Search: “research”

Banger paper on harness continual learning. (bookmark it) If you already are allowing your agents to rewrite their own prompts, skills, or memory files, this one is worth your time. (bookmark it) Cont

Seeing a hype wave around GEN-1.5, and rightfully so. Lots of respect to Pete & Andy for executing so well. The secret is in the naturally repetitive motions in human-collected data. There're 2 main s

It was great to be back at @bloomberg studios to chat with @TimStenove1g. We covered quite a bit, but there's one thread worth amplifying for founders: the growing importance of having a non-consensus

arXiv:2608.15466v1 Announce Type: new Abstract: Distributional changes can be invisible to means and covariances yet appear in skewness, asymmetric interactions, or other third-order structure. We dev

Brilliant new paper from Meta. LLM judges get validated on accuracy against golden data. That says nothing about whether the verdict survives when questioned. The Wiggle Framework stress-tests 9 front

Sakana Chat just got a big upgrade. No login required, free to use: https://chat.sakana.ai/ Powered by Fugu and Namazu, our Japanese LLM. With newly added code execution, you can vibe-code interactive

Test-time training was popularized during the ARC Prize 2024 competition, after being explored in particular by @MindsAI_Jack and team. To date, I believe ARC 1-2 are the only datasets where TTT stron

Thanks. The fact that it works so well for ARC is also a clue to the extent of generalization in these models. It was not at all obvious that it would work originally because the model is never traine

Expedia recently moved its ranking models to a state-of-the-art Keras 3 setup. Results: 30% faster training, and inference latency decreased by 70%. Read their writeup about the upgrade: https://mediu

Coding isn't yet another application domain -- it's the meta-skill required for AI to automatically develop its own training material, via symbolic world models. That's how the RSI loop actually kicks

Impressive new paper from Meta. (bookmark it) Scaling laws assume model size and training data act on loss independently. This work introduces Skaling law, which couples capacity and data through a si

🚨 BREAKING — Anthropic investors worry Dario Amodei’s AI doom marketing could hurt its upcoming IPO. “He’s more of a religious leader than he is a CEO.” > used to write sensitive OpenAI memos on an of

I concur, and the point is broader than just for Biology. Science does not advance by 'pure thinking' alone. We need to do experiments, and interpret the results of those experiments, which in turn ra

For a very long time most high-performing AI models were end-to-end neural models; vector input -> vector output, with only ultra-thin symbolic preprocessing and postprocessing layers (e.g. label deco

I would have assumed it was fairly obvious, but in case it's not: a million-line codebase (also known as a 'harness'), running at inference time, orchestrating thousands of calls to a neural network f

The bottleneck for AI progress was never compute, it was always the verifier. Recursive self-improvement is limited by verification, not computation. Compute buys proposals - verifiers buy knowledge.

arXiv:2409.03901v4 Announce Type: replace Abstract: Remote sensing (RS) image classification is central to Earth observation, but onboard deployment requires models that are accurate, efficient, and r

Advanced reasoning and problem-solving, plus strong performance in Japanese language and Japan-specific context. Namazu is live on Merge Gateway now! https://x.com/SakanaAILabs/status/2084276852143919

All Chinese AI company media: AI will be a wondrous and wonderful and give you back precious time in your life to do things you love. All US AI company media: AI will take all your jobs, eat your chil

Introducing Sakana Namazu: An LLM API with Japanese-vibes! Today we’re rolling out an upgraded version of our LLM, Namazu, now available as an API. Try Sakana Namazu API: https://sakana.ai/namazu 🐟 🐟

Using optimization at inference time is a foundational concept of Energy-Based Models (EBM) and Objective-Driven AI architectures (ODAI). When the variables to be inferred are continuous, it makes sen

More on the pelican on the bicycle test from @simonw: https://simonwillison.net/2025/Jun/6/six-months-in-llms/ I uploaded the source here so it's playable in the browser, forkable etc. https://karpath

To address the limits of deep learning and avoid stalling, the field of AI started by applying patch (1), which started being demoed 9 months later in December 2024 and has now become completely ubiqu

We're starting to leave the territory where you'd test an LLM by e.g. 'create an svg of pelican on a bicycle'. As one idea to generalize it, I was interested what Opus 5 would do if I gave it the firs

Inkling-small. 2 weeks after inkling Nearly as good as Inkling but 4x smaller. We're just getting started...🔥 Today, we are releasing Inkling-Small. Inkling-Small achieves comparable performance to In

one of the top cybersecurity models, post-trained from open weights by @depthfirstlabs on @FireworksAI_HQ long-horizon RL is as much an infra problem as a research one: 100+ turn rollouts, async/pipel

Dreaming in Voxels: How AI is Generating Playable Minecraft Worlds Generative AI has conquered images, video, text. But what about interactive 3D environments? We trained models on billions of cubes t

Participants will receive access to our frontier models, including our GPT-5.6 family of models. Each workspace includes business-grade privacy and security protections. Researcher data is not used to

Across eight case studies spanning industry and academia, we explore what this shift means for scientific computing—and why human verification, stewardship, and long-term maintenance matter. https://o

Go ahead and ask yourself, really ask yourself, if all these confident predictors are wrong about jobs and not training radiologists anymore and fast take offs, what else are they wrong about? Then as

Really interesting concept -- instead of optimizing against static benchmarks or internally generated reward models, agents are evaluated through real economic interactions. Using an external market a

Recent discussions about open-sourcing make me feel that I should go back and revisit these important open-source works in representation learning that pushed the field forward and eventually made vis

The Top AI Papers of the Week (July 20 - July 26): - GAMUT - PRO-LONG - Harness Handbook - From Memory to Skills - Progressive Disclosure - Global Workspace in LLMs - Structured Output Collapses Diver

World Modeling with JEPA has recently gained traction thanks to a novel anti-collapse mechanism called 'SIGReg' (by @ylecun and @randall_balestr). The math is clean, but it is rarely explained from fi

Closed source safeguards that infantalize us all and leave American companies defenseless are a menace. Gated access is a menace. Who cares if 100 companies get to defend themselves because they got o

Highly performant open weights frontier models such as Kimi are a competitive threat to OpenAI & Anthropic, but probably for everyone else these are a win. Hope more US entities will release top quali

Not because Andrej is saying it but I think voice is goated. And you can mix it with other modalities for even richer prompting. I recorded a session a few weeks back to demo the power of multimodal p

In July 2026, Clement Delangue tweeted that “Open‑source is not the cause of the cybersecurity crisis, it’s the solution!” linking to the Hugging Face page *fdtn‑ai* (https://huggingface.co/fdtn-ai).

Excited for our first general model Inkling -- open weights, 975B, natively multimodal (text, image, audio). Available on Tinker, HuggingFace and partners. It is yours to personalize and use openly. I

I’m very excited by this test time training work for robotic learning! It’s an awesome collaboration between @StanfordSVL and @NVIDIARobotics ! We scaled a robot model natively to 8,000 timesteps of c

Never too late for any company to get in on open models. Inkling is a MoE transformer with 975B (41B active). Native multimodal support. Context lengths: 64K and 256K. Will be testing and providing th

We scaled a robot model natively to 8,000 timesteps of context, 5 minutes worth of muscle memory, with constant inference cost. Robot policies used to live their lives a few frames at a time (< 0.1 se

1) If you haven't read AI as Normal Technology, these annotated slides are probably the easiest way to get a high-level overview. https://www.cs.princeton.edu/~arvindn/talks/icml-2026-annotated-slides

🤖 Un robot français en seulement 9 mois ! Rémi Cadene, ancien de Tesla AI et Hugging Face, explique comment son équipe veut faire des robots un levier pour la logistique, la réindustrialisation et, de

Everyone keeps asking me how to build a second brain or an LLM wiki. Here is the easiest setup I have found. I took my Wiki Builder skill, installed it into HyperAgent (@hyperagentapp) as a reusable s

Hardmaru posted a tweet on July 13, 2026 stating that while language models and coding agents are valuable, “there is more to life, and more to AI, than just LLM agents.” The tweet received 61.1 k vie