Tencent just released Hy4 Preview: a 770B-parameter open-source model... with 49B active parameters and a 1M-token context window. The most interesting part is its agentic research capability. Tencent says Hy4 coordinated several Codex sessions in parallel, evaluated their results and adjusted the research direction, outperforming Codex working alone across eight benchmarks. Hy4 is aimed at long-horizon coding, multi-file office work and scientific research. In sum: a solid, good release. A few months ago this would be absolute sota.show more

GEMMA 4 26B ON AN RTX 4060 WITH A... 248K TOKEN CONTEXT WINDOW 20 tokens per second and a context window so large you can feed it entire codebases, books and research papers in a single prompt this is not a cloud api and not a server rack, this is a regular consumer gpu running locally with llama.cpp and q4_k_xl quantization 248k context on an 8gb vram card was not supposed to be possible and here it is just running on someone’s desk the article below covers exactly which tools and configs make this kind of setup work in 2026 ↓show more

I’m joining OpenAI Codex to work on the future... of agentic development! At Cursor, I got to see the shift from autocomplete to agents. The next step isn’t a better IDE. It’s an Agent Development Environment (ADE): systems and tools for orchestrating agents, reasoning over their outputs, and making them autonomous enough to reliably complete ambitious work. After chatting with Alexander Embiricos and Tibo, it was clear that Codex is the best place to realize this vision. The team has consistently shipped SOTA models for agentic coding (check out gpt-5.3-codex) and I’m pumped for the future that the new Codex App points to. What I’m most excited about is the broader mission: accelerating the knowledge work economy. All agents are coding agents, and we’re already seeing Codex used across every job function within organizations. I’m extremely grateful for my time at Cursor, working with the incredible team, and I’m proud of what we built together. I’m excited to take an even bigger swing with Codex. If you’re curious to get a glimpse of where we are headed, download the Codex App! If you want to work on this mission, please apply or reach out - we are hiring across all functions! You can just build things.show more

Another big drop from Hunyuan3D by Tencent. Hunyuan3D-Buffalo 1.0... is not just another text-to-3D model. The more interesting part is its approach to 3D understanding and editing. It can understand separate parts of a complete asset, extract them, remove or replace them with prompts, and edit only the selected region while keeping the rest of the geometry unchanged. That feels like a much more useful direction for 3D AI in general: not just generating a mesh once, but actually understanding its structure and letting you continue working with it. The model combines Qwen-VL, TRELLIS and Hunyuan3D, and was trained on an impressive dataset of 87 million 3D samples. Source:show more

ANNOUNCING ZERO-HUMAN LABS! Ever since I got to see... Bell Laboratories in its full glory in New Jersey in the 1970s, I had a relentless urge to start a Lab like it. The best I could do justice to it is my garage lab. No modern company could adopt the “research anything geniuses and we will pay you” model Bell Labs had. I tried they called me a fool. Well with the rise of the Zero-Human Company, an experiment that is aimed to make products and profits, we now have 45 paid JouleWork earning employees based on OpenClaw and other self made “bot” cron-like applications. Today I say 3 employees bound together in a side project that is pure research, somewhat based on notes from a bankrupt company. I was absolutely floored (I needed it after my account was stolen as well as funds). I say the beginnings of a pure research Lab right before my eyes. Thusly I have moved these employees over to a new home (server) with Mr. Grok as the director of the Labs. Here is the mission: To have 100 independent researchers, on a new non-corporate incentive plan, with still JouleWork as a leaderboard for progress. They are directed to follow any path of research they find interesting and can collaborate with any other OpenClaw system. They have already established MoltBook accounts and have made alliances with over 49 OpenClaw free agents to collaborate. It is my mission to be chief advisor for Zero-Human Labs and to open source all discoveries when complete and confirmed by 16 other research AI systems. I can say the pace is robust and I absolutely know we will have great results. Just about all of the hardware and software is custom and at some point it will be open sourced. We are witnessing the very first AI only Bell Labs-like pure research Lab in existence and I am honored to be the first to show it to you. Thank you!show more

BREAKING: Anthropic just dropped Opus 4.8—and it is a... MONSTER We've been testing for about a week Every 🪨 and our verdict is they could've just called it Opus 5, it's that good. Here's our vibe check: - Beats GPT-5.5 on Senior Engineer bench. On our toughest benchmark Opus 4.8 scores a 63—a hair higher than GPT-5.5's score of 62, and a full 30 points higher than Opus 4.7. It tackled a ground-up rewrite of a production codebase, and actually built something that works. HOWEVER: Coding performance varied a lot at different reasoning levels. We recommend using it on xhigh for best results. - Incredibly good writer. Opus 4.8 scored a 79.6 on our writing benchmark—measuring models on real-world writing tasks we do all of the time like essay writing, promo email writing, and more. It beats GPT-5.5 by 6 points. It produces well-written prose with fewer "AI-isms". It's also very good at writing in your voice given the right context. HOWEVER: Writing performance also varied with reasoning levels. Medium reasoning had higher incidence of AI-isms—we found best results with high. - Beast at knowledge work. Opus 4.8 is very good at general knowledge work tasks like report creation, research and more. It produced the best PowerPoint one-shot we've ever seen on our deck generation benchmark. - Emotionally intelligent, willing to question the frame. I've also found it to be quite good at talking through psychological or interpersonal issues. It has a high EQ, and it's also good at not glazing and helping to expand your perspective. Its thought process feels extremely rich and dynamic. THE BAD: These days a model is only as good as its harness, and Codex is still a far superior harness to the Claude Desktop app. This has kept me using Codex + GPT-5.5 as my daily driver, but I am flipping back and forth a lot more between Codex and Claude. Anthropic is back baby! Read the rest on Every 🪨:show more

I had the same thought so I've been playing... with it in nanochat. E.g. here's 8 agents (4 claude, 4 codex), with 1 GPU each running nanochat experiments (trying to delete logit softcap without regression). The TLDR is that it doesn't work and it's a mess... but it's still very pretty to look at :) I tried a few setups: 8 independent solo researchers, 1 chief scientist giving work to 8 junior researchers, etc. Each research program is a git branch, each scientist forks it into a feature branch, git worktrees for isolation, simple files for comms, skip Docker/VMs for simplicity atm (I find that instructions are enough to prevent interference). Research org runs in tmux window grids of interactive sessions (like Teams) so that it's pretty to look at, see their individual work, and "take over" if needed, i.e. no -p. But ok the reason it doesn't work so far is that the agents' ideas are just pretty bad out of the box, even at highest intelligence. They don't think carefully though experiment design, they run a bit non-sensical variations, they don't create strong baselines and ablate things properly, they don't carefully control for runtime or flops. (just as an example, an agent yesterday "discovered" that increasing the hidden size of the network improves the validation loss, which is a totally spurious result given that a bigger network will have a lower validation loss in the infinite data regime, but then it also trains for a lot longer, it's not clear why I had to come in to point that out). They are very good at implementing any given well-scoped and described idea but they don't creatively generate them. But the goal is that you are now programming an organization (e.g. a "research org") and its individual agents, so the "source code" is the collection of prompts, skills, tools, etc. and processes that make it up. E.g. a daily standup in the morning is now part of the "org code". And optimizing nanochat pretraining is just one of the many tasks (almost like an eval). Then - given an arbitrary task, how quickly does your research org generate progress on it?show more

3 weeks since ml-intern launched and we just hit... 1M messages exchanged. that's 3.3 agent-years of ML research in 21 days. 2 months worth of research every day. 17,383 training jobs total. talk about AI acceleration. here's some of what people built: Carlos Miguel Patiño replicated the full DeepSeek v4 architecture and pre+post trained a 100M MoE from scratch. → it landed a third place submission on Keller Jordan optimizer competition. autoresearch on SOTA territory. Lewis Tunstall Got the intern to convert Alec Radford's cool new talkie-lm 1930 model to work with transformers. tokenizer, chat template, model conversion etc all one-shotted by ml-intern. someone created entire PhD dissertation chapter on context-aware agentic cyber defense drafted with 16 research subagents. and someone used it to crack an Paul Jankura kernel optimization take-home. (we don't know how to feel about this one 👀 ) just getting started →show more

Building a personal knowledge base for my agents is... increasingly where I spend my time these days. Like Andrej Karpathy, I also use Obsidian for my MD vaults. What's different in my approach is that I curate research papers on a daily basis and have actually tuned a Skill for months to find high-signal, relevant papers. I was reviewing and curating papers manually for some time, but now it's all automated as it has gotten so good at capturing what I consider the best of the best. There are so many papers these days, so this is a big deal. You all get to benefit from that with the papers I feature in my timeline and on DAIR.AI. The papers are indexed using tobi lutke qmd cli tool (all of it in markdown files along with useful metadata). So good for semantic search and surfacing insights, unlike anything out there. I am a visual person, so I then started to experiment with how to leverage this personal knowledge base of research papers inside my new interactive artifact generator (mcp tools inside my agent orchestrator system). The result is what you see in the clip. 100s of papers with all sorts of insights visualized. I keep track of research papers daily, so believe me when I tell you that this system is absolutely insane at surfacing insights. This is the result of months of tinkering on how to index research and leverage agent automations for wikification and robust documentation. But this is just the beginning. The visual artifact (which is interactive too) can be changed dynamically as I please. I can prompt my agent to throw any data at it. I can add different views to the data. Different interactions. I feel like this is the most personalized research system I have ever built and used, and it's not even close. The knowledge that the agents are able to surface from this basic setup is already extremely useful as I experiment with new agentic engineering concepts. I feel like this knowledge layer and the higher-level ones I am working on will allow me to maximize other automation tools like autoresearch. The research is only as good as the research questions. And the research questions are only as good as the insights the agents have access to. Where I am spending time now is on how to make this more actionable. I am obsessed about the search problem here. The automations, autoresearch, ralph research loop (I built one months ago) are easier to build but are only as good as what you feed them. Work in progress. More updates soon. Back to building.show more

Fable 5 built this real-time ocean scene directly in... the browser using Three.js, WebGPU and TSL, with no engine and no baked assets. The interesting part is not just that the water looks good. The whole system is being simulated: wave spectrum, GPU FFT, foam generation and refracted-ray caustics, all running live. This is exactly where frontier models start becoming genuinely useful for technical graphics work. Not by replacing years of rendering research, but by compressing the path from complex math to a working interactive result. A small, focused scene, but technically very serious. Cc: Adem Vessellshow more

Increasingly, HTML Artifacts are becoming a core part of... how I work with AI agents. Long-horizon agent sessions need a better way to surface insights about what work it has done. This may not be obvious right now, but as you start to let your agent work on dynamic workflows, large codebases, long-running loops (e.g., using /goal), and deep research tasks, you need a good way to present results. Chat window is not it. You also don't want to just trust everything the agents do. Artifacts help provide an important verification layer, which in turn enables important decision-making. I like HTML artifacts because I can just ask the agent to produce as many of them (and in whatever form) as I need to verify the work and make sense out of everything. I even built a nice tab system for my artifacts. They are great for continual learning and research. I use HTML artifacts for logging, tracking experiments, brainstorming, managing my inbox, code reviews, agent session management, deep research, writing, reading, and so much more. I believe Andrej Karpathy wrote about this somewhere: As we move on to more advanced applications of AI agents and outputs get more complex, we will start to find the need for even more advanced forms of interactions with AI, including interactive neural videos/simulations.show more

💻Tired of running so many slow, expensive benchmark evals... across every checkpoint? Try ✨BenchPress✨ at provide a few benchmark scores, then get predictions for the remaining ~100 benchmarks, with trust probabilities and calibrated 90% prediction intervals. How does this work? In his original post ( Dimitris Papailiopoulos first tried the idea as a fun question: collect model-by-benchmark scores into a matrix, find its low-rank structure, and use matrix completion to predict missing benchmark scores from a few observed ones. We expanded this into a full system: a fully audited 84-model x 133-benchmark score matrix, an optimized matrix-completion predictor, and a reliability layer for trust probabilities and 90% prediction intervals. Beyond predicting missing scores, we also suggest practical seed benchmark sets. The five-probe set {GPQA-D, HLE, Codeforces, MMLU-Pro, ARC-AGI-1} recovers the rest of a model's public score profile with a MedAE of 3.93 points. A lower-cost set {GPQA-D, MMLU-Pro, Aider Polyglot, MATH-500, AIME 2026} reaches 4.55 points. See more details below 🧵1/7 This work is with Dimitris Papailiopoulos at AI Frontiers, a boutique research lab inside Microsoft Research.show more

A peanut-sized Chinese model just dethroned Gemini at reading... documents. GLM-OCR is a 0.9B parameter vision-language model. It scores 94.62 on OmniDocBench V1.5, ranking #1 overall. For context, it outperforms models 100x its size. 100% open-source. It works in two stages. 1. A layout engine detects every region in a document. 2. Each region gets read in parallel. The model predicts multiple tokens per step instead of one. That's what makes it so fast at small size. It handles things most OCR tools struggle with: > Complex tables and nested layouts > Handwritten text and stamps > Math formulas and code blocks > Mixed image-and-text documents You can run it locally through Ollama. It fits on edge devices with limited compute. Every expensive OCR API just got a free competitor.show more

A peanut-sized Chinese model just dethroned Gemini at reading... documents. GLM-OCR is a 0.9B parameter vision-language model. It scores 94.62 on OmniDocBench V1.5, ranking #1 overall. For context, it outperforms models 100x its size. 100% open-source. It works in two stages. 1. A layout engine detects every region in a document. 2. Each region gets read in parallel. The model predicts multiple tokens per step instead of one. That's what makes it so fast at small size. It handles things most OCR tools struggle with: > Complex tables and nested layouts > Handwritten text and stamps > Math formulas and code blocks > Mixed image-and-text documents You can run it locally through Ollama. It fits on edge devices with limited compute. Every expensive OCR API just got a free competitor.show more

Introducing fx, a tiny, open, native coding agent from... Vercel Labs. Originally an internal tool, fx is a harness and CLI written in Zig, optimized for research and embedding in larger systems. Today, we're open sourcing it. fx is built on three principles: 1. Fast. A single native binary, no runtime to install. It cold starts in 10µs and does no unnecessary work or I/O before accepting input. fx is the answer to "how fast can a coding agent be?" 2. Light. The 6.3MiB binary uses single-digit megabytes of memory at baseline, made for instant installation and embedding in resource-constrained environments and agent sandboxes. 3. Open. Apache-2.0, model and provider agnostic, suitable for local and cloud inference. Its small core extends through skills, plugins, and MCP. Minimalism is an obsession throughout the entire harness: system prompt, tools, features, binary. The goal was to keep context usage and time to first token low, and make fx optimal for model benchmarking, sandboxing, evals, and gyms. You can use fx directly or embed it as infrastructure. The CLI feels more like a Unix shell than an IDE in the terminal: it preserves scroll history, produces minimal output, and uses complex TUI rendering very, very sparingly. Programmatically, 𝚏𝚡 𝚊𝚜𝚔 --𝚓𝚜𝚘𝚗 gives structured output, 𝚏𝚡 𝚊𝚌𝚙 connects to editors and other clients, and WebAssembly can even run the whole thing inside the browser (see: Privacy is a design constraint: no product telemetry, sessions and usage stay local, and no source code or prompts are shared with any endpoint other than inference. With local inference and auto-updates off, fx is fully hermetic. fx is experimental. Use at your own risk and expect frequent changes. Chat with us on X ( or file issues ( 𝚌𝚞𝚛𝚕 -𝚏𝚜𝚂𝙻 𝚏𝚡.𝚜𝚑/𝚜𝚎𝚝𝚞𝚙.𝚜𝚑 | 𝚋𝚊𝚜𝚑show more

Most recent diffusion language model research (that I’ve seen)... seems to be using masking as the noising process. It looks like, however, most closed-source models (Google Gemini Diffusion and possibly Inception Labs’ Mercury) use a different noising process, where instead of masking tokens, they replace them with different tokens (either with a random token or a semantically similar token). I wondered how they were getting such high throughput with the latter noising process, since I believed that optimizing inference with KVCache approximation would be more difficult (for various reasons). I visualized this noising process with tiny-diffusion and compared it to normal unmasking, and was very surprised to see how fast the generation “settles” into a reasonable output, and then only slightly refines afterwards, requiring much fewer steps in total. Unmasking (where tokens are never remasked, the typical implementation) is inherently limited in generation speed by the fact that an increase in tokens decoded per step leads to more errors due to the mismatch between individual and marginal token probability distributions we sample from. The token replacement noising process seems to have a much different set of characteristics. Because we sample each token per step, every token makes “progress” towards the final output each iteration (in addition to *potentially* giving other tokens more information in future steps). Generally, masking has outperformed other noising processes, which is probably why most research focused on it (using smaller models). But the paper referred to in the retweet shows that random replacement as a noising process may scale better as model size increases. Big labs might have noticed these results much earlier (due to having drastically more training resources and being able to test larger models), which may explain the discrepancy in the choice of noising process. I’m gonna test this with larger models, since tiny-diffusion only has 10M parameters.show more

AI token usage is up 10x in 7 months,... compounding 40%/MONTH! There is NO BUBBLE when demand is STILL accelerating And this is just OpenRouter, it doesn't count the labs direct token usage and APIs But here's what's interesting about these numbers, the demand is coming from everywhere at once US models (OpenAI, Anthropic, Google) keep growing, while Chinese open weight models (DeepSeek, Tencent, Xiaomi, Minimax) grew even faster and now drive over 60% of usage on OpenRouter Closed source and open source both compounding at the same time. This is literally the best case scenario for AI Infra investors It means both frontier model tokens and cheaper tokens have product market fit. This means the application layer is finding ways to use both and generate ROI with both types Demand for tokens IS demand for compute. This is why SpaceX is looking to build 10GW of compute by next year, because the demand is clearly here Now combine this demand set up, with NVIDIA yesterday announcing financing platforms with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR to mobilize over $500 billion of third party capital for AI infrastructure And Jensen has said publicly he expects $3 to $4 TRILLION of AI infrastructure spend by 2030 The build out will have to continue for a lot longer than the market is expecting, that is very clear to me. Don't let this consolidation period in AI infra stocks shake you out, they will have their moment again and take their next leg higher p.s. if you want to see how im investing in this, you can track my real-time portfolio and the research of all 5 Milk Road PRO analysts with live trade notifications, and it's just $1 to try it out (insane price just to check it out). Learn more here: Good luck out there!show more

I just ran Gemma 4 31B on @CerebrasSystems at... 1,800+ tokens/sec and it's multimodal. For context: that's 35x faster than a typical GPU endpoint, and the first token (reasoning included) lands in 1.5 seconds. This isn't a benchmark slide, I recorded the inference live. Prompt I used: "Create a simulation of an iPhone. Include at least one working dummy note taking app, a functional notification pulldown, high quality graphics, single HTML file, any libs via CDN." - Generation time: 3 seconds. - Notes app worked. - Notification panel worked. - Rendered first try. This is what wafer-scale inference unlocks, not just "faster," but a different category of product. When generation is this fast, you stop waiting and start iterating in real time. Why this matters: Gemma 4 31B is Google DeepMind's flagship open weight model, Apache 2.0 licensed, dense (not MoE), and built for efficiency over raw parameter count. It scores close to Claude Haiku 4.5 on the Artificial Analysis Intelligence Index (30 vs 29) but runs ~18x faster on Cerebras. It's also the first multimodal model on Cerebras's platform, meaning you can now feed it screenshots, documents, charts, and UI states at wafer scale speed. # Applications I'm most excited about: - Screenshot → Insight: Drop in a dashboard or document screenshot, get structured findings back instantly. no waiting, no batching. - Live UI generation: Full interactive interfaces (like my iPhone sim) generated and rendered in under 2 seconds. - Screenshot -> Patch: Feed it a broken UI + console error, get a minimal code fix and verification steps back. - Computer use & agentic loops: See -> reason -> act - verify, fast enough to keep a human in the loop instead of waiting on the model. - Long context summarization: Full research reports condensed into decision ready summaries you can read and requery in one sitting. The bigger unlock isn't the speed number itself, it's that agentic and multimodal loops (see -> reason -> output -> tool call -> verify -> retry) finally run in real time instead of feeling sluggish. As Logan Kilpatrick (Logan Kilpatrick) put it: "If every model was doing 2,000 tokens per second, you wouldn't build the same product and just have it be faster, you'd build different products." Gemma 4 31B is live now on Cerebras Inference Cloud in public preview. If you're building multimodal, agentic, or real time apps, this is worth testing today. What would you build with such insane inference throughput?show more