Loading video...

Video Failed to Load

There was a problem loading this video. This could be due to a temporary network issue or the video might be unavailable.

AMAZON SENIOR DEVELOPER BUILT A CONTEXT PIPELINE THAT DECIDES WHAT THE MODEL EVEN GETS TO SEE Most teams still dump the entire codebase into every prompt and hope it sorts itself out. He ranked every file by relevance to the task instead of how recently it was touched. A... show more

wast3

3,882 subscribers

23,868 views • 17 days ago •via X (Twitter)

Anya Rossi• Live Now

Private livecam show

0 Comments

No comments available

Comments from the original post will appear here

Related Videos

I MADE MY AI AGENT 10X FASTER WITHOUT CHANGING THE MODEL not a smarter model, not a bigger context window, not another clever prompt the same kind of AI that designs vaccines for viruses we have not even met yet was spending two minutes opening the wrong files just to hand me a brief from three months ago the problem was never capability, it was the scaffolding that piled up around my agent by accident, folder by folder an agent does not think in your categories, it searches from scratch every single time, and your tidy human folders are a maze to it the fix was almost stupidly small, one index file at the root of each big folder and a few numbers in front of the folder names slowest task dropped from 2 minutes to 26 seconds, fastest ones hit 10, zero model changes capability is cheap when the scaffolding around it is broken the article breaks down the whole system in 15 minutes ↓

I MADE MY AI AGENT 10X FASTER WITHOUT CHANGING THE MODEL not a smarter model, not a bigger context window, not another clever prompt the same kind of AI that designs vaccines for viruses we have not even met yet was spending two minutes opening the wrong files just to hand me a brief from three months ago the problem was never capability, it was the scaffolding that piled up around my agent by accident, folder by folder an agent does not think in your categories, it searches from scratch every single time, and your tidy human folders are a maze to it the fix was almost stupidly small, one index file at the root of each big folder and a few numbers in front of the folder names slowest task dropped from 2 minutes to 26 seconds, fastest ones hit 10, zero model changes capability is cheap when the scaffolding around it is broken the article breaks down the whole system in 15 minutes ↓

shmidt

36,479 views • 1 month ago

if you use claude code, this will save you real money. the problem: every time you make an edit, your ai rereads the whole codebase to figure out what changed. tens of thousands of tokens, every turn, for context it already had. this repo fixes it. it’s called code review graph, and it just maps your entire codebase: every file, every function, every connection laid out so you can see the actual shape of your project how to set up (2 min): 1. pip install code-review-graph 2. code-review-graph install - auto-configures Claude Code, Cursor, Codex, Gemini CLI + more 3. code-review-graph build change one function, and it traces exactly what that touches... every caller, every dependent file, every test. your ai only reads what's affected, not the whole repo. and the savings are wild. a task that used to burn ~100,000 tokens (about a dollar) now runs closer to a penny.

if you use claude code, this will save you real money. the problem: every time you make an edit, your ai rereads the whole codebase to figure out what changed. tens of thousands of tokens, every turn, for context it already had. this repo fixes it. it’s called code review graph, and it just maps your entire codebase: every file, every function, every connection laid out so you can see the actual shape of your project how to set up (2 min): 1. pip install code-review-graph 2. code-review-graph install - auto-configures Claude Code, Cursor, Codex, Gemini CLI + more 3. code-review-graph build change one function, and it traces exactly what that touches... every caller, every dependent file, every test. your ai only reads what's affected, not the whole repo. and the savings are wild. a task that used to burn ~100,000 tokens (about a dollar) now runs closer to a penny.

Alvaro Cintas

74,076 views • 21 days ago

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 views • 9 months ago

🚨 Do you understand what Claude just quietly dropped while everyone was distracted? 1 million tokens. Let me explain what that actually means because the number alone doesn't hit right. > A senior engineer joins a company and spends 3 to 6 months just reading code.. Understanding how things connect. Learning where the bugs hide. Why that one file nobody touches exists. It takes months because a codebase is massive and human memory is small. > Claude just loaded the entire thing in one prompt. 30 seconds. Every file, Every function, Every line. All of it. Sitting in memory like it's been working there for years. And it scored highest among every single frontier model. Not GPT.. Not Gemini, Nobody. > Yesterday Amazon's AI nuked production because it couldn't see the full picture - it made a decision with partial context and deleted everything. Today an AI can hold 1 million tokens of context at once. That's the fix. That's the "before and after" moment for AI coding. > 600 images in one request. Entire PDFs. Full repos. And they dropped it on a Friday on all plans like it was a patch note. The scariest AI updates aren't the ones with press conferences. They're the ones that drop in a tweet at 6pm and change everything by Monday morning.

🚨 Do you understand what Claude just quietly dropped while everyone was distracted? 1 million tokens. Let me explain what that actually means because the number alone doesn't hit right. > A senior engineer joins a company and spends 3 to 6 months just reading code.. Understanding how things connect. Learning where the bugs hide. Why that one file nobody touches exists. It takes months because a codebase is massive and human memory is small. > Claude just loaded the entire thing in one prompt. 30 seconds. Every file, Every function, Every line. All of it. Sitting in memory like it's been working there for years. And it scored highest among every single frontier model. Not GPT.. Not Gemini, Nobody. > Yesterday Amazon's AI nuked production because it couldn't see the full picture - it made a decision with partial context and deleted everything. Today an AI can hold 1 million tokens of context at once. That's the fix. That's the "before and after" moment for AI coding. > 600 images in one request. Entire PDFs. Full repos. And they dropped it on a Friday on all plans like it was a patch note. The scariest AI updates aren't the ones with press conferences. They're the ones that drop in a tweet at 6pm and change everything by Monday morning.

Tuki

206,309 views • 5 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

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

this is f*cking gold. I told Fable 5 it's a retiring engineer on its last week, and its final task is to leave everything behind for the team replacing it. It read my entire git history. Every mistake, every dead end, every fix that finally worked. Then it started writing them into skill files so the cheaper models replacing it never repeat what I got wrong. Fable leaves July 12. What it leaves behind is up to you. (full breakdown in the article below)

this is f*cking gold. I told Fable 5 it's a retiring engineer on its last week, and its final task is to leave everything behind for the team replacing it. It read my entire git history. Every mistake, every dead end, every fix that finally worked. Then it started writing them into skill files so the cheaper models replacing it never repeat what I got wrong. Fable leaves July 12. What it leaves behind is up to you. (full breakdown in the article below)

Prajwal Tomar

125,537 views • 1 month ago

SOMEONE MAPPED CHAOS INTO A NAVIGABLE SPACE AND YOUR HERMES AGENT NEEDS THE SAME THING thousands of chaotic data points, each given a position, navigation becomes instant because the structure does the work your vault is the same chaos, hermes lands in it and opens files at random because nothing tells it where to start one index file per major folder with a clear starting point changes everything 2 minutes per task drops to 10 seconds, same agent, same model full breakdown in the article below ↓

SOMEONE MAPPED CHAOS INTO A NAVIGABLE SPACE AND YOUR HERMES AGENT NEEDS THE SAME THING thousands of chaotic data points, each given a position, navigation becomes instant because the structure does the work your vault is the same chaos, hermes lands in it and opens files at random because nothing tells it where to start one index file per major folder with a clear starting point changes everything 2 minutes per task drops to 10 seconds, same agent, same model full breakdown in the article below ↓

leopardracer

23,629 views • 1 month ago

THIS IS HOW FABLE 5 MOVES WHEN YOUR VAULT IS BUILT RIGHT direct path. zero wandering. hits the target on the first try. without structure it's the opposite. 7 files opened. 2 minutes wasted. brief from 3 months ago still missing. one index file per major folder. gives the agent a direct line to what it needs. same task dropped from 2 minutes to 10 seconds. same model. nothing else changed. build the path or watch it search in the dark. full breakdown in the article below ↓

THIS IS HOW FABLE 5 MOVES WHEN YOUR VAULT IS BUILT RIGHT direct path. zero wandering. hits the target on the first try. without structure it's the opposite. 7 files opened. 2 minutes wasted. brief from 3 months ago still missing. one index file per major folder. gives the agent a direct line to what it needs. same task dropped from 2 minutes to 10 seconds. same model. nothing else changed. build the path or watch it search in the dark. full breakdown in the article below ↓

Avid

908,459 views • 1 month ago

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 views • 22 days ago

WTF, GROK BOT JUST MADE AI AGENTS AVAILABLE TO LITERALLY ANYONE – CREATING CONTENT HAS NEVER BEEN THIS EASY, EVEN IF YOU'VE NEVER MADE ANYTHING BEFORE Content was never a talent problem. It's a headcount problem. One person doing research, design, copy, analytics, timing and publishing – that's six jobs. The switching between them is what kills consistency, not a lack of ideas. Here's what one of these setups actually looks like. A Chief of Staff sits in the middle and routes every task. Nothing lands on the human. → Researcher tracks what's actually moving and pulls real sources instead of guesswork → Writer turns that research into finished copy, ready to review → Visualiser gets fed a few reference visuals once, then ships everything in that style → Analyst reads the numbers and tells the rest of the team what worked → Scheduler owns timing and holds the queue → Publisher ships it The part that makes it work: every agent on Grok Bot gets its own persistent computer, browser and file system – and they all share memory. So the research is already sitting inside the draft before the draft starts. No copy-pasting between tools. No approving every step. No human in the middle. You can even teach an agent a repetitive task by recording yourself doing it once. Start recording, do the thing, stop. It learns the pattern. And that's the real shift. Nobody needs AI to tell them what to post. They need it to delete the 40 steps between the idea and the post. Everyone has a backlog of things they've meant to make for months. This is what starts clearing it. Full breakdown of the setup in the article below ↓

WTF, GROK BOT JUST MADE AI AGENTS AVAILABLE TO LITERALLY ANYONE – CREATING CONTENT HAS NEVER BEEN THIS EASY, EVEN IF YOU'VE NEVER MADE ANYTHING BEFORE Content was never a talent problem. It's a headcount problem. One person doing research, design, copy, analytics, timing and publishing – that's six jobs. The switching between them is what kills consistency, not a lack of ideas. Here's what one of these setups actually looks like. A Chief of Staff sits in the middle and routes every task. Nothing lands on the human. → Researcher tracks what's actually moving and pulls real sources instead of guesswork → Writer turns that research into finished copy, ready to review → Visualiser gets fed a few reference visuals once, then ships everything in that style → Analyst reads the numbers and tells the rest of the team what worked → Scheduler owns timing and holds the queue → Publisher ships it The part that makes it work: every agent on Grok Bot gets its own persistent computer, browser and file system – and they all share memory. So the research is already sitting inside the draft before the draft starts. No copy-pasting between tools. No approving every step. No human in the middle. You can even teach an agent a repetitive task by recording yourself doing it once. Start recording, do the thing, stop. It learns the pattern. And that's the real shift. Nobody needs AI to tell them what to post. They need it to delete the 40 steps between the idea and the post. Everyone has a backlog of things they've meant to make for months. This is what starts clearing it. Full breakdown of the setup in the article below ↓

SCOTTY BEAM

4,731,272 views • 4 days ago

This Chinese developer launched Llama 70B locally on a MacBook on a plane and for a full 11 hours without internet ran client projects. He was sitting by the window on a transatlantic flight with a MacBook Pro M4 with 64 GB of memory. WiFi on board cost $25 for the flight. He declined. No cloud API, no connection to Anthropic or OpenAI servers, no internet at all. Just a local Llama 3.3 70B on bf16 and his own orchestrator script. The model runs through llama.cpp. Generation speed, 71 tokens per second. Context around 60,000 tokens. Memory usage, 48.6 GiB out of 64. Battery at takeoff, 3 hours 21 minutes. And he gave the orchestrator this system prompt before takeoff: "You are an offline orchestrator running on a single MacBook. There is no network. The only resources you have are local files in /Users/dev/work, the Llama 70B inference server at localhost:8080, and a battery budget of 3 hours 21 minutes. Process the queue at /Users/dev/work/queue.jsonl (one client task per line). For each task: draft → run local evals → save artefact to /Users/dev/work/done/. Save context checkpoints every 12 tasks so you can resume after a battery swap. Stop only on empty queue or when battery drops below 5%." So the system knows exactly what resources it is running on. It knows it has no connection to the outside world for the next 11 hours. It knows it has finite memory and a finite battery. It knows the human will not intervene until the plane lands. The system runs in 1 loop. Takes a task from the queue, runs it through inference, saves the artifact, writes a checkpoint. Task after task, just like that. And only when the battery drops below 5% does the orchestrator automatically pause, waits for the laptop to switch to the backup power bank, and continues from the last checkpoint. Here is what the system actually writes in his log during the flight: "saved context checkpoint 8 of 12 (pos_min = 488, pos_max = 50118, size = 62.813 MiB)" "restored context checkpoint (pos_min = 488, pos_max = 50118)" "prompt processing progress: n_tokens = 50 / 60 818" "task 37016 done | tps = 71 s tokens text → /Users/dev/work/done/proposal_westside.md" Outside the window, clouds, blue sky, and no WiFi. On the tray, 1 MacBook, an open terminal on 2 screens, and an inference server on localhost. From what I have observed, this is the cleanest offline AI workflow I have seen in the past year: 11 hours of flight, $0 for WiFi, and the entire client queue closed before landing.

This Chinese developer launched Llama 70B locally on a MacBook on a plane and for a full 11 hours without internet ran client projects. He was sitting by the window on a transatlantic flight with a MacBook Pro M4 with 64 GB of memory. WiFi on board cost $25 for the flight. He declined. No cloud API, no connection to Anthropic or OpenAI servers, no internet at all. Just a local Llama 3.3 70B on bf16 and his own orchestrator script. The model runs through llama.cpp. Generation speed, 71 tokens per second. Context around 60,000 tokens. Memory usage, 48.6 GiB out of 64. Battery at takeoff, 3 hours 21 minutes. And he gave the orchestrator this system prompt before takeoff: "You are an offline orchestrator running on a single MacBook. There is no network. The only resources you have are local files in /Users/dev/work, the Llama 70B inference server at localhost:8080, and a battery budget of 3 hours 21 minutes. Process the queue at /Users/dev/work/queue.jsonl (one client task per line). For each task: draft → run local evals → save artefact to /Users/dev/work/done/. Save context checkpoints every 12 tasks so you can resume after a battery swap. Stop only on empty queue or when battery drops below 5%." So the system knows exactly what resources it is running on. It knows it has no connection to the outside world for the next 11 hours. It knows it has finite memory and a finite battery. It knows the human will not intervene until the plane lands. The system runs in 1 loop. Takes a task from the queue, runs it through inference, saves the artifact, writes a checkpoint. Task after task, just like that. And only when the battery drops below 5% does the orchestrator automatically pause, waits for the laptop to switch to the backup power bank, and continues from the last checkpoint. Here is what the system actually writes in his log during the flight: "saved context checkpoint 8 of 12 (pos_min = 488, pos_max = 50118, size = 62.813 MiB)" "restored context checkpoint (pos_min = 488, pos_max = 50118)" "prompt processing progress: n_tokens = 50 / 60 818" "task 37016 done | tps = 71 s tokens text → /Users/dev/work/done/proposal_westside.md" Outside the window, clouds, blue sky, and no WiFi. On the tray, 1 MacBook, an open terminal on 2 screens, and an inference server on localhost. From what I have observed, this is the cleanest offline AI workflow I have seen in the past year: 11 hours of flight, $0 for WiFi, and the entire client queue closed before landing.

Blaze

1,841,161 views • 3 months ago

Rio de Janeiro just became the first city in the world to start reforesting itself with AI drones and the more i read about how it works the cooler it gets: the reason a city would even need this is that dead land is brutally hard to bring back when cattle farming or mining wrecks a piece of land, the soil turns hard and dry and basically dies. left alone it can stay like that for decades the only fix used to be huge crews planting seedlings by hand. one person covers about a hectare a day. at that pace a real forest takes years and a fortune, so most wrecked land just stays dead the company Rio hired is called MORFO. their answer is one drone plus an AI model doing the work of that entire crew it starts with the drone scanning the whole area from above from that scan, the AI studies the soil, the water, the slope, the plants already growing nearby it uses all of that to pick which native species have the best shot at surviving in each exact spot, choosing from a catalog of 300+ local plants once it knows what goes where, the drone flies back over and fires biodegradable seed pods into the ground, 180 every minute each pod holds seeds, nutrients, moisture. a little starter kit for surviving in dead soil flying like that, one drone covers up to 50 hectares a day. the work of a 50-person planting crew and it actually works. they tested it on Brazilian pasture that years of cattle farming had killed. a few months after planting, that same land had grass, bushes, small trees growing again the system keeps learning after the drones leave too. satellites watch what actually grows back, so each new project starts smarter than the last my favorite detail: the AI even decides where NOT to plant it left 16% of one 8,420-hectare site untouched because it detected the forest there was already regrowing on its own easily one of the coolest AI applications i've seen this year

Rio de Janeiro just became the first city in the world to start reforesting itself with AI drones and the more i read about how it works the cooler it gets: the reason a city would even need this is that dead land is brutally hard to bring back when cattle farming or mining wrecks a piece of land, the soil turns hard and dry and basically dies. left alone it can stay like that for decades the only fix used to be huge crews planting seedlings by hand. one person covers about a hectare a day. at that pace a real forest takes years and a fortune, so most wrecked land just stays dead the company Rio hired is called MORFO. their answer is one drone plus an AI model doing the work of that entire crew it starts with the drone scanning the whole area from above from that scan, the AI studies the soil, the water, the slope, the plants already growing nearby it uses all of that to pick which native species have the best shot at surviving in each exact spot, choosing from a catalog of 300+ local plants once it knows what goes where, the drone flies back over and fires biodegradable seed pods into the ground, 180 every minute each pod holds seeds, nutrients, moisture. a little starter kit for surviving in dead soil flying like that, one drone covers up to 50 hectares a day. the work of a 50-person planting crew and it actually works. they tested it on Brazilian pasture that years of cattle farming had killed. a few months after planting, that same land had grass, bushes, small trees growing again the system keeps learning after the drones leave too. satellites watch what actually grows back, so each new project starts smarter than the last my favorite detail: the AI even decides where NOT to plant it left 16% of one 8,420-hectare site untouched because it detected the forest there was already regrowing on its own easily one of the coolest AI applications i've seen this year

Ole Lehmann

13,123 views • 1 month 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.

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 🚀

244,567 views • 1 month ago

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. Repeat until stop_reason is end_turn Step 4 is the whole thing, the write-back is what makes it an agent The model has to see what actually happened before it decides the next move That's the entire loop... understand this cold before you reach for a framework

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. Repeat until stop_reason is end_turn Step 4 is the whole thing, the write-back is what makes it an agent The model has to see what actually happened before it decides the next move That's the entire loop... understand this cold before you reach for a framework

Daniel San

12,514 views • 2 months ago

SOMEONE TURNED THEIR TEAM'S TASK TRACKER INTO A 3D ISLAND instead of a boring list of tasks, your teams work is a little island that grows as you get stuff done > you assign tasks right in slack, just type who its for, the points, and the due date > finish a task and you get to place a building on the island > get your work rejected and the building collapses into rubble > the rubble stays there forever, so everyone can see it > each new sprint starts a fresh island so over time the island fills up with buildings for all the work your team actually finished, and the rubble is a reminder of what got rejected. its open source, so any team can set it up. way more fun than staring at a to do list all day

SOMEONE TURNED THEIR TEAM'S TASK TRACKER INTO A 3D ISLAND instead of a boring list of tasks, your teams work is a little island that grows as you get stuff done > you assign tasks right in slack, just type who its for, the points, and the due date > finish a task and you get to place a building on the island > get your work rejected and the building collapses into rubble > the rubble stays there forever, so everyone can see it > each new sprint starts a fresh island so over time the island fills up with buildings for all the work your team actually finished, and the rubble is a reminder of what got rejected. its open source, so any team can set it up. way more fun than staring at a to do list all day

Om Patel

12,952 views • 1 month ago

FABLE 5 + HIGGSFIELD BUILD $35K SITES FOR ~$27. BUT EVERYONE HIDES THE 4 THINGS THAT BREAK. HERE THEY ARE - WITH THE FIX. everyone shows the finished build. nobody shows where it breaks on the way there. this is what actually goes wrong, and how to fix each one: → BREAK 1: the clips look generic fix: never prompt "cinematic hero video." give Higgsfield the exact beat - "a slow push through fog as the logo resolves." specificity is the whole game. → BREAK 2: the scroll janks on mobile fix: tell Claude to scrub a compressed video, not full-res, and cap ScrollTrigger to 60fps. Lenis + will-change, never raw scroll listeners. → BREAK 3: it loads like molasses fix: one line - "extract frames, convert to WebP, lazy-load everything below the fold." a heavy hero is what kills these builds. → BREAK 4: the polish reads fake fix: film grain, bloom, vignette at 10-15% - not 100%. subtle reads $35k. cranked reads AI slop. taste is the dial. get those four right and nobody can tell it wasn't a studio. CONNECT HIGGSFIELD (MCP): add it as a custom connector in Claude Code: - mcp_servers: - higgsfield: - url: " one OAuth flow. Claude generates and pulls clips directly - no exporting by hand. the tools ship the site in an hour. these four fixes are what make it look like it took a month. Follow me, reply "BRAKE" and I'll send you the full step-by-step Playbook. full breakdown in the article 👇

ZEUS⚡️

11,765 views • 1 month ago

whoever leaked this has bigger balls than sense Google Research and MIT ran the same agent jobs 260 different ways for Nature last month: they held the prompts, the tools and the compute budget identical and moved nothing but the wiring between the agents, and the same work swung from 70% worse than a single agent to 80.8% better, averaging out at 0.0% i ran my own single agent against the task list first and it cleared 6 of 10 alone, already past the line where a crew starts subtracting this is Graph Engineering, the layer that decides whether a crew is worth 80% more or 70% less, and it installs into the agent you already pay for: - score your solo agent on the real task first: above roughly 45% success that study predicts zero to negative returns from any crew you put around it - under that line, put one supervisor over the fan out: crews with no correction step amplified their own errors to 17.2x the single agent rate, supervised aggregation held it to 4.4x - give every worker one output and let none of them read a peer's draft, so a wrong step reaches the supervisor instead of four other agents - run the comparison again after every model upgrade, because a better model raises your baseline and a higher baseline is what makes a crew stop paying - keep the single agent alive as the control, the only number that says the wiring is earning its calls turns out the shape does not travel: the biggest win came off a finance task under one supervisor and the worst collapse off a planning task with independent agents my position, and it is the arguable one: a crew is a bet on your own diagram, and the model you pick moves that bet less than one arrow does bookmark this, the three moves that draw those arrows before you pay for one extra call are in the post below ↓

whoever leaked this has bigger balls than sense Google Research and MIT ran the same agent jobs 260 different ways for Nature last month: they held the prompts, the tools and the compute budget identical and moved nothing but the wiring between the agents, and the same work swung from 70% worse than a single agent to 80.8% better, averaging out at 0.0% i ran my own single agent against the task list first and it cleared 6 of 10 alone, already past the line where a crew starts subtracting this is Graph Engineering, the layer that decides whether a crew is worth 80% more or 70% less, and it installs into the agent you already pay for: - score your solo agent on the real task first: above roughly 45% success that study predicts zero to negative returns from any crew you put around it - under that line, put one supervisor over the fan out: crews with no correction step amplified their own errors to 17.2x the single agent rate, supervised aggregation held it to 4.4x - give every worker one output and let none of them read a peer's draft, so a wrong step reaches the supervisor instead of four other agents - run the comparison again after every model upgrade, because a better model raises your baseline and a higher baseline is what makes a crew stop paying - keep the single agent alive as the control, the only number that says the wiring is earning its calls turns out the shape does not travel: the biggest win came off a finance task under one supervisor and the worst collapse off a planning task with independent agents my position, and it is the arguable one: a crew is a bet on your own diagram, and the model you pick moves that bet less than one arrow does bookmark this, the three moves that draw those arrows before you pay for one extra call are in the post below ↓

Argona

888,475 views • 10 days ago

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 views • 15 days ago

THIS MAN HASN’T OPENED HIS TASK MANAGER IN A WEEK AND HIS ENTIRE LIFE IS STILL ON TRACK he dumped everything into one folder before bed and closed his laptop woke up to a system that had already sorted every note and flagged every stalled project and assigned every task his coworkers are still living in their inboxes the only difference is one folder called CAPTURE and claude running at 8pm every night every idea goes in and claude files it and nothing requires a decision and nothing gets lost

THIS MAN HASN’T OPENED HIS TASK MANAGER IN A WEEK AND HIS ENTIRE LIFE IS STILL ON TRACK he dumped everything into one folder before bed and closed his laptop woke up to a system that had already sorted every note and flagged every stalled project and assigned every task his coworkers are still living in their inboxes the only difference is one folder called CAPTURE and claude running at 8pm every night every idea goes in and claude files it and nothing requires a decision and nothing gets lost

leopardracer

313,569 views • 3 months ago

last night I built a calorie tracking app, fully automated whilst I was asleep... all of these commits you can see here were on autopilot. I first spent an hour creating a large spec list/plan, split into phases and individual tasks. Kept asking Opus 4.5 to see if there's anything it's unsure of. nailed down the architecture (Bun, Nativewind, MySQL, EffectTS, TanStack Query/Mutate, Barcode Scanner API, Nutrition API details), and the schema. then I built my own version of the Ralph Wiggum bot, which is basically a .md file with a list of commands the AI can use to track task status, latest context, and a loop to keep resetting the context window on task completion. when I woke up in the morning the app was fully working and all tasks were complete. there's also 479 tests across 17 files. I don't know how to feel. This is insane. It works.

last night I built a calorie tracking app, fully automated whilst I was asleep... all of these commits you can see here were on autopilot. I first spent an hour creating a large spec list/plan, split into phases and individual tasks. Kept asking Opus 4.5 to see if there's anything it's unsure of. nailed down the architecture (Bun, Nativewind, MySQL, EffectTS, TanStack Query/Mutate, Barcode Scanner API, Nutrition API details), and the schema. then I built my own version of the Ralph Wiggum bot, which is basically a .md file with a list of commands the AI can use to track task status, latest context, and a loop to keep resetting the context window on task completion. when I woke up in the morning the app was fully working and all tasks were complete. there's also 479 tests across 17 files. I don't know how to feel. This is insane. It works.

Richie

85,555 views • 7 months ago