Загрузка видео...

Не удалось загрузить видео

Возникла проблема при загрузке этого видео. Это может быть связано с временными проблемами сети или видео может быть недоступно.

На главную

Let me explain the agent loop, simple It's the core of every agentic system, and the part most people overcomplicate It's just this: 1. Send messages to the model 2. Model responds, maybe calls a tool 3. You run the tool 4. Append the result back to messages 5.... show more

Daniel San

32,970 subscribers

12,514 просмотров • 2 месяцев назад •via X (Twitter)

Anya Rossi• Live Now

Private livecam show

Комментарии: 0

Нет доступных комментариев

Здесь появятся комментарии из оригинального поста

Похожие видео

an agent is four parts in a loop. you own one. the other three break it. that's why the demo works and prod doesn't. you can't debug what you can't see. 1) the prompt → what you tell the model each turn. you own this one. good. 2) the context window → what it sees right now. the framework fills it with junk, and you never notice until it rots. 3) the tools → what it can do. you own the list, not when or why it fires them. 4) the control flow → what happens next, when to stop. the framework owns this. it's what breaks at 80%. own all four and your agent stops being a magic trick that works on stage and dies on call. this isn't my idea. it's the 12-factor agents guide (24k stars) github: the whole thing every serious builder ends up rewriting their stack around. full breakdown in the article below.

an agent is four parts in a loop. you own one. the other three break it. that's why the demo works and prod doesn't. you can't debug what you can't see. 1) the prompt → what you tell the model each turn. you own this one. good. 2) the context window → what it sees right now. the framework fills it with junk, and you never notice until it rots. 3) the tools → what it can do. you own the list, not when or why it fires them. 4) the control flow → what happens next, when to stop. the framework owns this. it's what breaks at 80%. own all four and your agent stops being a magic trick that works on stage and dies on call. this isn't my idea. it's the 12-factor agents guide (24k stars) github: the whole thing every serious builder ends up rewriting their stack around. full breakdown in the article below.

Hanako

38,184 просмотров • 1 месяц назад

New open-source agent harness just landed! I got early access to TrueForge by TrueFoundry and have been running it locally for the past few days. The harness layer deserves as much attention as the model, and open source matters here because you can inspect the loop, run it on your own infrastructure, and swap to the latest or cheaper models. TrueForge handles the runtime work that makes an agent reliable. It drives the tool-calling loop, manages context, coordinates subagents, and executes code in a sandbox, with any model you choose. Every tool call re-sends the growing context to the model, so in practice the harness controls most of what an agent costs to run. A few things stood out from my testing and their published benchmarks. Vendor-Neutral by design. It runs OpenAI, Anthropic, and Google models alongside open-weight models like Kimi, GLM, and DeepSeek. Model routing is a setting, and you can send each task to the model that fits it. On a 14-task enterprise agent benchmark, it matched the accuracy of Claude Managed Agents running the same Opus 4.8 model at roughly 30% lower cost per run (3.8M tokens vs 10M for the same answers). Routing the same tasks to GLM-5.2 held accuracy and brought cost down by about 75%, around $3 per run instead of $12. Fully self-hosted and Open Source (MIT License). I had it running locally with one command, with sandboxed code execution working out of the box. It's time to own your agent harness. Thanks to TrueFoundry for partnering on this post.

New open-source agent harness just landed! I got early access to TrueForge by TrueFoundry and have been running it locally for the past few days. The harness layer deserves as much attention as the model, and open source matters here because you can inspect the loop, run it on your own infrastructure, and swap to the latest or cheaper models. TrueForge handles the runtime work that makes an agent reliable. It drives the tool-calling loop, manages context, coordinates subagents, and executes code in a sandbox, with any model you choose. Every tool call re-sends the growing context to the model, so in practice the harness controls most of what an agent costs to run. A few things stood out from my testing and their published benchmarks. Vendor-Neutral by design. It runs OpenAI, Anthropic, and Google models alongside open-weight models like Kimi, GLM, and DeepSeek. Model routing is a setting, and you can send each task to the model that fits it. On a 14-task enterprise agent benchmark, it matched the accuracy of Claude Managed Agents running the same Opus 4.8 model at roughly 30% lower cost per run (3.8M tokens vs 10M for the same answers). Routing the same tasks to GLM-5.2 held accuracy and brought cost down by about 75%, around $3 per run instead of $12. Fully self-hosted and Open Source (MIT License). I had it running locally with one command, with sandboxed code execution working out of the box. It's time to own your agent harness. Thanks to TrueFoundry for partnering on this post.

elvis

11,303 просмотров • 7 дней назад

let me save you 3 hours of head scratching. if you're running local models like Qwen3.5-35B-A3B through Claude Code via llama.cpp's Anthropic endpoint, the chain will break every 3 to 5 minutes. tool call fails. flow stops. you reprompt. it recovers. 2 minutes later it stops again. the model is fine. the harness chokes on local inference latency. switch to OpenCode. same localhost endpoint. same model. same GPU. the chain doesn't break. the tradeoff: OpenCode sometimes loops. the model forgets what it already read and repeats the same tool call. but a loop you can interrupt. a broken chain kills your momentum and you start over. watch both side by side. proprietary agent vs open source agent. same 3B model. different failure modes. pick your poison.

let me save you 3 hours of head scratching. if you're running local models like Qwen3.5-35B-A3B through Claude Code via llama.cpp's Anthropic endpoint, the chain will break every 3 to 5 minutes. tool call fails. flow stops. you reprompt. it recovers. 2 minutes later it stops again. the model is fine. the harness chokes on local inference latency. switch to OpenCode. same localhost endpoint. same model. same GPU. the chain doesn't break. the tradeoff: OpenCode sometimes loops. the model forgets what it already read and repeats the same tool call. but a loop you can interrupt. a broken chain kills your momentum and you start over. watch both side by side. proprietary agent vs open source agent. same 3B model. different failure modes. pick your poison.

Sudo su

72,538 просмотров • 5 месяцев назад

ANTHROPIC JUST DROPPED THE OFFICIAL GUIDE TO PROMPTING FABLE 5. This is the most important prompting framework I've seen. Bookmark this before you forget. Most people treat Fable 5 like a chatbot. That's the mistake. > don't over-engineer prompts — it degrades output. > use /loop for autonomous multi-step work. > give it the goal, not step-by-step commands. > add a memory file. it learns from past runs. > spin up 50+ subagents for complex tasks. Fable 5 isn't an assistant. it's a consultant that leads the work. Read it before you write another prompt. Claude → Fable 5 → Autonomous Work → Real Output → Money

ANTHROPIC JUST DROPPED THE OFFICIAL GUIDE TO PROMPTING FABLE 5. This is the most important prompting framework I've seen. Bookmark this before you forget. Most people treat Fable 5 like a chatbot. That's the mistake. > don't over-engineer prompts — it degrades output. > use /loop for autonomous multi-step work. > give it the goal, not step-by-step commands. > add a memory file. it learns from past runs. > spin up 50+ subagents for complex tasks. Fable 5 isn't an assistant. it's a consultant that leads the work. Read it before you write another prompt. Claude → Fable 5 → Autonomous Work → Real Output → Money

Kirill

414,789 просмотров • 1 месяц назад

this is the next $100B opportunity in ai , most will miss it's harness engineering what this agentic engineer reveals is insane >The model is almost irrelevant. The harness is everything >every failure is a signal about what the environment needs. >when agent throughput far exceeds human attention, corrections are cheap and waiting is expensive most people will ignore and bookmark. be different.

this is the next $100B opportunity in ai , most will miss it's harness engineering what this agentic engineer reveals is insane >The model is almost irrelevant. The harness is everything >every failure is a signal about what the environment needs. >when agent throughput far exceeds human attention, corrections are cheap and waiting is expensive most people will ignore and bookmark. be different.

Avid

497,726 просмотров • 5 месяцев назад

Do you use Runway Gen-2 to make AI videos? This is for you! I built a simple tool to help "extend" Runway videos beyond the 4 second clip limit. Just choose a video, and the tool will give you the final frame -- so that you can drop this back into Runway, and generate another AI video based on it. Let me know if this is useful to you! Link is at bottom of the video and it's live to try right now.

Do you use Runway Gen-2 to make AI videos? This is for you! I built a simple tool to help "extend" Runway videos beyond the 4 second clip limit. Just choose a video, and the tool will give you the final frame -- so that you can drop this back into Runway, and generate another AI video based on it. Let me know if this is useful to you! Link is at bottom of the video and it's live to try right now.

Benjamin De Kraker

68,384 просмотров • 3 лет назад

this is the important part to understand what is going on. you can see the front wheel lose traction then the back of the car dips due to weight transfer. it was AFTER that the ICE agent draws his weapon.

this is the important part to understand what is going on. you can see the front wheel lose traction then the back of the car dips due to weight transfer. it was AFTER that the ICE agent draws his weapon.

Phil Labonte 🇺🇸

85,981 просмотров • 7 месяцев назад

sorry, they just did WHAT someone gave a machine one disease name, the leading cause of blindness in the developed world with 1.5 million americans already in its path, and it came back pointing at a drug that has sat in pharmacies for years under a different label: 551 papers read in 30 minutes against the 294 hours a human would have needed, and the loop that did it is public on GitHub most agent setups answer one question at a time, so the ceiling on the work is the quality of the question you happened to think of this one was handed a single question and wrote the second one itself. turns out that follow-up is where the real find was: a target called ABCA1, upregulated threefold, in an experiment no human ordered i read the whole paper looking for the trick, and the trick is structural. that is the second question, and it is the gap between an assistant and a factory: - hand the loop a field rather than a task: it was given a disease, and choosing the mechanism was part of its job - make it rank before it spends: 151 papers in, ten candidate mechanisms out, scored against each other before anything touched a bench - split reading from judging, so the agent that forms the theory is a different agent from the one grading it - close every cycle on physical reality: the verdict was an experiment, and another model's opinion was never allowed to stand in for one - feed each result back as the next question rather than a log line, which is the step almost nobody builds - search what already passed inspection first: the winner was an approved compound with a safety file already on record - write down what the round learned before opening the next one, so round two starts where round one stopped my read, and i think it is the uncomfortable one: reading was the entire bottleneck in that field, and everybody spent the decade optimising the writing. people ran every physical experiment here, the analysis agent needs a domain expert writing its prompts, and the authors decline to call this the leap it resembles. the thinking got replaced, and the hands did not so the question i cannot answer for my own setup: which step of your loop still stops dead until you sit down and type something bookmark this one. the four parts that turn one model into a line that runs like this, the queue, the rooms, the write permissions and the gate, are built file by file in the piece below ↓

sorry, they just did WHAT someone gave a machine one disease name, the leading cause of blindness in the developed world with 1.5 million americans already in its path, and it came back pointing at a drug that has sat in pharmacies for years under a different label: 551 papers read in 30 minutes against the 294 hours a human would have needed, and the loop that did it is public on GitHub most agent setups answer one question at a time, so the ceiling on the work is the quality of the question you happened to think of this one was handed a single question and wrote the second one itself. turns out that follow-up is where the real find was: a target called ABCA1, upregulated threefold, in an experiment no human ordered i read the whole paper looking for the trick, and the trick is structural. that is the second question, and it is the gap between an assistant and a factory: - hand the loop a field rather than a task: it was given a disease, and choosing the mechanism was part of its job - make it rank before it spends: 151 papers in, ten candidate mechanisms out, scored against each other before anything touched a bench - split reading from judging, so the agent that forms the theory is a different agent from the one grading it - close every cycle on physical reality: the verdict was an experiment, and another model's opinion was never allowed to stand in for one - feed each result back as the next question rather than a log line, which is the step almost nobody builds - search what already passed inspection first: the winner was an approved compound with a safety file already on record - write down what the round learned before opening the next one, so round two starts where round one stopped my read, and i think it is the uncomfortable one: reading was the entire bottleneck in that field, and everybody spent the decade optimising the writing. people ran every physical experiment here, the analysis agent needs a domain expert writing its prompts, and the authors decline to call this the leap it resembles. the thinking got replaced, and the hands did not so the question i cannot answer for my own setup: which step of your loop still stops dead until you sit down and type something bookmark this one. the four parts that turn one model into a line that runs like this, the queue, the rooms, the write permissions and the gate, are built file by file in the piece below ↓

Argona

32,475 просмотров • 18 дней назад

Probably you know them already, the tips works to me are -Simplify the shapes of the character and animate first that simple forms -Create model sheets and trace it -Use the key draws to create the inbetweens -Sculpt a simple model, pose it and copy what you see (1/2) +

Probably you know them already, the tips works to me are -Simplify the shapes of the character and animate first that simple forms -Create model sheets and trace it -Use the key draws to create the inbetweens -Sculpt a simple model, pose it and copy what you see (1/2) +

Tlauz - Open Comms

357,382 просмотров • 1 год назад

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.

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 🚀

244,885 просмотров • 1 месяц назад

I STOPPED REVIEWING MY OWN AGENT, SOMETHING ELSE DOES IT NOW I used to read every diff it produced and approve most of them, because an agent grading itself always says the work is good. -> Now a second model with different instructions tries to break the work first, and I only read what survived. Here is what is actually in the folder that took over the night shift: • the brief > CONTRACT.md -- what it may touch, and what it may never touch. > VISION.md -- the destination, so turn 47 still knows why it started. • the gate > judge/ -- a different model, never the one that wrote the code. > break-it.md -- it opens the page, clicks, screenshots, reports back. > -- no opinion, just zero or non-zero. > shift.yml -- 03:30 every night, laptop closed. • the memory > receipts/ -- one folder per night, dated and graded. > STATE.md -- where it stopped and what it escalated. > lessons.log -- the flaky test, written down once instead of rediscovered weekly. • the brakes > caps.json -- turn limit, retry limit, spend limit. > -- written on day one, used never. The generator decides what your loop can produce -> The judge decides what it refuses to produce. One of those is the part everyone builds -> The other is why most loops quietly fail. Bookmark it & Read Full breakdown below ↓

I STOPPED REVIEWING MY OWN AGENT, SOMETHING ELSE DOES IT NOW I used to read every diff it produced and approve most of them, because an agent grading itself always says the work is good. -> Now a second model with different instructions tries to break the work first, and I only read what survived. Here is what is actually in the folder that took over the night shift: • the brief > CONTRACT.md -- what it may touch, and what it may never touch. > VISION.md -- the destination, so turn 47 still knows why it started. • the gate > judge/ -- a different model, never the one that wrote the code. > break-it.md -- it opens the page, clicks, screenshots, reports back. > -- no opinion, just zero or non-zero. > shift.yml -- 03:30 every night, laptop closed. • the memory > receipts/ -- one folder per night, dated and graded. > STATE.md -- where it stopped and what it escalated. > lessons.log -- the flaky test, written down once instead of rediscovered weekly. • the brakes > caps.json -- turn limit, retry limit, spend limit. > -- written on day one, used never. The generator decides what your loop can produce -> The judge decides what it refuses to produce. One of those is the part everyone builds -> The other is why most loops quietly fail. Bookmark it & Read Full breakdown below ↓

slash1s

37,556 просмотров • 25 дней назад

Alright, now that we know *what* an agent is, how does it actually work? When you ask for help on a task, the agent plans a series of steps and executes them directly in the application on your behalf, using the tools it has access to. Say you are booking a local service or trying to organize your inbox (which typically takes multiple steps): the AI model first plans how to achieve the task using its existing knowledge and then interacts with your inbox to execute the task. The agent will continue until it is confident the task has been successfully completed.

Alright, now that we know what an agent is, how does it actually work? When you ask for help on a task, the agent plans a series of steps and executes them directly in the application on your behalf, using the tools it has access to. Say you are booking a local service or trying to organize your inbox (which typically takes multiple steps): the AI model first plans how to achieve the task using its existing knowledge and then interacts with your inbox to execute the task. The agent will continue until it is confident the task has been successfully completed.

Google AI

22,487 просмотров • 9 месяцев назад

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.

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 просмотров • 1 месяц назад

Those building AI coding agents will eventually hit the same wall: No matter how smart the model is, the moment you ask it to write production-ready code with the latest libraries, it starts hallucinating. Everyone is building coding agents these days, but most of them run into the same issue: They're coding with outdated documentation. The root cause isn't the model. It's the retrieval layer. If you are building an AI agent that needs information from specific domains such as code, law, academia, and finance Sign up, connect it to your agent, and you'll see the difference immediately.

Those building AI coding agents will eventually hit the same wall: No matter how smart the model is, the moment you ask it to write production-ready code with the latest libraries, it starts hallucinating. Everyone is building coding agents these days, but most of them run into the same issue: They're coding with outdated documentation. The root cause isn't the model. It's the retrieval layer. If you are building an AI agent that needs information from specific domains such as code, law, academia, and finance Sign up, connect it to your agent, and you'll see the difference immediately.

Jafar Najafov

19,801 просмотров • 1 месяц назад

Visualizer of our MultiAgentRouter 🤖 The MultiAgentRouter is an all-new multi-agent structure that leverages a hierarchical pattern to select the most specialized agent for your task. Here's how it works: Step 1. You give a task. Step 2. The Boss Agent Routes your task to the most specialized Agent Step 3. The selected agent returns your response! Get started with it now below ⬇️ Thanks to WE!SS for the visualizer!!

Visualizer of our MultiAgentRouter 🤖 The MultiAgentRouter is an all-new multi-agent structure that leverages a hierarchical pattern to select the most specialized agent for your task. Here's how it works: Step 1. You give a task. Step 2. The Boss Agent Routes your task to the most specialized Agent Step 3. The selected agent returns your response! Get started with it now below ⬇️ Thanks to WE!SS for the visualizer!!

swarms

32,372 просмотров • 1 год назад

seedance 2.0 + my v2 AI UGC prompting system is giving insane results i spent the last 24 hours generating over 200 seedance 2.0 videos to figure out the best prompting framework system for AI UGC this video was made with 1 prompt and 1 tool, no editing was done to the video this was just a prompt to a video this is by far the best model i've ever used and the craziest part is that it can be fully automated this is the first time we can actually automate high quality ai ugc at this level bytedance owns tiktok so this model is trained on millions of high quality ugc videos. you just need to know how to extract that and call it in your prompt. we are so early... it's insane

seedance 2.0 + my v2 AI UGC prompting system is giving insane results i spent the last 24 hours generating over 200 seedance 2.0 videos to figure out the best prompting framework system for AI UGC this video was made with 1 prompt and 1 tool, no editing was done to the video this was just a prompt to a video this is by far the best model i've ever used and the craziest part is that it can be fully automated this is the first time we can actually automate high quality ai ugc at this level bytedance owns tiktok so this model is trained on millions of high quality ugc videos. you just need to know how to extract that and call it in your prompt. we are so early... it's insane

Miko

81,462 просмотров • 6 месяцев назад

2.2 million TikTok Shop Views Just one of the 348+ videos my AI Agent made today The account is @seedsynergy This is a real life money printer 1. like 2. comment "AI AGENT" & 3. RT I'll send the info on the tool to you. (must be following me Ellie Jones )

2.2 million TikTok Shop Views Just one of the 348+ videos my AI Agent made today The account is @seedsynergy This is a real life money printer 1. like 2. comment "AI AGENT" & 3. RT I'll send the info on the tool to you. (must be following me Ellie Jones )

Ellie Jones

21,882 просмотров • 1 год назад

Stanford researchers did it again. They just built the agent-native version of Git. When an agent works on a longer task, the run builds up a lot of state. This includes files edited/created, a dev server, a database, installed packages, KV cache, etc. Say the agent is at step 10 and makes a mistake, maybe it misreads a traceback and rewrites a file that was actually fine. The tests start failing, and the run goes off track, although everything through step eight was correct. By default, the agent just tries to fix it, which creates more edits and tool calls. This burns more tokens and grows the context. The other options are a person stepping in to redirect it or restarting the whole run from step one. That's wasteful, because it pays for every model/tool call again and re-prefills the context. Moreover, since an agent's run is non-deterministic, it doesn't reproduce the same early steps anyway. The reason it's hard to just jump back exactly to a previous correct step and resume from there is that the trajectory is only a message log. It records what the agent said and which tools it called, but not the live state underneath. That state includes things like memory, open file handles, child processes, installed packages, /tmp, and KV cache. None of that is in the log. Git can version the files, but it doesn't snapshot the running process or the KV cache. Checking out step eight moves the files back, but the process is still sitting in step-ten memory with a cold cache. Shepherd is a runtime layer by Stanford that records the run as a trace of typed events rather than a flat log. Each agent-environment interaction becomes a commit, similar to Git, but it tracks the live run. Its commit includes the agent process and the filesystem together, copy-on-write, so a branch carries the actual state and not just the files. Going back to a previous step is then a single call that forks from that commit and continues from the exact state. The copy-on-write fork is roughly five times faster than docker commit, and because the prompt prefix through step eight is unchanged, the KV cache is reused over 95% on replay, so early steps aren't reprocessed again. Once the run can be forked, a meta-agent can sit on top and operate it. It watches the trace and reverts as soon as it looks wrong, before the bad write is committed. In practice, it's just Python calling fork, replay, and revert on the trace, rather than a separate control plane wired into the harness. Not everything is reversible though. Files and sandbox changes undo themselves, but a database write has no automatic undo, so it needs a matching undo step set up in advance. Something external, like a sent email or a real charge, can't be undone, so the supervisor's job there is to catch it before it fires. They tested this on a few public benchmarks. On CooperBench, where two agents work on the same codebase, adding a live supervisor took the pair-coding pass rate from 28.8% to 54.7%. It's still early and labeled alpha. The benefit mostly shows up when a run gets branched a lot over a heavy sandbox state, which is exactly where restarting wastes the most tokens and time. If Git was made to make file changes reversible, Shepherd is trying to do the same thing for a live agent run. Shepherd Repo: (don't forget to star it ⭐ ) That said, Shepherd reverts a bad step inside a run. The harness around it, the prompts, tools, and checks the supervisor relies on, still drifts across runs as models and dependencies change. Akshay wrote about making that harness repair itself, where a failing trace gets diagnosed, the fix is verified against the exact input that failed, and the failure is locked as a regression test so it can't recur. Read it below.

Stanford researchers did it again. They just built the agent-native version of Git. When an agent works on a longer task, the run builds up a lot of state. This includes files edited/created, a dev server, a database, installed packages, KV cache, etc. Say the agent is at step 10 and makes a mistake, maybe it misreads a traceback and rewrites a file that was actually fine. The tests start failing, and the run goes off track, although everything through step eight was correct. By default, the agent just tries to fix it, which creates more edits and tool calls. This burns more tokens and grows the context. The other options are a person stepping in to redirect it or restarting the whole run from step one. That's wasteful, because it pays for every model/tool call again and re-prefills the context. Moreover, since an agent's run is non-deterministic, it doesn't reproduce the same early steps anyway. The reason it's hard to just jump back exactly to a previous correct step and resume from there is that the trajectory is only a message log. It records what the agent said and which tools it called, but not the live state underneath. That state includes things like memory, open file handles, child processes, installed packages, /tmp, and KV cache. None of that is in the log. Git can version the files, but it doesn't snapshot the running process or the KV cache. Checking out step eight moves the files back, but the process is still sitting in step-ten memory with a cold cache. Shepherd is a runtime layer by Stanford that records the run as a trace of typed events rather than a flat log. Each agent-environment interaction becomes a commit, similar to Git, but it tracks the live run. Its commit includes the agent process and the filesystem together, copy-on-write, so a branch carries the actual state and not just the files. Going back to a previous step is then a single call that forks from that commit and continues from the exact state. The copy-on-write fork is roughly five times faster than docker commit, and because the prompt prefix through step eight is unchanged, the KV cache is reused over 95% on replay, so early steps aren't reprocessed again. Once the run can be forked, a meta-agent can sit on top and operate it. It watches the trace and reverts as soon as it looks wrong, before the bad write is committed. In practice, it's just Python calling fork, replay, and revert on the trace, rather than a separate control plane wired into the harness. Not everything is reversible though. Files and sandbox changes undo themselves, but a database write has no automatic undo, so it needs a matching undo step set up in advance. Something external, like a sent email or a real charge, can't be undone, so the supervisor's job there is to catch it before it fires. They tested this on a few public benchmarks. On CooperBench, where two agents work on the same codebase, adding a live supervisor took the pair-coding pass rate from 28.8% to 54.7%. It's still early and labeled alpha. The benefit mostly shows up when a run gets branched a lot over a heavy sandbox state, which is exactly where restarting wastes the most tokens and time. If Git was made to make file changes reversible, Shepherd is trying to do the same thing for a live agent run. Shepherd Repo: (don't forget to star it ⭐ ) That said, Shepherd reverts a bad step inside a run. The harness around it, the prompts, tools, and checks the supervisor relies on, still drifts across runs as models and dependencies change. Akshay wrote about making that harness repair itself, where a failing trace gets diagnosed, the fix is verified against the exact input that failed, and the failure is locked as a regression test so it can't recur. Read it below.

Avi Chawla

441,021 просмотров • 1 месяц назад

The architecture of this new world model is one of the most interesting things I've seen lately: Let me first explain how most world models work: They predict and render one frame at a time. If you are navigating in one of these worlds, and you look left, the model draws whatever looks right in the moment. Every time you change your viewpoint, the model has to imagine what should be there again, so it's very common for these models to "forget" what's in the world. For example, if you put a toy on the table, look away, then look back, the toy might not be there anymore. Tripo AI is releasing its Project Eden model, which works very differently: The model builds the world first, and then renders it based on that map. That map holds the real state of the world: the geometry, every object, where things are, what's already happened. The picture you see on screen gets generated from the map. This architecture flips the whole thing. Now, you get the following: 1. The world stops forgetting. Leave, come back, and the toy is still on the table because it lives in the map, not in the last frame you saw. 2. You can edit the world, and those changes persist for anyone who enters later. 3. Multiple people and AI agents can coexist in the world and see it from different perspectives. This is early research, but it's looking really promising. They just raised nearly $200M across two rounds to build it out. Tripo will be at SIGGRAPH 2026 (July 19–23, Los Angeles Convention Center). If you work in 3D, embodied AI, simulation, or anything spatial, go connect with them there.

The architecture of this new world model is one of the most interesting things I've seen lately: Let me first explain how most world models work: They predict and render one frame at a time. If you are navigating in one of these worlds, and you look left, the model draws whatever looks right in the moment. Every time you change your viewpoint, the model has to imagine what should be there again, so it's very common for these models to "forget" what's in the world. For example, if you put a toy on the table, look away, then look back, the toy might not be there anymore. Tripo AI is releasing its Project Eden model, which works very differently: The model builds the world first, and then renders it based on that map. That map holds the real state of the world: the geometry, every object, where things are, what's already happened. The picture you see on screen gets generated from the map. This architecture flips the whole thing. Now, you get the following: 1. The world stops forgetting. Leave, come back, and the toy is still on the table because it lives in the map, not in the last frame you saw. 2. You can edit the world, and those changes persist for anyone who enters later. 3. Multiple people and AI agents can coexist in the world and see it from different perspectives. This is early research, but it's looking really promising. They just raised nearly $200M across two rounds to build it out. Tripo will be at SIGGRAPH 2026 (July 19–23, Los Angeles Convention Center). If you work in 3D, embodied AI, simulation, or anything spatial, go connect with them there.

Santiago

30,244 просмотров • 2 месяцев назад