How well can Qwen3.5 models debug code? I built... BugFind-15 — 15 buggy snippets across Python, JS, Rust, and Go. Docker sandbox compiles and validates every fix. Two trap scenarios where the code is correct and the model must resist "fixing" it. Tested every Qwen3.5 size from 0.8B to 397B, plus Jackrong's popular distilled model (V2). The 0.8B scored 5%. The 2B scored 10%. At 4B, debugging ability jumps to 69%. The hardest scenario: BF-03, a Rust trap. The code compiles fine — format! borrows, it doesn't move. Not a single model figured this out. From 0.8B to 397B, every one of them "fixed" a bug that doesn't exist. Category C (subtle bugs — mutable defaults, integer overflow, slice aliasing) was 100% across every model 4B and above. Category D (red herring resistance) told the real story — can it resist fixing code that isn't broken? No model scored above 90%. Small models can't debug. Mid-size models fix obvious bugs but fall for traps. Large models fix the hard bugs but still invent problems that don't exist.show more

Got a 16GB GPU? You can run all of... these right now. Tested 4 Qwen3.5-based models on ToolCall-15 & BugFind-15: Models: - Qwen3.5:9b Q8 (Official) - Qwopus v3 Q8 by Jackrong - OmniCoder-9B by Tesslate - Qwen3.5-9b-Sushi-Coder by bigatuna Summary: - ToolCall-15: Qwopus v3 went perfect 30/30, Sushicoder beat base Qwen3.5 - BugFind-15: Omnicoder flipped the script and took #1 at 83% No single model won both, that's the fun part. Open source community is cooking.show more

Qwen3.6 35B A3B can't fill out a paper form... on its own. But give it NVIDIA's LocateAnything-3B — the #1 trending model on HuggingFace — as its eyes, and the two small models get it done together. (The test: place each element at the right pixel position on a blank form image, not type into a field.) Setup: > Qwen is the brain (main model), LocateAnything is the eyes (helper model acting as a tool). > I gave Qwen a new tool: ask "where's the email field?" and LocateAnything returns the exact x, y, width, height. > The blue boxes on the screen are its detections. Look how tight they are — it nails every field. Result: > Qwen3.6 35B A3B + LocateAnything-3B: form completed, all info correct. > Name, DOB, ID, gender, marital status, nationality, email, phone, address, postal code: all landed in the right field areas. > Character-box alignment still a touch loose, but every value is where it belongs. > 9m10s, 224.5k input, 24.3k output, 21 turns. Why it matters: > Qwen alone can't finish this test. Bolt on a 3B model that does exactly one thing > locate > and suddenly it can. > A combination of small models can do the work of a single large one.show more

"I'm not a human." Fed it to Qwen 3.5... 0.8B running locally on my Mac Studio M2 Ultra. It solved it. The CAPTCHA is fake. But sending images to the local model? Very real. I'm not breaking the internet. Yet.show more

✨ I revived my first AI startup from 6... years ago with Claude Code [ 💡 ] Back then it used GPT-3 (this was 2 years before ChatGPT existed!) to generate new startup ideas which then people can vote on And the best startup ideas rise to the top! Back then I made it because people complained they didn't have any ideas to build a startup This week I moved it to its own VPS and installed Claude Code and told it to fix everything, the DB had become big and there was stupid write operations on every page load that it made it very slow Claude Code is excellent at fixing all those small bugs from old projects and quickly fixing them As Garry Tan says "boil the oceans" as in before I'd not have the time to fix these kinds of projects, it wouldn't be worth it, I mean IdeasAI doesn't even make money, but now it takes me an hour to do this and it works again! I also upgraded GPT-3 to xAI's Grok 4.2 for new startup ideasshow more

My project has 39,205 lines of code, and Cursor... can't answer questions about it. Cursor's context seems to be capped at around 10,000 tokens. Unfortunately, this is not enough for any decent-sized project. If you have a large codebase, check out Augment Code. This thing is faaaast! I'm currently using their Visual Studio Code plugin, but you can also use them on JetBrains, Neovim, and even Vim. (I'm a Neovim fan, but Copilot's implementation for Neovim is nowhere as good as Augment Code.) Augment Code was gracious enough to sponsor this post. After you install their extension and run it for the first time, it will index your entire codebase. This is why it can answer questions as fast as it does, regardless of the size of your codebase. Augment Code supports chat and completions like every other AI coding assistant, but its killer feature is "Next Edit." When you make a change, two things happen: 1. The model analyzes the change to determine the ripple effects across your *entire* codebase. 2. The model suggests everything you need to update to ensure everything works correctly. This is pretty wild!show more

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.show more

Fable 5 comes back！It can now build playable game... prototypes. I think it is actually a signal for where AI coding is going. Making a game is not just “write some code.” Even a small browser game needs: game loop；character movement；collision logic；scoring system；UI states；physics tuning；visual feedback；bug fixing；playtesting This is why game prototyping is a great test for AI models. A model cannot fake it with a pretty answer. Either the game runs, or it does not. What impressed me about Fable 5 is that it is useful for the messy middle: turning an idea into mechanics, turning mechanics into code, debugging broken interactions, and iterating until the prototype feels playable. But here is the practical part: I would not use the strongest model for every step. For game building, I would split the workflow: 1. Fable 5 for game design + architecture 2. a fast coding model for routine implementation 3. a vision-capable model for screenshot/UI feedback 4. a cheaper model for docs, test cases, and small fixes 5. fallback when latency, cost, or output quality becomes a problem That is the real AI coding stack. Not “one magic model does everything.” More like: the right model, for the right task, at the right cost, with fallback when things break. This is why I’ve been looking at ZenMux ZenMux. ZenMux gives developers one gateway to access multiple leading AI models, with OpenAI / Anthropic / Google Vertex compatible APIs, cost tracking, quality benchmarks, auto-routing, and compensation when output quality, latency, or throughput falls short. If AI can now make games, the next question is not just “which model is strongest?” It is:how do we manage the whole model workflow Fable 5 shows the creative ceiling. ZenMux is closer to the infrastructure layer you need when AI coding becomes a real production habit.show more

Introducing HermesAgent-20, a new Bench Pack for BenchLocal. 20... scenarios extracted straight from the Hermes Agent source code, run against a REAL Hermes instance. The actual workload you'd put your model through. Why I built BenchLocal in the first place: most benchmarks are too abstract. We use local LLMs for practical work, and finding the right model for YOUR task efficiently is the single most important thing, especially when you're constrained to what fits on your machine. BenchLocal is a framework: providers, models, side-by-side comparison, all in one UI. Bench Packs are the unit of testing: ToolCall-15 and BugFind-15 shipped first, and when I launched the BenchLocal 0.1.0, added StructOutput, ReasonMath, InstructFollow, DataExtract. Now, HermesAgent-20 is the newest. Bench Packs install like VS Code extensions. The SDK is open, write your own, share it, grow the ecosystem. Here's the goal: a community-built, practical evaluation layer for the local LLM space. Early numbers on HermesAgent-20: > GLM 5.1 — 85 > Gemma4 31B — 83 > Qwen3.5 27B — 79 > MiniMax M2.7 — 76 Upgrade to the latest BenchLocal to install HermesAgent-20 (SDK update required).show more

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

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.show more

AI Is Moving Beyond “Generating Videos” — Toward “Generating... Worlds” Over the past two years, AI video models have advanced at an astonishing pace. From Runway and Pika to Sora and Veo, AI-generated videos have become increasingly realistic and more consistent with the physical laws of the real world. Many people believe the next objective is simply to generate videos that are longer, sharper, and more lifelike. But if we take a step back, we can see that the real transformation is not happening in video itself. It is happening in world models. What Is a World Model? In 1943, psychologist Kenneth Craik proposed an idea that would influence artificial intelligence research for decades. He argued that the human brain does not merely react to the outside world. Instead, it maintains an internal model of how the world works. Because we have this internal model, we can predict the outcome of an action before we actually take it. Before crossing a road, we estimate whether a car will pass by. Before catching a ball, we predict its trajectory. These abilities come from continuously simulating the world in our minds, rather than relying entirely on trial and error. This idea later became known by a more formal term: World Model. A world model does not describe a single image or a fixed video clip. It is an internal representation capable of continuously simulating the rules and dynamics of the real world. Why Is AI Research Turning Toward World Models? Because predicting “what comes next” is becoming increasingly central to how AI systems work. Language models predict the next token. Image models predict the next step in the denoising process. Video models predict the next frame. A world model, however, attempts to predict something broader: What should the world look like in the next moment? In 2018, David Ha and Jürgen Schmidhuber proposed in their paper World Models that an intelligent agent could first learn a model of the world, and then use that internal model to plan its actions. The Dreamer series later demonstrated that many complex tasks could be learned by training agents inside an “imagined world.” At the same time, the development of video models such as Sora and Veo led researchers to another realization: A model capable of continuously generating video has already learned, at least implicitly, many of the rules governing the real world. As a result, these two research directions have gradually begun to converge. But Video Is Not Yet a World This is where the distinction is often misunderstood. For a world model to support meaningful real-time interaction, it must solve several critical problems. Most video models today are essentially answering one question: What should the next frame look like? A true world model needs to answer much more: What happens if I take one step forward? If I walk behind a building and then return, will the building still be there? If I suddenly change the camera angle, will the entire space remain consistent? If I enter a command such as: “Summon a dragon.” Will the world respond immediately? In other words, a world model must do more than generate content. It must understand space. It must understand time. It must understand causality. And it must understand interaction. Moving from watching to participating is where the real difficulty of world models begins. World Models Are Entering the Interactive Era One of the latest attempts in this direction is Alaya World, recently open-sourced by Alaya World, or Alaya Lab. Instead of generating a fixed video clip, it generates a world that users can explore in real time. Users can begin with text, an image, or a video, enter the generated scene, move freely through it, and introduce new prompts at any moment during generation. The world responds immediately. According to the publicly released information, Alaya World provides: Real-time streaming generation at 720p and 24 FPS Stable continuous exploration for more than one minute The ability to switch prompts and trigger skills or events during generation Model weights and inference code released under the Apache 2.0 License Training code and datasets planned for future release What makes these capabilities important is not simply the technical specifications. It is that the generated “world” can now support continuous interaction. The official demo shows that users can genuinely control, transform, and explore the generated environment. AI Is Evolving From a Tool Into an Environment Over the past few years, most discussions around AI have focused on content generation. Generating text. Generating images. Generating videos. But world models raise a fundamentally different question: Can AI generate an environment that people can inhabit, explore, and continuously evolve? If the answer is yes, the impact will extend far beyond video generation. Game development, robotics training, embodied intelligence, digital twins, virtual production, and many other fields could be transformed by the development of world models. World models are still at a very early stage. Yet from Craik’s proposal of an internal mental model more than eighty years ago to the emergence of today’s interactive world-generation systems, a clear evolutionary path is beginning to take shape. Perhaps what AI is ultimately learning has never been limited to images, videos, or language. Perhaps it is learning the world itself. References GitHub: Technical Report:show more

Big win for open-source LLMs! DeepSeek V4 Pro holds... the top open-weights score on SWE-bench Verified, in the GPT-5.5 range. GLM 5.2 leads the open-weight intelligence index and sits near the closed frontier on long-horizon coding. But this leaderboard number is a weak proxy for real performance. It comes from one task set, run through one harness, served at one precision. The same weights can even score differently across providers, since many hosts quantize activations to fp8 and drift the model off its reference weights. Real performance is determined based on whether a model can read a repo, make coordinated edits across files, run the tests, and recover when one breaks. By that measure, the top open models hold up, but only inside the right harness. The teams that actually put DeepSeek V4 into production pipelines as a frontier substitute got there through the harness they built around the model, not by picking a stronger model. If you want to see this in practice, Cline (64k+ stars) has actually built that harness around open models, tuned so they run at production quality. And it's tuned so that these LLMs can run at production quality, with plan and act modes, checkpoints, and terminal feedback. ClinePass is the new access layer on top of it. It runs a curated set of those models inside Cline, narrowed to the ones tested for coding-agent use, with 2 to 5x the standard rate limits and no separate provider accounts, keys, or billing to track. The video below shows the setup, and I worked with the team to put this together. It runs alongside custom keys and local models as well, not in place of them.show more

50% cheaper Claude inference with just one line of... code change! - Remove → model="claude-opus-4-8" - Add → model="ship-like/claude-opus-4-8" I verified the cost saving in my own terminal by invoking the same Anthropic model with the same prompt. The underlying engineering by Ship is actually interesting, and the patterns can be used in any production LLM stack. Essentially, a trained model is a frozen artifact. Every request performs the same forward-pass, whether it extracts a date or refactors a module, because the compute decision was made at training time, before the request existed. Ship makes that decision at inference time instead. After seeing a request, it searches over executions, involving single models, cascades, ensembles, or harnesses with tools, and serves the cheapest one that will match the reference model's quality. This is not a basic router, because picking a cheaper model per query doesn't ensure the cheaper model preserves the original's behavior, like output shape, tool-call patterns, and refusals. Ship measures this equivalence directly. Outputs stay distributionally indistinguishable from the reference model, not token-identical, since two calls to the same model already differ, but they are indistinguishable in capability and behavior. Of course, some requests execute cheaply and some cost Ship more than the customer pays, but the price per request is still a flat 50% off either way, so the execution-cost variance moves off the application's bill entirely. The video below depicts the cost savings and output in my real invocation, and I partnered with the team to put this together.show more

🚨 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.show more

MiniMax M3 just dropped — their first natively multimodal... model. So I ran it through my form-filling test. (The model has to place each element at the right pixel position on a blank form image, not type into a field.) Verdict: it got everything on the paper. > Name, DOB, ID, gender, marital status, nationality, email, phone, address, postal code, all there. > Best character spacing I've seen yet: it actually calculates the gap between each character, clean across the DOB and number boxes > A few fields slightly misaligned, but every piece of data made it onto the form The reasoning chain is the interesting part: it does the easy fields first, then works into the tight one-char-per-box fields, reasoning through y-coordinates, baselines, and label clearance in obsessive detail. The cost: 40:33 and 126.7k output tokens. That's a long think — but it's MiniMax's first multimodal model, and it nailed the content.show more

✨ Grok Imagine Video is now live on Photo... AI It's hard to explain how impressive this is because of the speed that xAI got itself from literally nothing to the top of the leaderboards Six months ago Grok's video model was a joke, it wasn't even close to any of the video models out there, it looked cartoony and wasn't there and nobody took it seriously Now it's here and it's instantly the #1 video model out there now, it shot above Kling (which I used before on Photo AI and usually my favorite) and above Runway Gen 4.5 which was just launched 6 days ago! Mmore importantly it's now above xAI's biggest competitors' models: Google's Veo 3 and OpenAI's Sora 2 Being the best video model doesn't mean it's flawless: video is incredibly hard and actually because it looks so realistic now when it does make a mistakes it's even funnier One thing I noticed is that it still has a hard time with is voice, it does it well for a majority of the video but then slips up and produces unintelligible blabbering (which is really funny to hear) in both English (video 1: "it's where I find my naim", what's a "naim"?), and tested it in Portuguese too (video 3 at the end is unintelligible Portuguese I believe) In many ways Grok Imagine Video also reminds me of Sora, it has that weird but funny Sora conversation style But guys it's REALLY really really really close to getting perfect, we're so close to having full video productions being to be able done in AI, actually you already can if you just cut out the bad parts already Very exciting and I'm grateful I can experience thisshow more

this is the new benchmark for models. 10 iterations... on one generation is where Flux 3 Image starts to add a bunch of noise. I didn't understand at first why the image on the left was drifting so much, whereas on the right it wasn't. After doing a bit of research and reading up on the BFL docs, it turns out it's because when asking it to change hair, it's very hard to apply a stencil to the hair, so it regenerates the entire thing. Whereas when you're asking it to swap a T-shirt, a T-shirt is very easy to trace the outline for, which is basically in-painting. If your object is obvious precise region editing with Flux 3 Image is incredible. If it's not obvious the model is less good. but again you rarely need to go over 5+ iterations. Yesterday I was doing tests with Ideogram 4.5 and I thought it was amazing but I haven't put it through a stress test where the change I'm asking for isn't as easy to trace. I guess now that all image models are getting to the same quality of output, the new benchmark will be how good the vision model is at tracing and inpainting the edits we ask for. I've been doing a bunch of generating and Flux 3 still remains my favorite model for artistic and stylistic choices. It's available now in ElevenCreative.show more

A company built a model good enough to pass... as a person. 48% of people who talked to Griffin live (a human interaction model from Tavus) thought it was a real human, I was lucky to get an early access and can definitely agree to that. On NVIDIA's scoreboard for face-to-face AI, a real human scores 3.92 out of 5. Griffin scores 3.83. The next best system scores 2.80. Earlier systems stitched separate models together to hear, think, speak and animate, and every handoff added delay. Griffin does all of it in one system, reading your face and tone as well as your words and reacting while you're still talking.show more