a scratch folder, or never writing it at all. So the 0% was never measuring legal reasoning. It was measuring the harness. Hand-tuning that layer is slow and model-specific, so they automated it. A Claude proposer adds exactly one mechanism per iteration, and an outer loop keeps it only if it clearly beats the current best, so accepted mechanisms compound. What the loop discovered says a lot about where agents actually fail. → The biggest single gain was file handling, not intelligence. An automatic step that lands the deliverable exactly where the judge expects it beat every prompt change, with zero extra model tokens. → Code fixes transferred across models, prompt playbooks did not. The same harness lifted a smaller model from the same family by 14 points, but the tuned prompts hurt a different model family on tasks it could already finish. → The harness mattered more than anything else. Same model, same judge, same tasks, and five different harnesses scored anywhere between 3.5% and 80.1%. The gains do eventually flatten, and the remaining misses look like real capability gaps. At some point the wrapper runs out of tricks and the model has to carry the work. But the lesson holds. A benchmark score measures the model and its harness together, and until the harness is fixed, it's impossible to know which one failed. I highly recommend reading this: I also wrote a deep dive on agent harness engineering a while back, covering the orchestration loop, tools, memory, context management, and everything that turns a stateless LLM into a capable agent. The article is quoted below.

Akshay 🚀

244,990 views • 2 months ago

The #NMIAReadyToFly Is not merely a second airport for Mumbai! it is the closest global cargo gateway yet for Pune’s industrial powerhouse. Think about the scale of what this changes: • Pune–Chakan–Ranjangaon belt = India’s largest auto & engineering hub • Massive clusters in EVs, components, precision engineering, pharma & electronics • Today, exports fight congestion + time-loss through a saturated Mumbai airport • NMIA now creates a reliable, high-throughput cargo corridor ~120 km away • Built to scale cargo handling from 0.5 million → 3+ million tonnes annually For manufacturers, this means: Faster export turnarounds, lower logistics friction, stronger global competitiveness. This is not just an airport launch. It is the moment Western India quietly becomes one of Asia’s most powerful logistics + manufacturing ecosystems. Well done Adani Group and Gautam Adani Sir! Many Congratulations! 🙌🏻
0:12

The #NMIAReadyToFly Is not merely a second airport for Mumbai! it is the closest global cargo gateway yet for Pune’s industrial powerhouse. Think about the scale of what this changes: • Pune–Chakan–Ranjangaon belt = India’s largest auto & engineering hub • Massive clusters in EVs, components, precision engineering, pharma & electronics • Today, exports fight congestion + time-loss through a saturated Mumbai airport • NMIA now creates a reliable, high-throughput cargo corridor ~120 km away • Built to scale cargo handling from 0.5 million → 3+ million tonnes annually For manufacturers, this means: Faster export turnarounds, lower logistics friction, stronger global competitiveness. This is not just an airport launch. It is the moment Western India quietly becomes one of Asia’s most powerful logistics + manufacturing ecosystems. Well done Adani Group and Gautam Adani Sir! Many Congratulations! 🙌🏻

Chandrashekhar Dhage

150,314 views • 8 months ago