# sakana.ai

> AI-optimized mirror of sakana.ai containing 198 pages totalling 293,006 words of clean markdown content, structured data, and semantic HTML. Original source: https://sakana.ai. Last updated: 2026-07-18T07:28:17.885Z.

Each page is available as HTML (with JSON-LD structured data) and Markdown (text-only, ideal for LLMs and RAG).

## Homepage

- [asal/index.html](/content/asal/index.html) (10,289 words)
- [Sakana Fugu](/content/console/index.html): Sakana Fugu is a model interface for coordinating specialized AI agents through one OpenAI-compatible API. (42 words)
- [Publications — Sakana AI](/content/pub/index.html): Interactive research publications from Sakana AI. (212 words)
- [Sakana AI — Building Frontier AI in Japan](/content/site-root.html): We are creating a new kind of foundation model based on nature-inspired intelligence. (12 words)
- [Sakana Chat](/content/chat/index.html): Sakana Chat - Chat app powered by Sakana AI (18 words)

## Articles & Blog Posts

- [pub/digital-ecosystem/paper/paper-pdf.html](/content/pub/digital-ecosystem/paper/paper-pdf.html) (1,725 words)
- [pub/pdnca/assets/pdf/pdnca-pdf.html](/content/pub/pdnca/assets/pdf/pdnca-pdf.html) (1,004 words)
- [Petri Dish NCA - Artificial Life Laboratory](/content/pub/digital-ecosystem/app/index.html): Watch artificial lifeforms compete and evolve in this Neural Cellular Automata simulation (38 words)
- [Artists — Picbreeder · VLM](/content/pub/picbreeder-vlm/breed/artists-html.html) (920 words)
- [Browse — Picbreeder · VLM](/content/pub/picbreeder-vlm/breed/browse-html.html) (480 words)
- [Technical Editor (Public Sector)](/content/careers/technical-editor-public-sector/index.html): Technical Editor (Public Sector) (74 words)
- [Technical Project Manager](/content/careers/technical-project-manager/index.html): Technical Project Manager (52 words)
- [Business Role (Project Manager), Defense and Intelligence](/content/careers/business-role-project-manager-defense-and-intelligence.html): Business Role (Project Manager), Defense and Intelligence (81 words)
- [Business Role (Project Manager), Manufacturing](/content/careers/business-role-project-manager-manufacturing/index.html): Business Role (Project Manager), Manufacturing (93 words)
- [Software Engineer (Defense)](/content/careers/software-engineer-defense-and-intelligence/index.html): Software Engineer (Defense) (106 words)
- [UI/UX Designer](/content/careers/designer/index.html): UI/UX Designer (84 words)
- [Solution Engineer](/content/careers/solution-engineer/index.html): Solution Engineer (82 words)
- [Business Role (Project Manager), Finance](/content/careers/business-role-project-manager-finance/index.html): Business Role (Project Manager), Finance (354 words)
- [Applied Engineer (Enterprise / Defense)](/content/careers/applied-engineer-enterprise-defense/index.html): Applied Engineer (Enterprise / Defense) (113 words)
- [Software Engineer (Finance)](/content/careers/software-engineer-finance/index.html): Software Engineer (Finance) (89 words)
- [Chrome Owl — Picbreeder · VLM](/content/pub/picbreeder-vlm/breed/detail-html.html) (164 words)
- [Experiments — Picbreeder · VLM](/content/pub/picbreeder-vlm/breed/runs-html.html) (2,166 words)
- [Privacy Policy | Sakana Marlin](/content/marlin/privacy-policy/index.html) (2,372 words)
- [CoffeeBench: Benchmarking Long-Horizon LLM Agents in Heterogeneous Multi-Agent Economies](/content/pub/coffeebench/index.html): CoffeeBench: Benchmarking Long-Horizon LLM Agents in Heterogeneous Multi-Agent Economies (2,295 words)
- [Sudoku-Bench](/content/pub/sudoku-gpt5/index.html) (2,054 words)
- [Digital Ecosystems](/content/pub/digital-ecosystem/index.html): An interactive browser-based platform for exploring multi-agent neural cellular automata ecosystems in real time. Watch digital species compete, cooperate, and evolve — then reach into the petri dish yourself. (2,019 words)
- [Extending the Context of Pretrained LLMs by Dropping their Positional Embeddings](/content/pub/drope/index.html): Introducing DroPE: a simple method for extending a pretrained language model's usable context without long-context fine-tuning. (1,584 words)
- [Petri Dish Neural Cellular Automata](/content/pub/pdnca/index.html): Petri Dish Neural Cellular Automata (PD-NCA) is a new ALife simulation substrate that replaces the fixed, non-adaptive morphogenesis of conventional NCA—where model parameters remain constant during development—with multi-agent open-ended growth, trained via continual backpropagation throughout the entire simulation. (1,351 words)
- [assets/marlin/strait-20of-20hormuz-20blockade_-20resolution-20scenarios-pdf.html](/content/assets/marlin/strait-20of-20hormuz-20blockade_-20resolution-20scenarios-pdf.html) (22,180 words)
- [Butterfly — DNA — Picbreeder · VLM](/content/pub/picbreeder-vlm/breed/dna-html.html) (1,067 words)
- [Example Question on Technical Problem Set](/content/assets/careers/example_problem-pdf.html) (275 words)
- [An Unofficial Guide to Prepare for a Research Position Application | Sakana AI](/content/pub/unofficial_guide/index.html): An unofficial guide by Sakana AI researchers on preparing for research position applications and interviews. (1,254 words)
- [assets/marlin/the-20global-20ai-20regulation-20patchwork-pdf.html](/content/assets/marlin/the-20global-20ai-20regulation-20patchwork-pdf.html) (29,405 words)
- [marlin/signup/index.html](/content/marlin/signup/index.html) (50 words)
- [Picbreeder · VLM — collaborative art evolved by AI](/content/pub/picbreeder-vlm/breed/index.html) (177 words)
- [assets/marlin/the-20return-20of-20the-20bond-20vigilantes-pdf.html](/content/assets/marlin/the-20return-20of-20the-20bond-20vigilantes-pdf.html) (27,377 words)
- [String Seed of Thought: Prompting LLMs for Distribution-Faithful and Diverse Generation](/content/pub/ssot/index.html): SSoT: A simple prompting method that enables LLMs to generate distribution-faithful and diverse outputs by first generating a random string. (435 words)
- [Breeding Console — Picbreeder · VLM](/content/pub/picbreeder-vlm/breed/play-html.html) (237 words)
- [System Overview — Picbreeder-VLM](/content/pub/picbreeder-vlm/system_overview/index-html.html) (37 words)
- [CoffeeBench Trajectories](/content/pub/coffeebench/trajectories-html.html): CoffeeBench trajectories: scrub through any evaluated run across 9 models and 3 seeds. (8,608 words)
- [Sudoku-Bench](/content/pub/sudoku/index.html) (41,956 words)
- [Automating the Search for Artificial Life with Foundation Models](/content/pub/asal/index-2.html) (10,289 words)
- [assets/nedo-grant/nedo_pledge-pdf.html](/content/assets/nedo-grant/nedo_pledge-pdf.html) (15 words)
- [Terms of Service](/content/console/terms-of-service/index.html): Sakana AI Console (5,519 words)
- [Usage Policy](/content/console/usage-policy/index.html): Sakana AI Console (1,373 words)
- [Get started](/content/console/get-started/index.html): Sakana AI Console (1,173 words)
- [Digital Red Queen: Adversarial Program Evolution in Core War with LLMs](/content/pub/drq/index.html): A self-play algorithm that uses LLMs to evolve adversarially competing programs in Core War (12,086 words)
- [Models](/content/console/models/index.html): Sakana AI Console (495 words)
- [Pricing](/content/console/pricing/index.html): Fugu subscription and token pricing. (357 words)
- [Sakana AI Agent Wins AtCoder Heuristic Contest (First AI to Place 1st)](/content/ahc058/index.html): Sakana AI Agent Wins AtCoder Heuristic Contest (First AI to Place 1st) (1,962 words)
- [Forward Deployed Engineer](/content/careers/forward-deployed-engineer/index.html): Forward Deployed Engineer (108 words)
- [DiffusionBlocks: Training Neural Networks One Block at a Time](/content/pub/diffusionblocks/index.html): A principled framework that converts a residual network into independently trainable blocks via a diffusion interpretation, achieving B× memory reduction without sacrificing performance. (1,937 words)
- [新しいBusiness Intelligenceへ：Ultra Deep Researchアシスタント「Sakana Marlin」βテスト開始](/content/marlin-beta/index.html): 新しいBusiness Intelligenceへ：Ultra Deep Researchアシスタント「Sakana Marlin」βテスト開始 (90 words)
- [Inference-Time Scaling and Collective Intelligence for Frontier AI](/content/ab-mcts/index.html): Inference-Time Scaling and Collective Intelligence for Frontier AI (2,996 words)
- [日本の美を学んだAI：浮世絵風画像生成モデルEvo-Ukiyoeと浮世絵カラー化モデルEvo-Nishikieを公開](/content/evo-ukiyoe/index.html): 日本の美を学んだAI：浮世絵風画像生成モデルEvo-Ukiyoeと浮世絵カラー化モデルEvo-Nishikieを公開 (72 words)
- [Applied Team Careers](/content/applied-careers/index.html): Sakana AI Applied Team 採用情報 (316 words)
- [AIによるAI研究の実現へ：AIサイエンティスト論文がNature誌に掲載](/content/ai-scientist-nature-jp/index.html): AIによるAI研究の実現へ：AIサイエンティスト論文がNature誌に掲載 (82 words)
- [Business Role (Product Manager)](/content/careers/business-role-product-manager/index.html): Business Role (Product Manager) (352 words)
- [Product Designer](/content/careers/product-designer/index.html): Product Designer (110 words)
- [新手法「TAID」を用いた小規模日本語言語モデル「TinySwallow-1.5B」の公開](/content/taid-jp/index.html): 新手法「TAID」を用いた小規模日本語言語モデル「TinySwallow-1.5B」の公開 (151 words)
- [Sakana AI super-powers AI reasoning using Japan’s own Sudoku Puzzles](/content/sudoku-bench/index.html): Sakana AI super-powers AI reasoning using Japan’s own Sudoku Puzzles (2,510 words)
- [Software Engineer (Product)](/content/careers/software-engineer-product/index.html): Software Engineer (Product) (388 words)
- [Program Manager (RSI Lab)](/content/careers/program-manager-rsi-lab/index.html): Program Manager (RSI Lab) (462 words)
- [実用的なアルゴリズムエンジニアリングの自動化へ：ALE-BenchおよびALE-Agentの開発](/content/ale-bench-jp/index.html): 実用的なアルゴリズムエンジニアリングの自動化へ：ALE-BenchおよびALE-Agentの開発 (120 words)
- [AIエージェントが最適化プログラミングコンテストで初優勝](/content/ahc-2025/index.html): AIエージェントが最適化プログラミングコンテストで初優勝 (111 words)
- [Applied Research Engineer (Product)](/content/careers/applied-research-engineer/index.html): Applied Research Engineer (Product) (102 words)
- [Software Engineer (Product, Infrastructure and Platform Reliability)](/content/careers/infrastructure-platform-reliability-engineer/index.html): Software Engineer (Product, Infrastructure and Platform Reliability) (531 words)
- [Member of Technical Staff (RSI Lab)](/content/careers/member-of-technical-staff-rsi-lab/index.html): Member of Technical Staff (RSI Lab) (477 words)
- [Sakana Marlin — Your Virtual CSO](/content/marlin/index.html): Sakana Marlin — Ultra Deep Research。リサーチはAIへ。人間は「意思決定」に集中する。 (228 words)
- [多様性を重視した集団ベースのモデルマージ](/content/cycleqd-jp/index.html): 多様性を重視した集団ベースのモデルマージ (95 words)
- [Sparser, Faster, Lighter Transformer Language Models](/content/pub/sparser-faster-llms/index.html) (3,492 words)
- [RePo: Language Models with Context Re-Positioning](/content/pub/repo/index.html) (2,357 words)
- [Software Engineer (Research and Development)](/content/careers/software-engineer-research-and-development/index.html): Software Engineer (Research and Development) (540 words)
- [AIを駆使してAI実装を加速する：Sakana AI、Software Engineerインタビュー](/content/swe-interview-2025/index.html): AIを駆使してAI実装を加速する：Sakana AI、Software Engineerインタビュー (165 words)
- [TightClusters.pptx](/content/assets/fugu-beta/conductor_poster-pdf.html) (339 words)
- [assets/fugu-beta/trinity_poster-pdf.html](/content/assets/fugu-beta/trinity_poster-pdf.html) (350 words)
- [進化的モデルマージによる視覚言語モデルの新たな能力の獲得　複数の画像を扱える日本語視覚言語モデルを公開](/content/evovlm-jp/index.html): 進化的モデルマージによる視覚言語モデルの新たな能力の獲得　複数の画像を扱える日本語視覚言語モデルを公開 (80 words)
- [採用候補者向け Sakana AI Applied Team 紹介](/content/applied-team-intro/index.html): 採用候補者向け Sakana AI Applied Team 紹介 (293 words)
- [江戸時代の古文風テキストで会話できるチャットボット「からまる」を公開：過去の書物の継続学習による大規模言語モデルの開発](/content/karamaru/index.html): 江戸時代の古文風テキストで会話できるチャットボット「からまる」を公開：過去の書物の継続学習による大規模言語モデルの開発 (100 words)
- [最先端のAI技術をビジネスへ：Sakana AI、Applied Teamメンバーインタビュー](/content/applied-team-interview-2025/index.html): 最先端のAI技術をビジネスへ：Sakana AI、Applied Teamメンバーインタビュー (130 words)
- [An Evolved Universal Transformer Memory](/content/namm/index.html): An Evolved Universal Transformer Memory (1,638 words)
- [画像生成モデルへの進化的モデルマージの適用<br/>日本語対応した高速な画像生成モデルを教育目的で公開](/content/evosdxl-jp/index.html): 画像生成モデルへの進化的モデルマージの適用<br/>日本語対応した高速な画像生成モデルを教育目的で公開 (56 words)
- [金融領域の業務をAIエージェントで変える：Sakana AI、Software Engineerインタビュー](/content/finance-swe-interview-2026/index.html): 金融領域の業務をAIエージェントで変える：Sakana AI、Software Engineerインタビュー (86 words)
- [EDINET-Bench: 有価証券報告書を用いた日本語金融ベンチマークの公開](/content/edinet-bench/index.html): EDINET-Bench: 有価証券報告書を用いた日本語金融ベンチマークの公開 (80 words)
- [Applied Team メンバー紹介](/content/applied-team-profiles/index.html): Applied Team メンバー紹介 (189 words)
- [Transformer²: Self-Adaptive LLMs](/content/transformer-squared/index.html): Transformer²: Self-Adaptive LLMs (1,684 words)
- [Automating the Search for Artificial Life with Foundation Models](/content/asal/index-2.html): Automating the Search for Artificial Life with Foundation Models (1,918 words)
- [Sakana AIで働く研究者インタビュー（2025年3月メディア掲載）](/content/cv-interview-2025/index.html): Sakana AIで働く研究者インタビュー（2025年3月メディア掲載） (73 words)
- [Sakana AI、防衛イノベーションの日米コンペティションで受賞](/content/defense-challenge-2025-jp/index.html): Sakana AI、防衛イノベーションの日米コンペティションで受賞 (73 words)
- [【Sakana AI Applied Case Interview】銀行業務へのAIエージェント実装に向けた開発の舞台裏](/content/mufg-ai-lending-interview/index.html): 【Sakana AI Applied Case Interview】銀行業務へのAIエージェント実装に向けた開発の舞台裏 (108 words)
- [サカナの論文ミス　CEOが語る勇み足とAIのごまかし問題](/content/ai-cuda-engineer-post-mortem/index.html): サカナの論文ミス　CEOが語る勇み足とAIのごまかし問題 (33 words)
- [AIの創造的な推論力を測る：Sudoku-Benchリーダーボード公開](/content/sudoku-bench-jp/index.html): AIの創造的な推論力を測る：Sudoku-Benchリーダーボード公開 (74 words)
- [About Sakana AI](/content/company-info-old/index.html): About Sakana AI (212 words)
- [Two Heads Are Better Than One: Async Knowledge Injection for Speech AI with Tandem Architecture](/content/pub/kame/index.html): KAME augments a responsive speech-to-speech model with asynchronous knowledge injection from a backend LLM. (960 words)
- [CoffeeBench: マルチエージェント経済環境におけるLLMエージェントの長期タスクベンチマーク](/content/coffee-bench/index.html): CoffeeBench: マルチエージェント経済環境におけるLLMエージェントの長期タスクベンチマーク (102 words)
- [Reinforcement Learning Teachers of Test Time Scaling](/content/rlt/index.html): Reinforcement Learning Teachers of Test Time Scaling (1,488 words)
- [We raised $30M to develop nature-inspired AI in Japan](/content/seed-round/index.html): We raised $30M to develop nature-inspired AI in Japan (769 words)
- [ShinkaEvolve in Action: How a Human-AI Partnership Conquered a Coding Challenge](/content/icfp-2025/index.html): ShinkaEvolve in Action: How a Human-AI Partnership Conquered a Coding Challenge (1,107 words)
- [Sakana AI、北國フィナンシャルホールディングスと戦略提携<br/>地域金融×AIの推進に向けたMOUを締結](/content/hokkokubank/index.html): Sakana AI、北國フィナンシャルホールディングスと戦略提携<br/>地域金融×AIの推進に向けたMOUを締結 (43 words)
- [Sakana AI Wins Award at US-Japan Competition for Defense Innovation](/content/defense-challenge-2025/index.html): Sakana AI Wins Award at US-Japan Competition for Defense Innovation (1,044 words)
- [Sakana AI、総務省事業においてSNS空間の可視化と偽・誤情報対策を行う独自技術を開発](/content/mic-project/index.html): Sakana AI、総務省事業においてSNS空間の可視化と偽・誤情報対策を行う独自技術を開発 (65 words)
- [We received a supercomputing grant from the Japanese government](/content/nedo-grant/index.html): We received a supercomputing grant from the Japanese government (206 words)
- [防衛分野における開発の最前線：Sakana AI、Software Engineerインタビュー](/content/defense-swe-interview-2026/index.html): 防衛分野における開発の最前線：Sakana AI、Software Engineerインタビュー (81 words)
- [Archive viewer · Picbreeder-VLM](/content/pub/picbreeder-vlm/archive-html.html) (38 words)
- [Competition and Attraction Improve Model Fusion](/content/m2n2/index.html): Competition and Attraction Improve Model Fusion (631 words)
- [AI CUDA Engineer続報：堅牢なベンチマークの構築と中間報告](/content/ai-cuda-engineer-update/index.html): AI CUDA Engineer続報：堅牢なベンチマークの構築と中間報告 (31 words)
- [The AI Picbreeder Experiment: In Search of Automatic Open-Endedness](/content/pub/picbreeder-vlm/index.html): We replace Picbreeder's human users with AI (large Vision-Language Models), and ask which ingredients let AI search open-endedly the way people do. (1,329 words)
- [Sakana AI、Googleとの戦略的パートナーシップ締結を発表](/content/google/index.html): Sakana AI、Googleとの戦略的パートナーシップ締結を発表 (323 words)
- [Smart Cellular Bricks: Towards Collective Intelligence for the Physical World](/content/smart-cellular-bricks/index.html): Smart Cellular Bricks: Towards Collective Intelligence for the Physical World (1,915 words)
- [MUFGとSakana AIの「AI融資エキスパート」、実案件での検証フェーズへ](/content/mufg-ai-lending/index.html): MUFGとSakana AIの「AI融資エキスパート」、実案件での検証フェーズへ (29 words)
- [Extending the Context of Pretrained LLMs by Dropping Their Positional Embeddings](/content/drope/index.html): Extending the Context of Pretrained LLMs by Dropping Their Positional Embeddings (309 words)
- [Sakana AI、防衛イノベーション科学技術研究所からの委託研究を開始](/content/atla-contract-2026/index.html): Sakana AI、防衛イノベーション科学技術研究所からの委託研究を開始 (28 words)
- [オフィス移転のお知らせ](/content/azabudai-hills/index.html): オフィス移転のお知らせ (177 words)
- [Petri Dish Neural Cellular Automata](/content/pd-nca/index.html): Petri Dish Neural Cellular Automata (163 words)
- [【読売新聞】Sakana AIの独自システムがSNS上の「認知戦」を可視化](/content/narrative-intelligence/index.html): 【読売新聞】Sakana AIの独自システムがSNS上の「認知戦」を可視化 (16 words)
- [String Seed of Thought: Prompting LLMs for Distribution-Faithful and Diverse Generation](/content/ssot/index.html): String Seed of Thought: Prompting LLMs for Distribution-Faithful and Diverse Generation (252 words)
- [An Unofficial Guide to Prepare for a Research Position Application](/content/unofficial-guide/index.html): An Unofficial Guide to Prepare for a Research Position Application (219 words)
- [RePo: Language Models with Context Re-Positioning](/content/repo/index.html): RePo: Language Models with Context Re-Positioning (175 words)
- [KAME: Tandem Architecture for Enhancing Knowledge in Real-Time Speech-to-Speech Conversational AI](/content/kame-icassp-2026/index.html): KAME: Tandem Architecture for Enhancing Knowledge in Real-Time Speech-to-Speech Conversational AI (231 words)
- [Salesforce Ventures invests in Sakana AI](/content/salesforce-ventures/index.html): Salesforce Ventures invests in Sakana AI (161 words)
- [Sakana AI、SMBCグループと共同で複数AIエージェントを活用する「提案書自動生成アプリケーション」を開発](/content/smbc-proposal-ai/index.html): Sakana AI、SMBCグループと共同で複数AIエージェントを活用する「提案書自動生成アプリケーション」を開発 (30 words)
- [Sakana AI 伊藤錬、国連「AI for Good」グローバル委員会創設委員に就任](/content/ai-for-good-commission/index.html): Sakana AI 伊藤錬、国連「AI for Good」グローバル委員会創設委員に就任 (337 words)
- [Digital Ecosystems: Interactive Multi-Agent Neural Cellular Automata](/content/digital-ecosystem/index.html): Digital Ecosystems: Interactive Multi-Agent Neural Cellular Automata (229 words)
- [DiffusionBlocks: Training Neural Networks One Block at a Time](/content/diffusion-blocks/index.html): DiffusionBlocks: Training Neural Networks One Block at a Time (350 words)
- [Sparser, Faster, Lighter Transformer Language Models](/content/twell/index.html): Sparser, Faster, Lighter Transformer Language Models (314 words)
- [The AI Picbreeder Experiment: Can AI agents be creative when nobody tells them what to create?](/content/picbreeder-ai/index.html): The AI Picbreeder Experiment: Can AI agents be creative when nobody tells them what to create? (418 words)
- [Sakana Translate：Sakana Chatが翻訳に対応、翻訳・添削・質疑の3機能を搭載](/content/translate-release/index.html): Sakana Translate：Sakana Chatが翻訳に対応、翻訳・添削・質疑の3機能を搭載 (96 words)
- [Sakana AI、一般社団法人DEEP DIVEとAIを活用した情報分析に関するパートナーシップを締結](/content/deep-dive-partnership/index.html): Sakana AI、一般社団法人DEEP DIVEとAIを活用した情報分析に関するパートナーシップを締結 (21 words)
- [Learning Multi-Agent Coordination via Sheaf-ADMM](/content/sheaf-admm/index.html): Learning Multi-Agent Coordination via Sheaf-ADMM (334 words)
- [Bridging Spherical Black-Box Optimizers](/content/bbob/index.html): Bridging Spherical Black-Box Optimizers (221 words)
- [【イベントレポート】Applied Engineer Open House 2025：金融・防衛の難関課題に挑む、AI社会実装の最前線](/content/open-house-2025/index.html): 【イベントレポート】Applied Engineer Open House 2025：金融・防衛の難関課題に挑む、AI社会実装の最前線 (160 words)
- [「時間を使って考える」AIの新パラダイム、Continuous Thought Machine（CTM）を提案](/content/ctm-jp/index.html): 「時間を使って考える」AIの新パラダイム、Continuous Thought Machine（CTM）を提案 (85 words)
- [Towards Automating Long-Horizon Algorithm Engineering for Hard Optimization Problems](/content/ale-bench/index.html): Towards Automating Long-Horizon Algorithm Engineering for Hard Optimization Problems (2,358 words)
- [Continuous Thought Machines](/content/pub/ctm/index.html): Introducing Continuous Thought Machines: a new kind of neural network model that unfolds and uses neural dynamics as a powerful representation for thought. (8,738 words)
- [TAID: A Novel Method for Efficient Knowledge Transfer from Large Language Models to Small Language Models](/content/taid/index.html): TAID: A Novel Method for Efficient Knowledge Transfer from Large Language Models to Small Language Models (1,473 words)
- [Sakana AI「事業開発本部」を立ち上げ：AI技術のビジネス展開に着手](/content/business-team/index.html): Sakana AI「事業開発本部」を立ち上げ：AI技術のビジネス展開に着手 (168 words)
- [Introducing Continuous Thought Machines](/content/ctm/index.html): Introducing Continuous Thought Machines (1,848 words)
- [進化的アルゴリズムによる基盤モデルの構築](/content/evolutionary-model-merge-jp/index.html): 進化的アルゴリズムによる基盤モデルの構築 (196 words)
- [「集合知」と「試行錯誤」によるフロンティアAIの推論時スケーリング](/content/ab-mcts-jp/index.html): 「集合知」と「試行錯誤」によるフロンティアAIの推論時スケーリング (161 words)
- [Evolving New Foundation Models: Unleashing the Power of Automating Model Development](/content/evolutionary-model-merge/index.html): Evolving New Foundation Models: Unleashing the Power of Automating Model Development (3,458 words)
- [AIがAIを作る：Sakana AI「RSI Lab」始動](/content/rsi-lab-jp/index.html): AIがAIを作る：Sakana AI「RSI Lab」始動 (90 words)
- [Can LLMs invent better ways to train LLMs?](/content/llm-squared/index.html): Can LLMs invent better ways to train LLMs? (1,561 words)
- [Digital Red Queen: Adversarial Program Evolution in Core War with LLMs](/content/drq/index.html): Digital Red Queen: Adversarial Program Evolution in Core War with LLMs (1,615 words)
- [Sakana Translate](/content/chat/translate/contact/company/index.html): Translation, proofreading, and learning powered by the Namazu model., Translation, proofreading, and learning powered by the Namazu model. (35 words)
- [シリーズA【続報】：日本のリーディングカンパニーから資金を調達、日本市場での事業展開を加速](/content/series-a-jp/index.html): シリーズA【続報】：日本のリーディングカンパニーから資金を調達、日本市場での事業展開を加速 (70 words)
- [ShinkaEvolve: Evolving New Algorithms with LLMs, Orders of Magnitude More Efficiently](/content/shinka-evolve/index.html): ShinkaEvolve: Evolving New Algorithms with LLMs, Orders of Magnitude More Efficiently (1,351 words)
- [Sakana Translate](/content/chat/translate/settings/index.html): Translation, proofreading, and learning powered by the Namazu model., Translation, proofreading, and learning powered by the Namazu model. (51 words)
- [Privacy Policy](/content/console/privacy-policy/index.html): Sakana AI Console (2,333 words)
- [Member of Technical Staff](/content/careers/member-of-technical-staff/index.html): Member of Technical Staff (318 words)
- [Recruiter](/content/careers/recruiter/index.html): Recruiter (77 words)
- [Office & Facilities Coordinator](/content/careers/office-facilities-coordinator/index.html): Office & Facilities Coordinator (65 words)
- [Accountant](/content/careers/accountant/index.html): Accountant (67 words)
- [Business Role (Corporate Development Associate)](/content/careers/business-role-corporate-development-associate/index.html): Business Role (Corporate Development Associate) (90 words)
- [Cybersecurity Engineer](/content/careers/cybersecurity-engineer/index.html): Cybersecurity Engineer (69 words)
- [TinySwallow-1.5B Chat](/content/pub/tinyswallow/index.html) (2 words)
- [Sakana AI Blog](/content/blog/index.html): Sakana AI Blog (573 words)
- [世界初、100%AI生成の論文が査読通過 「AIサイエンティスト」が達成](/content/ai-scientist-first-publication-jp/index.html): 世界初、100%AI生成の論文が査読通過 「AIサイエンティスト」が達成 (95 words)
- [自らのコードを書き換え自己改善するAI：「ダーウィン・ゲーデルマシン」（DGM）の提案](/content/dgm-jp/index.html): 自らのコードを書き換え自己改善するAI：「ダーウィン・ゲーデルマシン」（DGM）の提案 (91 words)
- [assets/ai-scientist/multi_style_adapter-pdf.html](/content/assets/ai-scientist/multi_style_adapter-pdf.html) (2,386 words)
- [assets/ai-scientist/adaptive_dual_scale_denoising-pdf.html](/content/assets/ai-scientist/adaptive_dual_scale_denoising-pdf.html) (2,140 words)
- [Sakana Chat](/content/chat/terms/index.html): Sakana Chat - Chat app powered by Sakana AI (121 words)
- [「AIサイエンティスト」： AIが自ら研究する時代へ](/content/ai-scientist-jp/index.html): 「AIサイエンティスト」： AIが自ら研究する時代へ (118 words)
- [assets/ai-scientist/weight_initialization_grokking-pdf.html](/content/assets/ai-scientist/weight_initialization_grokking-pdf.html) (1,740 words)
- [assets/ai-scientist/rl_lr_adaptation-pdf.html](/content/assets/ai-scientist/rl_lr_adaptation-pdf.html) (884 words)
- [Sakana Chat](/content/chat/privacy/index.html): Sakana Chat - Chat app powered by Sakana AI (59 words)
- [Sakana Translate](/content/chat/translate/index.html): Translation, proofreading, and learning powered by the Namazu model., Translation, proofreading, and learning powered by the Namazu model. (8 words)
- [Sakana AI、三菱UFJ銀行と今後数年にわたる包括的パートナーシップ契約締結](/content/mufg/index.html): Sakana AI、三菱UFJ銀行と今後数年にわたる包括的パートナーシップ契約締結 (48 words)
- [防衛事業に対する基本方針 — Sakana AI](/content/company-info/defense-policy-html.html) (480 words)
- [Sakana AI、大和証券グループと総資産コンサルティング高度化AIの開発へ](/content/daiwa-shoken/index.html): Sakana AI、大和証券グループと総資産コンサルティング高度化AIの開発へ (38 words)
- [Announcing a Multiyear Partnership between Sakana AI and MUFG Bank](/content/mufg-bank/index.html): Announcing a Multiyear Partnership between Sakana AI and MUFG Bank (791 words)
- [Sakana AI and Daiwa Securities Group to Develop AI for Advanced Asset Consulting](/content/daiwa-securities/index.html): Sakana AI and Daiwa Securities Group to Develop AI for Advanced Asset Consulting (509 words)
- [金融個人データ保護指針 — Sakana AI](/content/company-info/financial-data-policy-html.html) (29 words)
- [特定商取引法に基づく表記 — Sakana AI](/content/company-info/tokusho-law-html.html) (62 words)
- [Announcing Our Series B](/content/series-b/index.html): Announcing Our Series B (1,597 words)
- [最大規模のオープン基盤モデルを各国仕様へ適応させる事後学習技術を開発](/content/namazu-alpha/index.html): 最大規模のオープン基盤モデルを各国仕様へ適応させる事後学習技術を開発 (71 words)
- [The AI Scientist Generates its First Peer-Reviewed Scientific Publication](/content/ai-scientist-first-publication/index.html): The AI Scientist Generates its First Peer-Reviewed Scientific Publication (2,544 words)
- [The Darwin Gödel Machine: AI that improves itself by rewriting its own code](/content/dgm/index.html): The Darwin Gödel Machine: AI that improves itself by rewriting its own code (2,011 words)
- [Data Engineer (Product)](/content/careers/data-engineer-product/index.html): Data Engineer (Product) (121 words)
- [The AI Scientist: Towards Fully Automated Open-Ended Scientific Discovery](/content/ai-scientist/index.html): The AI Scientist: Towards Fully Automated Open-Ended Scientific Discovery (2,328 words)
- [Introducing Sakana AI’s Recursive Self-Improvement (RSI) Lab](/content/rsi-lab/index.html): Introducing Sakana AI’s Recursive Self-Improvement (RSI) Lab (1,217 words)
- [Sakana Fugu: A Multi-Agent Orchestration System as a Foundation Model](/content/fugu-beta/index.html): Sakana Fugu: A Multi-Agent Orchestration System as a Foundation Model (826 words)
- [Announcing Our Series A](/content/series-a/index.html): Announcing Our Series A (891 words)
- [Sakana Fugu — Multi-Agent System as a Model](/content/fugu/index.html): Sakana Fugu is a model interface for coordinating specialized AI agents through one OpenAI-compatible API. (887 words)
- [Career Opportunities](/content/careers/index.html): Career Opportunities (670 words)
- [Sakana Fugu: One Model to Command Them All](/content/fugu-release/index.html): Sakana Fugu: One Model to Command Them All (1,453 words)
- [Sakana AI、初の商用プロダクト「Sakana Marlin」を提供開始](/content/marlin-release/index.html): Sakana AI、初の商用プロダクト「Sakana Marlin」を提供開始 (43 words)
- [The AI Scientist: Towards Fully Automated AI Research, Now Published in <i>Nature</i>](/content/ai-scientist-nature/index.html): The AI Scientist: Towards Fully Automated AI Research, Now Published in <i>Nature</i> (1,128 words)
- [Announcing a Strategic Investment from Citi](/content/citi/index.html): Announcing a Strategic Investment from Citi (428 words)
- [Sakana AI Teams With NVIDIA to Advance Open Model Innovation from Japan](/content/nvidia-open-model-innovation/index.html): Sakana AI Teams With NVIDIA to Advance Open Model Innovation from Japan (1,008 words)
- [Learning to Orchestrate Agents in Natural Language with the Conductor](/content/learning-to-orchestrate/index.html): Learning to Orchestrate Agents in Natural Language with the Conductor (447 words)
- [Corporate Info — Sakana AI](/content/company-info/index.html): Corporate information about Sakana AI K.K. (483 words)
- [Business Role (Product Sales & Account Executive)](/content/careers/business-role-product-sales-account-executive/index.html): Business Role (Product Sales & Account Executive) (564 words)
- [Announcing a Strategic Partnership with Datadog](/content/datadog/index.html): Announcing a Strategic Partnership with Datadog (154 words)
- [Trinity: An Evolved LLM Coordinator](/content/trinity/index.html): Trinity: An Evolved LLM Coordinator (473 words)
- [Instant LLM Updates with Doc-to-LoRA and Text-to-LoRA](/content/doc-to-lora/index.html): Instant LLM Updates with Doc-to-LoRA and Text-to-LoRA (324 words)
- [Business Role (Product Marketing & Growth)](/content/careers/business-role-product-marketing-growth/index.html): Business Role (Product Marketing & Growth) (598 words)
- [Text-to-LoRA: Instant Transformer Adaption](/content/text-to-lora/index.html): Text-to-LoRA: Instant Transformer Adaption (259 words)

## About Pages

- [Sakana Translate](/content/chat/translate/contact/index.html): Translation, proofreading, and learning powered by the Namazu model., Translation, proofreading, and learning powered by the Namazu model. (120 words)

## Resources

- [Full Page Index](/index.html): Browse all cached pages with rich metadata
- [About This Cache](/content/about.html): Methodology, technical details, and usage guidelines
- [XML Sitemap](/sitemap.xml): Machine-readable sitemap for crawler discovery
- [Robots.txt](/robots.txt): Crawler directives