正在加载视频...

视频加载失败

返回首页

This new AI model is crazy fast at coding 🤯 It fixed a broken backend in one prompt and around 83 seconds. I tested Space Bunny inside OpenCode with a FastAPI repo. It had 3 failing tests caused by broken order filtering, unsafe money calculations and incorrect product ranking....

mhdfaran's profile picture

Farhan

78,116 subscribers

60,437 次观看 • 8 天前 •via X (Twitter)

教育科学技术
LIVE

Anya Rossi• Live Now

Private livecam show

3 条评论

Andrew Bolis 的头像
Andrew Bolis8 天前

Adding edge case tests on its own is a nice touch.

Abdul Shakoor 的头像
Abdul Shakoor8 天前

Fast code is nice, but fast bug hunting is the real test.

Genius💡💹🧲 🤖 的头像
Genius💡💹🧲 🤖8 天前

fastapi repo bugs gone that fast honestly kinda scary good

相关视频

I put Space Bunny, the most-used model on OpenCode this week to the test on a broken launch dashboard. With one prompt and no hints, it: → Found two separate logic bugs → Fixed a broken 700% progress calculation → Repaired the task filters → Improved the result from 2/6 to 6/6 passing tests Fast, methodical, and reliable with tools, not just patching the UI, but verifying the fix. Space Bunny is currently free to try on OpenCode
0:27

I put Space Bunny, the most-used model on OpenCode this week to the test on a broken launch dashboard. With one prompt and no hints, it: → Found two separate logic bugs → Fixed a broken 700% progress calculation → Repaired the task filters → Improved the result from 2/6 to 6/6 passing tests Fast, methodical, and reliable with tools, not just patching the UI, but verifying the fix. Space Bunny is currently free to try on OpenCode

Darshal Jaitwar

24,922 次观看 • 4 天前

WeiboAI dropped VibeThinker-3B, so I had to try it locally. this is a 3B model, not a giant frontier system. in the video I load it on my DGX Spark, give it 3 small algorithm problems, stream the actual model output live, then run the generated python through pytest. no benchmark screenshot no canned answer just a tiny local reasoner writing code and real tests deciding if it worked it went 3/3.
2:35

WeiboAI dropped VibeThinker-3B, so I had to try it locally. this is a 3B model, not a giant frontier system. in the video I load it on my DGX Spark, give it 3 small algorithm problems, stream the actual model output live, then run the generated python through pytest. no benchmark screenshot no canned answer just a tiny local reasoner writing code and real tests deciding if it worked it went 3/3.

Joey

16,865 次观看 • 3 个月前

Chinese researchers did it again! OpenBMB just open-sourced MiniCPM5-2B, a dense 2B-parameter model built for reasoning, coding, and tool use on resource-constrained hardware. Artificial Analysis ranked it highest among models under 4B in its Agentic Index comparison. It scored 20, while Granite 4.2 8B scored 9. The model is particularly strong at coding and tool calling, so I tested both capabilities locally. I pulled it onto my machine, connected it to a constrained CI repair agent, and gave it one issue: > A customer reports that retrying checkout with the same idempotency key returns a larger total. The first request returns $109, while the retry returns $118. Find the root cause, fix it without changing the public API contract, and verify the complete test suite. The Python checkout service had 18 tests. Sixteen passed, while two failed on the retry path. The agent could list files, search code, read selected ranges, run approved tests, apply a patch, and inspect its diff. It reproduced the failure, then followed the checkout and idempotency paths through the repository. The model found that shipping was added to mutable order state before the cached result was checked. On retry, the same order already contained shipping, so the calculation added it again. It generated a narrow patch that moved the idempotency check ahead of the mutation without changing the public API. The agent ran the targeted tests and the complete suite. All 18 tests passed. The model was never told which file contained the issue or what change to make. Each test result, search result, and code inspection determined its next action. The video below shows the full trajectory, including the investigation, tool calls, generated patch, diff, and final verification. Everything ran 100% locally on my machine throughout the run. MiniCPM5-2B supports llama.cpp, Ollama, vLLM, SGLang, iOS, Android, and HarmonyOS for local deployment. The model weights, training recipes, reasoning datasets, and UltraX data-refinement system are open-source. GitHub Repo: A 2B model can now inspect a repository, reason across multiple files, modify code, and verify its patch while remaining small enough to target local hardware.
0:29

Chinese researchers did it again! OpenBMB just open-sourced MiniCPM5-2B, a dense 2B-parameter model built for reasoning, coding, and tool use on resource-constrained hardware. Artificial Analysis ranked it highest among models under 4B in its Agentic Index comparison. It scored 20, while Granite 4.2 8B scored 9. The model is particularly strong at coding and tool calling, so I tested both capabilities locally. I pulled it onto my machine, connected it to a constrained CI repair agent, and gave it one issue: > A customer reports that retrying checkout with the same idempotency key returns a larger total. The first request returns $109, while the retry returns $118. Find the root cause, fix it without changing the public API contract, and verify the complete test suite. The Python checkout service had 18 tests. Sixteen passed, while two failed on the retry path. The agent could list files, search code, read selected ranges, run approved tests, apply a patch, and inspect its diff. It reproduced the failure, then followed the checkout and idempotency paths through the repository. The model found that shipping was added to mutable order state before the cached result was checked. On retry, the same order already contained shipping, so the calculation added it again. It generated a narrow patch that moved the idempotency check ahead of the mutation without changing the public API. The agent ran the targeted tests and the complete suite. All 18 tests passed. The model was never told which file contained the issue or what change to make. Each test result, search result, and code inspection determined its next action. The video below shows the full trajectory, including the investigation, tool calls, generated patch, diff, and final verification. Everything ran 100% locally on my machine throughout the run. MiniCPM5-2B supports llama.cpp, Ollama, vLLM, SGLang, iOS, Android, and HarmonyOS for local deployment. The model weights, training recipes, reasoning datasets, and UltraX data-refinement system are open-source. GitHub Repo: A 2B model can now inspect a repository, reason across multiple files, modify code, and verify its patch while remaining small enough to target local hardware.

Akshay 🚀

313,956 次观看 • 22 天前

what the hell did this free stealth model just build 💀 I asked Space Bunny, a Flash-level model, to make Minecraft in one HTML file and left it running it came back with a playable survival game inside my browser I could explore the generated world, break and place blocks, and switch items through the hotbar easily the best result I have got from a coding model on this test Space Bunny is free on OpenCode for the next week prompt below:
2:19

what the hell did this free stealth model just build 💀 I asked Space Bunny, a Flash-level model, to make Minecraft in one HTML file and left it running it came back with a playable survival game inside my browser I could explore the generated world, break and place blocks, and switch items through the hotbar easily the best result I have got from a coding model on this test Space Bunny is free on OpenCode for the next week prompt below:

Israfil

42,443 次观看 • 9 天前

A Robinhood Chain receipt machine that remembers what the changelog forgets. It's called RUMZO. Most tools show you a builder token's chart. RUMZO shows you its commits and its contract at the same moment, then saves the proof so the story can't be rewritten later. [give it a token address and a public GitHub repo] → Pulls source, tests, docs and manifests at one pinned commit. → Reads launch stage, fees, buyback and locker flags at one pinned block. → Saves both sides as a receipt with commit, block number and observation time. → Run it again later and it diffs the two receipts. If a dev says the tests are done, the fees changed or the locker is on, RUMZO does not read the announcement and guess. It already has the receipt from last time. Which source files changed, whether tests and docs moved with them, which contract settings flipped in between - all of it side by side. Missing data stays marked unknown, nothing gets filled in. Local web UI, a CLI, JSON and Markdown exports. And the whole thing is intentionally READ ONLY. No wallet. No signer. No repo code executed. No buy button. Small raccoon. Lots of receipts. Open source, repo in the first reply ↓
0:30

A Robinhood Chain receipt machine that remembers what the changelog forgets. It's called RUMZO. Most tools show you a builder token's chart. RUMZO shows you its commits and its contract at the same moment, then saves the proof so the story can't be rewritten later. [give it a token address and a public GitHub repo] → Pulls source, tests, docs and manifests at one pinned commit. → Reads launch stage, fees, buyback and locker flags at one pinned block. → Saves both sides as a receipt with commit, block number and observation time. → Run it again later and it diffs the two receipts. If a dev says the tests are done, the fees changed or the locker is on, RUMZO does not read the announcement and guess. It already has the receipt from last time. Which source files changed, whether tests and docs moved with them, which contract settings flipped in between - all of it side by side. Missing data stays marked unknown, nothing gets filled in. Local web UI, a CLI, JSON and Markdown exports. And the whole thing is intentionally READ ONLY. No wallet. No signer. No repo code executed. No buy button. Small raccoon. Lots of receipts. Open source, repo in the first reply ↓

Nekt0

57,088 次观看 • 26 天前

I wanted to try something quick with Muse Spark 1.3, so I had it build this pirate platformer. It was incredibly fast and worked flawlessly at one-shot! Now I’m enhancing the artwork, sprites and adding new levels 😊This is such a fun coding model, it just works, fast and cheap!
0:45

I wanted to try something quick with Muse Spark 1.3, so I had it build this pirate platformer. It was incredibly fast and worked flawlessly at one-shot! Now I’m enhancing the artwork, sprites and adding new levels 😊This is such a fun coding model, it just works, fast and cheap!

Derya Unutmaz, MD

29,951 次观看 • 1 个月前

watch this anon. i gave NVIDIA's biggest model ever a single task. 100 minutes and 440,000 tokens later, it had rendered nothing. not one important thing on the screen. this is Nemotron 3 Ultra. 550 billion parameters, a hybrid Mamba Transformer MoE, the largest model NVIDIA has ever shipped, and they built it specifically for long-running agentic coding. so i handed it exactly that: build a 3D scene from a spec, multiple files, iterate until the tests pass. the same task a frontier model one shotted in minutes. i genuinely wanted to be impressed. it ran for an hour and forty. burned through 440,000 tokens. wrote every file, passed its own tests, and proudly printed "task complete."the browser was blank. the 3D scene never rendered. not once. and the long horizon agentic behavior was genuinely good. it stayed on task the whole hour and forty, wrote real multi-file code, drove its own tools without derailing. it just couldn't turn any of that into something that actually runs. here's the part that gets me. it's a text model, it cannot see its own output. so it sat there looping on a broken vision tool, trying to "look" at the page, hitting error after error, never once reasoning its way out. it declared victory on an empty screen because it had no way to know the screen was empty. to be fair, i genuinely don't know what quant the NIM was serving, so maybe some of that's on the serving, not the model. but the biggest model NVIDIA has ever made, on the exact task it was designed for, couldn't tell it had built nothing in 100 minutes. same task on a local model, below thread👇.
4:07

watch this anon. i gave NVIDIA's biggest model ever a single task. 100 minutes and 440,000 tokens later, it had rendered nothing. not one important thing on the screen. this is Nemotron 3 Ultra. 550 billion parameters, a hybrid Mamba Transformer MoE, the largest model NVIDIA has ever shipped, and they built it specifically for long-running agentic coding. so i handed it exactly that: build a 3D scene from a spec, multiple files, iterate until the tests pass. the same task a frontier model one shotted in minutes. i genuinely wanted to be impressed. it ran for an hour and forty. burned through 440,000 tokens. wrote every file, passed its own tests, and proudly printed "task complete."the browser was blank. the 3D scene never rendered. not once. and the long horizon agentic behavior was genuinely good. it stayed on task the whole hour and forty, wrote real multi-file code, drove its own tools without derailing. it just couldn't turn any of that into something that actually runs. here's the part that gets me. it's a text model, it cannot see its own output. so it sat there looping on a broken vision tool, trying to "look" at the page, hitting error after error, never once reasoning its way out. it declared victory on an empty screen because it had no way to know the screen was empty. to be fair, i genuinely don't know what quant the NIM was serving, so maybe some of that's on the serving, not the model. but the biggest model NVIDIA has ever made, on the exact task it was designed for, couldn't tell it had built nothing in 100 minutes. same task on a local model, below thread👇.

Sudo su

32,589 次观看 • 3 个月前

Zuckerberg says Meta's Llama race changes in the next 12-18 months: most of the code will be written by AI - not autocomplete, but agents that run tests and improve the model loop. "We're not trying to build a general developer tool. We are trying to build a coding agent and an AI research agent that basically advances Llama research specifically." "Sometime in the next 12 to 18 months, we'll reach the point where most of the code that's going towards these efforts is written by AI." "And I don't mean autocomplete." "I'm talking more like you give it a goal, it can run tests, it can improve things, it can find issues." "It writes higher quality code than the average very good person on the team already." The hidden bottleneck is not chat UX. It is closing the loop between AI research, code, tests, and self-improvement inside the lab.
1:00

Zuckerberg says Meta's Llama race changes in the next 12-18 months: most of the code will be written by AI - not autocomplete, but agents that run tests and improve the model loop. "We're not trying to build a general developer tool. We are trying to build a coding agent and an AI research agent that basically advances Llama research specifically." "Sometime in the next 12 to 18 months, we'll reach the point where most of the code that's going towards these efforts is written by AI." "And I don't mean autocomplete." "I'm talking more like you give it a goal, it can run tests, it can improve things, it can find issues." "It writes higher quality code than the average very good person on the team already." The hidden bottleneck is not chat UX. It is closing the loop between AI research, code, tests, and self-improvement inside the lab.

Karl Mehta

36,347 次观看 • 3 个月前

My new favorite way to build with agents is to organize them into a factory. This is the 4-step workflow: 1) Station. - A custom order gets its own station, so the order does not interfere with the rest of production. - In software, this station is a branch or a work tree. 2) Assembly line. - The agent writes code, changes files, and adds structure. 3) Quality control. - Does it turn on? - Does it fit? - Does it break under pressure? - In software, you run tests, open a preview, and read the logs and screenshots. 4) Shipping. - The product goes out the door. - You merge, deploy the PR, and write release notes. If a product fails quality control, you send feedback to the product team. Then you do the loop again. Write each step of your agent workflow as a factory job.
2:32

My new favorite way to build with agents is to organize them into a factory. This is the 4-step workflow: 1) Station. - A custom order gets its own station, so the order does not interfere with the rest of production. - In software, this station is a branch or a work tree. 2) Assembly line. - The agent writes code, changes files, and adds structure. 3) Quality control. - Does it turn on? - Does it fit? - Does it break under pressure? - In software, you run tests, open a preview, and read the logs and screenshots. 4) Shipping. - The product goes out the door. - You merge, deploy the PR, and write release notes. If a product fails quality control, you send feedback to the product team. Then you do the loop again. Write each step of your agent workflow as a factory job.

The Startup Ideas Podcast (SIP) 🧃

13,017 次观看 • 17 天前

Max Tynan on the gap nobody else is solving. AI coding tools are single-player by design. One developer, one repo, one prompt. Enterprise software is a team sport. Software Factory is the multiplayer layer. Work Orders run inside Cursor, VS Code, or Claude Code through MCP. Requirements, Blueprints, and Tests stay connected in the Knowledge Graph. Engineers keep their IDE. The team gets a system around it.
0:47

Max Tynan on the gap nobody else is solving. AI coding tools are single-player by design. One developer, one repo, one prompt. Enterprise software is a team sport. Software Factory is the multiplayer layer. Work Orders run inside Cursor, VS Code, or Claude Code through MCP. Requirements, Blueprints, and Tests stay connected in the Knowledge Graph. Engineers keep their IDE. The team gets a system around it.

8090

102,280 次观看 • 4 个月前

ANTHROPIC LEAKED A 6-AGENT ASSEMBLY LINE WHERE ONE PROMPT GOES IN AND A FINISHED PR COMES OUT your name shows up in this process exactly once - at the very end, for 5 minutes. prompt → planner → coder → tester → critic → scribe → PR Opus 5 plans once - a bad plan executed by five perfect agents is still a bad result. Sonnet 5 runs 3 parallel lanes - and they only work because the planner made the steps independent. tests loop until green and red goes back to the coder - the critic never sees a broken build. Opus 5 as the critic rejects to the plan, not the code - tests ask "does it run", the critic asks "should it exist". the PR description writes itself from the trace, not from memory - cut this seat and you get 8 PRs nobody can tell apart. one human step in the whole thing: approve or send back - 5 minutes instead of 5 hours. save the setup - 2 models, 6 seats, 2 return paths ↓
0:47

ANTHROPIC LEAKED A 6-AGENT ASSEMBLY LINE WHERE ONE PROMPT GOES IN AND A FINISHED PR COMES OUT your name shows up in this process exactly once - at the very end, for 5 minutes. prompt → planner → coder → tester → critic → scribe → PR Opus 5 plans once - a bad plan executed by five perfect agents is still a bad result. Sonnet 5 runs 3 parallel lanes - and they only work because the planner made the steps independent. tests loop until green and red goes back to the coder - the critic never sees a broken build. Opus 5 as the critic rejects to the plan, not the code - tests ask "does it run", the critic asks "should it exist". the PR description writes itself from the trace, not from memory - cut this seat and you get 8 PRs nobody can tell apart. one human step in the whole thing: approve or send back - 5 minutes instead of 5 hours. save the setup - 2 models, 6 seats, 2 return paths ↓

Sprytix

39,283 次观看 • 1 个月前

How to use GitHub Actions to build a pipeline that tests your code, trains a model, and publishes a new version of it. Bonus: I let GitHub Workspace Copilot do a lot of the work. (This is the best "AI developer" I've tested so far by a mile.) Here is the full video:
18:00

How to use GitHub Actions to build a pipeline that tests your code, trains a model, and publishes a new version of it. Bonus: I let GitHub Workspace Copilot do a lot of the work. (This is the best "AI developer" I've tested so far by a mile.) Here is the full video:

Santiago

148,109 次观看 • 2 年前

ANTHROPIC HAS AN OFFICIAL PROMPT LIBRARY FOR CLAUDE CODE Most people have never opened it. So they write every prompt from scratch. It is copy-paste, tagged by task and by role, across the whole lifecycle: discover → design → build → ship → operate Straight from the page: > what would break if I deleted this helper? > plan this refactor, list the files, do not touch code yet > write tests for this, run them, fix what fails > the test is failing, find out why and fix it This is not a cheat sheet. It is a map of what Claude Code already does for you. Bookmark it before you forget it exists.
0:33

ANTHROPIC HAS AN OFFICIAL PROMPT LIBRARY FOR CLAUDE CODE Most people have never opened it. So they write every prompt from scratch. It is copy-paste, tagged by task and by role, across the whole lifecycle: discover → design → build → ship → operate Straight from the page: > what would break if I deleted this helper? > plan this refactor, list the files, do not touch code yet > write tests for this, run them, fix what fails > the test is failing, find out why and fix it This is not a cheat sheet. It is a map of what Claude Code already does for you. Bookmark it before you forget it exists.

shmidt

524,325 次观看 • 3 个月前

🚨 This Chinese model is INSANE !! I just tested Tencent's Hunyuan AI model to build a professional landing page & the output is mind blowing 🤯 Here's my exact prompt: "Create a modern landing page for a crypto related product with beautiful hero section, 3d elements, pricing table, and a whitelist form" What it generated: → Clean, responsive front-end code → Complex UI components with smooth interactions → Built-in debugging (it fixed its own errors) → Production-ready in minutes This 295B parameter model has 2 insane agents: → Coding Agent (handles the entire front-end) → Working Agent (task planning & multi-tool workflows) It also has a context window of 256k tokens !! Watch the video to see the full build 🔥 Try it: Tencent Hy
0:48

🚨 This Chinese model is INSANE !! I just tested Tencent's Hunyuan AI model to build a professional landing page & the output is mind blowing 🤯 Here's my exact prompt: "Create a modern landing page for a crypto related product with beautiful hero section, 3d elements, pricing table, and a whitelist form" What it generated: → Clean, responsive front-end code → Complex UI components with smooth interactions → Built-in debugging (it fixed its own errors) → Production-ready in minutes This 295B parameter model has 2 insane agents: → Coding Agent (handles the entire front-end) → Working Agent (task planning & multi-tool workflows) It also has a context window of 256k tokens !! Watch the video to see the full build 🔥 Try it: Tencent Hy

AGAFE

110,528 次观看 • 2 个月前

THIS GUY USED OPUS 4.8 + KIMI K2.6 TO CUT HIS CODING BILL FROM $4,000 TO $700/MO. WITH KIMI RUNNING A 300-AGENT CODING FLOOR, HE STOPPED PAYING CLAUDE TO DO EVERYTHING kimi does the cheap heavy lifting. it can spin up hundreds of agents, push through thousands of steps, write the rough code, expand files, draft tests and handle the repetitive work that burns the most tokens opus 4.8 only comes in where the money is worth spending. first to plan the spec and define the rules, then again to tear apart the output, catch weak logic and flag the bugs that a fast swarm can miss that is what changed the economics. kimi handled the bulk of the volume for a fraction of the price, while opus stayed in the loop as the architect and the reviewer instead of the full-time builder most people still use one model for every step and wonder why their costs explode. this guy split the jobs properly. kimi runs wide and cheap. opus goes deep and skeptical. one builds fast, the other makes sure it should ship the real edge is not finding one perfect model. it is knowing where expensive intelligence actually matters and where cheap parallel output is already enough. that is how a $4,000 workflow turns into a $700 system
1:28

THIS GUY USED OPUS 4.8 + KIMI K2.6 TO CUT HIS CODING BILL FROM $4,000 TO $700/MO. WITH KIMI RUNNING A 300-AGENT CODING FLOOR, HE STOPPED PAYING CLAUDE TO DO EVERYTHING kimi does the cheap heavy lifting. it can spin up hundreds of agents, push through thousands of steps, write the rough code, expand files, draft tests and handle the repetitive work that burns the most tokens opus 4.8 only comes in where the money is worth spending. first to plan the spec and define the rules, then again to tear apart the output, catch weak logic and flag the bugs that a fast swarm can miss that is what changed the economics. kimi handled the bulk of the volume for a fraction of the price, while opus stayed in the loop as the architect and the reviewer instead of the full-time builder most people still use one model for every step and wonder why their costs explode. this guy split the jobs properly. kimi runs wide and cheap. opus goes deep and skeptical. one builds fast, the other makes sure it should ship the real edge is not finding one perfect model. it is knowing where expensive intelligence actually matters and where cheap parallel output is already enough. that is how a $4,000 workflow turns into a $700 system

Gipp 🦅

16,989 次观看 • 4 个月前

deepseek v4.1 flash vs secret model tested both with the same prompt at the highest reasoning available > v4.1 flash took 2 hours to finish both tests and cost $2.60 > secret model finished both in 30 minutes, can’t reveal cost yet v4.1 flash spent a lot of time overthinking and running random tests, which slowed it down thanks AIHubmix for providing api access so i could test v4.1 flash hoping that team will cut down the overthinking before the full release, but for now here are the results which model did better here?
0:41

deepseek v4.1 flash vs secret model tested both with the same prompt at the highest reasoning available > v4.1 flash took 2 hours to finish both tests and cost $2.60 > secret model finished both in 30 minutes, can’t reveal cost yet v4.1 flash spent a lot of time overthinking and running random tests, which slowed it down thanks AIHubmix for providing api access so i could test v4.1 flash hoping that team will cut down the overthinking before the full release, but for now here are the results which model did better here?

J A Z I I

29,132 次观看 • 25 天前

i been running Qwen3.5-35B-A3B UD-Q4_K_XL through Claude Code since llama.cpp merged the Anthropic endpoint. configured it in minutes. everything was great. projects grew from single scripts to multifile systems with 8 modules and 3,000+ lines. then the chains started breaking. 3 to 5 minutes of pure autonomy and suddenly it stops. tool call fails. reprompt. it recovers. 2 minutes later it stops again. the model is fine. the harness is the bottleneck. saw a comment suggesting OpenCode. installed it. pointed it at the same localhost endpoint running the same model on the same GPU. the game is different. instead of stopping on a bad tool call it just keeps going. on wrong read it adjusts. if file not found it retries. the flow is unbroken. i watched it plan a refactor across 8 files, read every module, and start building without a single pause. in Claude Code that same task would have stopped 4 times. the tradeoff is sometimes it loops. same tool call repeated because the model loses track of what it already read. but here is the thing. i choose loops over pauses. a loop you can interrupt and redirect. a broken chain stops the flow and you have to reprompt to get it moving again. someone is solving this at the core level and i have a feeling it is the open source community. the fact that i can run this level of autonomous coding intelligence on a single consumer GPU with 24gb VRAM at 112 tokens per second. respect to the chinese labs. respect to the open source builders making this possible.
2:00

i been running Qwen3.5-35B-A3B UD-Q4_K_XL through Claude Code since llama.cpp merged the Anthropic endpoint. configured it in minutes. everything was great. projects grew from single scripts to multifile systems with 8 modules and 3,000+ lines. then the chains started breaking. 3 to 5 minutes of pure autonomy and suddenly it stops. tool call fails. reprompt. it recovers. 2 minutes later it stops again. the model is fine. the harness is the bottleneck. saw a comment suggesting OpenCode. installed it. pointed it at the same localhost endpoint running the same model on the same GPU. the game is different. instead of stopping on a bad tool call it just keeps going. on wrong read it adjusts. if file not found it retries. the flow is unbroken. i watched it plan a refactor across 8 files, read every module, and start building without a single pause. in Claude Code that same task would have stopped 4 times. the tradeoff is sometimes it loops. same tool call repeated because the model loses track of what it already read. but here is the thing. i choose loops over pauses. a loop you can interrupt and redirect. a broken chain stops the flow and you have to reprompt to get it moving again. someone is solving this at the core level and i have a feeling it is the open source community. the fact that i can run this level of autonomous coding intelligence on a single consumer GPU with 24gb VRAM at 112 tokens per second. respect to the chinese labs. respect to the open source builders making this possible.

Sudo su

67,104 次观看 • 7 个月前

MINIMAX OPEN SOURCED A CODING AGENT THAT LIVES IN YOUR TERMINAL its called MiniMax Code CLI. i gave it one prompt, 19 minutes later i had a working interactive 3d globe. → 249 countries → 14 animated flight arcs across 16 hubs → click a country and the camera flies to it → fresnel atmosphere, bloom, the full pass it did it in 9 phases at about 96 tok/s, committing after every one. Mcode is MIT and it is not locked to MiniMax models, you can bring your own key and point it at claude, gpt, or a local model. setup in your terminal: 1.⁠ ⁠curl -fsSL | bash 2.⁠ ⁠cd into the project and run mcode 3.⁠ ⁠run mcode init . first, that is the step that writes the AGENTS.md the AGENTS.md is doing more work here than the model. build commands, what not to touch, and a rule that it verifies before it calls anything done i will drop the repo below
1:26

MINIMAX OPEN SOURCED A CODING AGENT THAT LIVES IN YOUR TERMINAL its called MiniMax Code CLI. i gave it one prompt, 19 minutes later i had a working interactive 3d globe. → 249 countries → 14 animated flight arcs across 16 hubs → click a country and the camera flies to it → fresnel atmosphere, bloom, the full pass it did it in 9 phases at about 96 tok/s, committing after every one. Mcode is MIT and it is not locked to MiniMax models, you can bring your own key and point it at claude, gpt, or a local model. setup in your terminal: 1.⁠ ⁠curl -fsSL | bash 2.⁠ ⁠cd into the project and run mcode 3.⁠ ⁠run mcode init . first, that is the step that writes the AGENTS.md the AGENTS.md is doing more work here than the model. build commands, what not to touch, and a rule that it verifies before it calls anything done i will drop the repo below

Israfil

25,336 次观看 • 11 天前

Alright guys, wow. Qwen 3.8 27B is basically Qwen 4. What we have here is one of the best one-shot shark survival games I've made with any model. I ran this same prompt through fable, and the biggest thing all models struggle with in this test is the top-down view of the sharks. Fable even had all of the fins reversed and details jumbled on the first turn. Before now, it was really a 2 to 8 turn test to get something final regardless of model. Qwen 3.8 27B just nailed this in one output, no harness at all, just purely the model thoroughly thinking before outputting the final result. On every asset used, different shark variants, everything, it all looks incredibly tight. The boat movement, shark speed and engagement, everything is so much better than I expected, and I was expecting a lot. Imagine what 8 turns would do in a harness? This confirms to me that while Qwen 3.8 27B thinks a lot, for the first time in a local model on the base weights, IT MIGHT ACTUALLY BE WORTH IT This is a Frontier Lab model result, and it ran on my 5090 at 60tps in Q5_K_M. I am just in complete astoundment at what Qwen has accomplished here, and I am incredibly grateful that this model is open source. The speed will also increase drastically with optimisations. This is just an incredible day. What a blessing. More to come!
4:28

Alright guys, wow. Qwen 3.8 27B is basically Qwen 4. What we have here is one of the best one-shot shark survival games I've made with any model. I ran this same prompt through fable, and the biggest thing all models struggle with in this test is the top-down view of the sharks. Fable even had all of the fins reversed and details jumbled on the first turn. Before now, it was really a 2 to 8 turn test to get something final regardless of model. Qwen 3.8 27B just nailed this in one output, no harness at all, just purely the model thoroughly thinking before outputting the final result. On every asset used, different shark variants, everything, it all looks incredibly tight. The boat movement, shark speed and engagement, everything is so much better than I expected, and I was expecting a lot. Imagine what 8 turns would do in a harness? This confirms to me that while Qwen 3.8 27B thinks a lot, for the first time in a local model on the base weights, IT MIGHT ACTUALLY BE WORTH IT This is a Frontier Lab model result, and it ran on my 5090 at 60tps in Q5_K_M. I am just in complete astoundment at what Qwen has accomplished here, and I am incredibly grateful that this model is open source. The speed will also increase drastically with optimisations. This is just an incredible day. What a blessing. More to come!

Kyle Hessling

24,917 次观看 • 1 个月前