Recent posts
I’ve been thinking about how agents can learn inside world models for years. We decided to scale up our RSI Lab to bridge recursive self-improvement with physical AI and robotics. We are looking for frontier researchers and engineers to join us in Tokyo: https://sakana.ai/careers/member-of-technical-staff-rsi-lab/
Huge congratulations to @JeffDean and the legendary founding team on the launch! 🚀 I share a deep conviction in this mission. Automating the scientific method will profoundly alter the trajectory of AI over the next few years. Bringing the AI Scientist into a loop of recursive self-improvement will fundamentally change the landscape of our field. I strongly believe this automated experimental approach is the next major paradigm shift in AI beyond building large foundation models. It is a true honor to be included on this slide alongside such an incredible group of alumni, and I am excited to see what you will build next.
Our general approach is to automate the experimental loop. We think this approach is broadly applicable across many different fields of science and engineering. We’ll initially focus on ML research and engineering, but believe the approach can help with important subproblems in nearly every one of the fourteen <at>NAE Grand Challenge problems. We think doing this well requires strong expertise in machine learning as well as large-scale systems.
After rigorous testing, our joint AI project with Daiwa Securities is entering the full-scale production phase. We're bringing our agentic AI systems to @Daiwa_JP’s wealth management teams to accelerate complex market analysis in volatile markets. Big milestone for Sakana AI!
大和証券との共同AIプロジェクトが本格開発フェーズへ移行します。 https://t.co/O9lCoa6Idc マーケット情報の収集・分析に関する技術検証を通じて有用性を確認できたため、ウェルスマネジメント業務支援AIの本番開発を開始します。 Sakana AIのAIエージェント技術を活用し、お客さまと向き合う時間 の創出とコンサルティング品質のさらなる向上を支援していきます。
Sakana Namazu: An LLM API with Japanese-vibes! 🎏 Built for Japanese enterprises, featuring frontier-level reasoning and built-in agentic tools. 開発者の皆様、大変お待たせしました！Sakana Chatのモデルが遂にAPIとして公開です。ぜひお試しください！ Blog: https://sakana.ai/namazu-api#English
🐟 Sakana Namazu API 公開 🐟 本日、Sakana AIは大規模言語モデル「Namazu」をアップデートし、API「Sakana Namazu（サカナ・ナマズ）」として提供を開始しました。 Sakana Namazu API: https://sakana.ai/namazu 🐟
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 to generate fully playable, structured worlds. Blog: https://pub.sakana.ai/dream-cubed 🧱
We are excited to share our latest work, together with @nyuniversity: "Dream-Cubed: Controllable Generative Modeling in Minecraft by Training on Billions of Cubes." Blog: https://t.co/dASGUTAPIY Paper: https://t.co/EQbWkWgztP Code: https://t.co/QQBh2HmF6i Generative AI has made incredible progress in language modeling, far beyond other modalities, where words and tokens offer a natural compositional unit for scalable training. This is similar to Minecraft and many other popular video games, where developers rely on cubes, tiles, and other discrete primitives to build rich, interactive worlds. In this work, we show that using cubes as tokens allows large transformers to do the same. Our contribution is two-fold: 1/ We release Dream-Cubed to the research community, a large-scale dataset of Minecraft worlds designed for generative modeling. Our data comprises tens of billions of carefully-balanced cubes from procedurally generated Minecraft terrain and high-quality human-authored maps (obtained with the authors' consent). 2/ We use our data to train a family of powerful transformers for efficient generation of interactive 3D environments at cube resolution. We show how our models allow players to mold the world around them by generating structures, terrain, and maps that are immediately editable and playable. Using high-quality data, we demonstrate that these models can be successfully trained with different training objectives, including both continuous and discrete diffusion, unlocking targeted inpainting, large-scale outpainting, and user-conditioned generation of infinitely sized worlds with fine-grained block-level control.
Running K3 locally on m5 max at 0.3 tok/s This is probably as slow as it will ever be😅
That's a bit slow. Streaming directly the K3 official hugging face 1.6TB of weights in mxfp4 in an m5 max 128gb.
Open ecosystems are the foundation of a healthy AI industry. We have always believed that collective intelligence is the future. Proud that @SakanaAILabs is standing alongside global tech leaders to sign the open-weights letter. 🐟
創業以来、オープンソースコミュニティから多くを学び、また研究成果の公開を通じてそこに貢献してきました。オープンなエコシステムが健全なAI産業と技術主権を支える重要な基盤の一つであると考えており、その発展を支持します。 このたび、Sakana AIは、オープンウェイトAIモデルに関する公開書簡 「Open Weights and American AI Leadership」に署名しました。 書簡は、オープンウェイトモデルがAIへのアクセス拡大、競争の促進、ユーザーによるコントロール、そして安全性の向上に寄与することを訴えるものです。 PDF：https://t.co/Etp6p12EFf 全文：https://t.co/QVU0RK5giu 🐟
Fugu-Ultra now works with Claude Code 🐡
Announcing the Claude Code-compatible interface for our new Fugu-Ultra v1.1! 🐡 Put a dynamically coordinated team of frontier models to work inside the coding workflow you already know. Instead of relying on a single model to write, debug, and execute your code, you can now orchestrate a diverse pool of state-of-the-art models directly from your terminal. Put the whole school to work on your next task: https://t.co/B3LTWK4IEc 🐟
今夜はデニーズのドリンクバーでオープンモデルの未来について語り合うしかない ☕️
Our team just shipped Fugu-Ultra v1.1! 🐡 By dynamically orchestrating the latest frontier models, we pushed performance up by 7.9 points. We are now beating Fable 5 in complex coding and reasoning tasks without even having Fable 5 in our agent pool. Collective intelligence is the future.
Announcing Fugu-Ultra v1.1 🐡 We’ve been thrilled by the reception to the Fugu model family. Thanks to everyone who tried it, shared feedback, and trusted Fugu with real work. Today, we’re releasing Fugu-Ultra v1.1 → https://t.co/hhO6qTawgb Upgraded to incorporate the latest frontier models, resulting in stronger performance across every benchmark shown, including gains of up to 7.9 points over v1.0, with particularly strong results on ProgramBench and Terminal Bench 2.1. Fugu-Ultra v1.1 is more capable across coding, agentic tasks, and advanced reasoning, and available at the same price as Fugu-Ultra v1.0 The frontier keeps moving, and Fugu keeps getting better.
Incredibly proud of the Sakana AI team. We have developed an orchestration model right here out of Japan that achieves state-of-the-art performance on real-world cybersecurity benchmarks! 🎌
Introducing Fugu-Cyber: an update to our Fugu orchestration model. It achieves state-of-the-art performance on real-world security benchmarks, matching cyber-focused frontier models like GPT-5.5-Cyber and Mythos Preview. https://sakana.ai/fugu-cyber-release 🐡
Real brains follow Dale's principle: a neuron can either excite its neighbors or suppress them, but never both. Standard deep learning ignores this and uses backpropagation. In our new paper, Diffusing Blame, we fix this disconnect. By introducing a routing method that broadcasts error signals directly to the hidden layers, we can train networks made of dedicated positive and negative neurons to strictly obey Dale's principle, all without relying on backprop! This method works surprisingly well on image recognition tasks despite the strict biological constraints. We also achieved competitive, backprop-free reinforcement learning on complex locomotion tasks and the open-ended Craftax environment. It is neat to see that representation learning remains possible even when we force deep learning to play by the rules of real neurons.
Introducing "Diffusing Blame": can a neural network learn competitively while strictly obeying Dale's principle, the rule that real neurons follow? We show it can, across both image classification and reinforcement learning. 🧠 Accepted at #ALIFE2026 https://t.co/oSsfRCpvc7 Real neurons generally follow Dale’s principle: each neuron is predominantly excitatory or inhibitory. Standard artificial networks usually ignore this constraint, allowing every unit to mix positive and negative outgoing weights. Backprop makes the gap even wider. Its backward pass needs exact transposed copies of the forward weights, the so-called "weight transport problem,” which biology doesn’t seem to have a mechanism for. So we asked: can a network that strictly enforces Dale's principle still learn well, without weight transport? Our approach builds on Error Diffusion (ED), a local rule that routes a single global error signal directly to every hidden unit, where each layer is split into separate excitatory and inhibitory streams with four non-negative weight matrices, so a synapse's sign comes from fixed population identity rather than a learnable weight. Our main contribution is to extend ED from binary to multi-class problems via modulo error routing. We then asked whether this routing mechanism could provide useful credit signals in the noisy setting of reinforcement learning. During PPO training on Ant, Humanoid, and HalfCheetah, we compared each local ED update with the corresponding true backpropagation gradient. Among the routing schemes we tested, modulo routing consistently produced the strongest alignment. Taken together, these results show that Dale-constrained networks can still learn without transporting weights backward, suggesting a potential path toward learning rules that are both effective and more biologically plausible.
We’re excited to collaborate with NVIDIA to build the next generation of Fugu orchestration models together, by incorporating leading open-weights models.
Sakana AI Teams With NVIDIA to Advance Open Model Innovation from Japan We're announcing the next phase of our collaboration with NVIDIA. We're bringing NVIDIA's open model stack, including the Nemotron family, into Sakana Fugu, our multi-agent orchestration system. https://t.co/bsxjRkux0a Rather than relying solely on scaling individual monolithic models, our approach focuses on collective intelligence. Sakana Fugu operates as an intelligent orchestrator behind a single API, dynamically selecting, coordinating, and combining the strengths of multiple models for each task. This architecture keeps our system modular, adaptable, and resilient. As a natural next step to expand Fugu's capabilities, we're integrating NVIDIA Nemotron as a specialized agent, complementing the frontier and open models Fugu already orchestrates. Nemotron helps demonstrate how open models become far more useful when orchestrated within agentic systems rather than used in isolation. This collaboration creates a reinforcing cycle. Fugu gains a deeper pool of specialized capabilities, while NVIDIA can evaluate how its models perform when coordinated within complex, multi-step workflows. These real-world signals can continuously improve both the models and the orchestration layer. By combining Sakana AI's Japan-born collective-intelligence approach with NVIDIA's open models and accelerated computing, we aim to shape a future of AI that is modular, collaborative, and open by design.
Language models and coding agents are great, but there is more to life, and more to AI, than just LLM agents.
How do physical systems achieve collective intelligence and self-repair without a central brain? A new paper published today in Nature Communications by my Sakana AI colleague Sebastian Risi (@risi1979), along with co-authors from IT University of Copenhagen and Autodesk Research, presents a beautiful realization of biologically inspired robotics: Smart Cellular Bricks. The team built a system of physical 3D cubic units that can collectively infer their global shape and autonomously guide their own damage recovery using purely local interactions. Here is a deep dive into the paper’s key contributions: 1/ Neural Cellular Automata-based Architecture: Modular robots usually rely on central processors. This system flips that paradigm. Every block independently runs the exact same neural network on local microcontrollers. With no master plan or global coordinates, they communicate only with immediate neighbors. By passing continuous state vectors, hundreds of bricks achieve global consensus on their shape in under 3 minutes. 2/ Emergent Biological Morphogens: How does a block know it is part of a chair, not a table? The network’s internal memory automatically learns to establish continuous gradients across the structure. This beautifully mirrors how biological morphogens give positional info to developing cells. The bricks naturally form left-right, radial, and head-to-tail axes to align their identity. 3/ Performance and Generalization: Validated in large-scale simulations, the networks transferred seamlessly to nearly 200 physical hardware bricks, achieving a 100% convergence rate. Instead of rigid template-matching, the system infers broad categories. Even when tested on unseen variations, like an asymmetric table with five random legs, the collective correctly classified the structure. 4/ Fault Tolerance and Autonomous Damage Recovery: Hardware fails in the real world. This system easily tolerates up to 15% module failure without losing accuracy. By predicting spatial damage directions, the cells pinpointed missing components with 95% accuracy. They actively use these local signals to guide a self-repair process, regenerating back into the intended morphology. I believe this is a significant piece of research, bridging collective intelligence and Physical AI. This work represents the first successful physical realization of large-scale, decentralized 3D self-recognition and damage detection. By moving away from centralized control, this architecture paves the way for highly adaptive smart materials and resilient robotics that can survive and repair themselves. Read the full open-access paper: https://t.co/friVEdvfT6 Congratulations to the team on this achievement!
We are pleased to share our latest research, now published in Nature Communications: “Smart Cellular Bricks: Physical Modules That Recognize Their Own Shape and Repair Themselves.” Blog: https://t.co/FMkbCv8mpP Paper: https://t.co/rxzbpU3zTu A long-running theme in our work is collective intelligence: the idea that sophisticated, robust behavior can emerge from many simple parts following local rules, with no central controller, as it does in a colony, a tissue, or a brain. We had mostly studied this in software and simulation. So this time we asked a simple question. Do the same decentralized principles hold up in the physical world, where communication is noisy and modules fail? To find out, we built a collection of simple cubic bricks. Each brick runs the same small neural network and talks only to the bricks it is physically connected to. No brick is told its position, or which shape it is part of. Yet from these purely local exchanges, the collective converges on the correct global shape, locates where modules are missing or damaged, and can even guide its own repair, inspired by how living tissue self-organizes and regenerates after injury. For us, this is a first step in a broader direction: taking the principles of collective intelligence we have studied in software and letting them emerge, decentralized and robust, in the physical world. In the future, we imagine smart materials that let structures sense and report damage on their own, and LEGO-like systems that recognize their own configuration and adapt in real time, pointing toward environments that are more robust, adaptive, and regenerative. This work is a collaboration between Sakana AI, IT University of Copenhagen and Autodesk.
One of my first journeys in neural networks started over a decade ago with implementing CPPN-NEAT! Back then, I built a clone of ‘Picbreeder’ not only to study the mechanics of neural nets, but to explore the human creativity process itself, and generate some cool abstract art. Neurogram: https://t.co/ytwBChaHu7 Gallery: https://t.co/aAaUGomkv2 Today, things have come full circle. We are now trying to use modern VLMs and frontier LLM agents within open-ended exploration algorithms. We want to see if we can finally computationally derive the underlying mechanics of human creativity: serendipity, memory, exploration versus exploitation, and novelty search. Can modern AI actually replicate the magic of human open-endedness? Dive into our new AI Picbreeder Experiment here: https://t.co/hrLhTsIMfe
The AI Picbreeder Experiment: Can AI agents be creative when nobody tells them what to create? Blog: https://t.co/qsMwcB3N5D In our new #GECCO2026 paper, "In Search of the Ingredients of Open-Endedness: Replicating Picbreeder with Large Vision-Language Models", in collaboration with MIT and NYU, we revisit Picbreeder, a lost website where people collaboratively evolved images without any predefined objective. Users simply selected images they found interesting, allowing unexpected forms such as faces, animals, vehicles, and skulls to emerge gradually across many generations and many different people. We recreated this process using vision-language model agents. The agents explore a shared archive, choose images to branch from, evolve new candidates, publish their favorites, and evaluate the creations of other agents. There is no target image and no explicit definition of what counts as progress. The results reveal both the promise and current limitations of AI-driven open-ended discovery. Compared with humans, VLM agents tend to keep circling back to the same kinds of images and concepts. They repeatedly select similar parents, make smaller conceptual leaps, and often refine an existing idea rather than abandoning it in search of something genuinely unexpected. However, introducing a diverse population of agent personalities substantially improves exploration. In some runs, diverse agent populations approached or matched the human archive on measures of semantic diversity and produced more balanced evolutionary trees. We also find intriguing evidence that open-ended evolution can produce more robust representations. A skull evolved by the agents changes smoothly when its underlying neural representation is perturbed, less fractured than a skull directly optimized with gradient descent, although still less cleanly disentangled than one evolved collectively by humans. But perhaps the most interesting result is the gap that remains. Humans appear better at turning fortunate accidents into sustained creative discoveries: recognizing when something unexpected is worth pursuing, refining it, and then making a larger conceptual leap. The AI agents often notice interesting patterns too, but are more likely to become trapped in them. We still do not fully understand what enables humans to navigate open-ended search in this way, or what ingredient(s) current AI systems are missing. For now, the results suggest that there remains something important about human creativity that AI agents have not yet learned to reproduce. This paper will be presented at #GECCO2026 and is nominated for a best paper award! Please check out the interactive blog and technical paper for more details! Read our full paper: https://t.co/QnxVWLzjez 🐟
We just launched Sakana Translate! https://t.co/KeleCLOLOT I personally rely on this tool every day. Huge congratulations to the team for shipping this! Standard translation tools often miss the deep nuance of Japanese business honorifics, cultural concepts, and internet slang. We built a tool that actually translates the context and tone.
🐟️ Sakana Translate公開 🐟️ 本日、Sakana AIはチャットサービス「Sakana Chat」に新機能「Sakana Translate」を追加しました。 日本語・英語・中国語の双方向翻訳に対応します。 Sakana Translateを試す：https://translate.sakana.ai
I’m looking to hire a Program Manager to help manage Sakana AI’s fast growing Recursive Self-Improvement (RSI) Lab 🚀 RSI Lab (English): https://sakana.ai/rsi-lab/ RSI Lab (日本語): https://sakana.ai/rsi-lab-jp/ Job Description: https://sakana.ai/careers/program-manager-rsi-lab/
【採用情報】プログラムマネージャー(RSI Lab)のポジションをオープンしました🚀 https://t.co/57Eyb23kws RSI Labの研究活動を支えるプログラムマネージャーを募集します。トップクラスの研究者・エンジニアが研究に専念できる環境をつくる役割です。 このような役割を担っていただきます。 ・予算管理・スケジュール管理を含む研究オペレーション全般 ・リサーチャーと技術的な会話をしながら、計画と実態のギャップ調整 ・社外パートナーとの窓口としてのコミュニケーション 予算管理やプロジェクトマネジメント、対外折衝などの実務経験があり、ビジネスレベル以上の英語力をお持ちの方を歓迎します。 研究を支える立場から、AIの次のパラダイムづくりに関わりたい方、ぜひご応募ください🐟
Excited to partner with @OpenRouter ⚡ Products like OpenRouter Fusion and Sakana Fugu have sparked a serious conversation about dependency and resilience in AI. I believe this is just the start of a great architectural shift to come in AI development.
Fugu-Ultra is now live on @OpenRouter! ⚡ We share a core vision with the OpenRouter team: the future of AI isn’t a single monolithic model, but the collective intelligence of the world’s best models working together. Try it: https://openrouter.ai/sakana/fugu-ultra 🐡
