#sakanaai — Public Fediverse posts

Sakana AI startet den Beta-Test für den autonomen Rechercheassistenten Sakana Marlin. Das System arbeitet bis zu acht Stunden nach einem Initial-Prompt und generiert umfangreiche Strategie-Reports. Technisch setzt Sakana auf die Suchmethode AB-MCTS und mehrfache LLM-Aufrufe zur Hypothesenevaluierung. Die Verifikation langer Dokumente bleibt fehleranfällig.

#SakanaAI #LLM #Automation #KI #News
https://www.all-ai.de/news/news26/sakana-marlin-ki

https://winbuzzer.com/2026/03/25/harvard-physicist-claude-ai-two-week-physics-research-xcxwbn/

Anthropic Says Harvard Physicist Used Claude to Complete Year of Research in Two Weeks

#AI #Anthropic #Claude #Physics #Science #ClaudeAI #ClaudeOpus4 #Research Productivity #SakanaAI #MatthewSchwartz #VibePhysics

https://winbuzzer.com/2026/03/25/harvard-physicist-claude-ai-two-week-physics-research-xcxwbn/

Anthropic Says Harvard Physicist Used Claude to Complete Year of Research in Two Weeks

#AI #Anthropic #Claude #Physics #Science #ClaudeAI #ClaudeOpus4 #Research Productivity #SakanaAI #MatthewSchwartz #VibePhysics

https://winbuzzer.com/2026/03/25/harvard-physicist-claude-ai-two-week-physics-research-xcxwbn/

Anthropic Says Harvard Physicist Used Claude to Complete Year of Research in Two Weeks

#AI #Anthropic #Claude #Physics #Science #ClaudeAI #ClaudeOpus4 #Research Productivity #SakanaAI #MatthewSchwartz #VibePhysics

https://winbuzzer.com/2026/03/25/harvard-physicist-claude-ai-two-week-physics-research-xcxwbn/

Anthropic Says Harvard Physicist Used Claude to Complete Year of Research in Two Weeks

#AI #Anthropic #Claude #Physics #Science #ClaudeAI #ClaudeOpus4 #Research Productivity #SakanaAI #MatthewSchwartz #VibePhysics

https://winbuzzer.com/2026/03/25/harvard-physicist-claude-ai-two-week-physics-research-xcxwbn/

Anthropic Says Harvard Physicist Used Claude to Complete Year of Research in Two Weeks

#AI #Anthropic #Claude #Physics #Science #ClaudeAI #ClaudeOpus4 #Research Productivity #SakanaAI #MatthewSchwartz #VibePhysics

Exciting breakthrough at Sakana AI! 🚀 Enterprise AI agents are revolutionizing optimization strategies by tackling complex computational challenges. Their innovative approach reconstructs solutions using advanced algorithms that push the boundaries of artificial intelligence performance. Curious how they're reshaping enterprise tech? #SakanaAI #AIOptimization #EnterpriseAgents #ArtificialIntelligence

🔗 https://aidailypost.com/news/sakana-ai-breakthrough-enterprise-agents-overcome-optimization

Exciting breakthrough at Sakana AI! 🚀 Enterprise AI agents are revolutionizing optimization strategies by tackling complex computational challenges. Their innovative approach reconstructs solutions using advanced algorithms that push the boundaries of artificial intelligence performance. Curious how they're reshaping enterprise tech? #SakanaAI #AIOptimization #EnterpriseAgents #ArtificialIntelligence

🔗 https://aidailypost.com/news/sakana-ai-breakthrough-enterprise-agents-overcome-optimization

With a ¥400 billion valuation after a big funding round, Sakana AI is targeting defense and banking markets. https://www.japantimes.co.jp/business/2025/11/17/companies/sakana-ai-valuation/?utm_medium=Social&utm_source=mastodon #business #companies #sakanaai #ai #mufg #banks #startups #tech

Since its 2023 launch, the value of Sakana AI has soared past $1 billion, making it Japan's fastest startup to reach so-called unicorn status. https://www.japantimes.co.jp/business/2025/09/10/companies/sakana-ai-interview/?utm_medium=Social&utm_source=mastodon #business #companies #sakanaai #ai #mufg #banks

#SakanaAI has introduced #MultiLLM #ABMCTS, a technique that enables multiple #LLMs to #collaborate on #complextasks. By combining the strengths of #differentmodels, the system outperforms individual LLMs by 30% on the ARC-AGI-2 benchmark. The open-source #TreeQuest #framework allows developers to implement this approach for their own tasks. https://venturebeat.com/ai/sakana-ais-treequest-deploy-multi-model-teams-that-outperform-individual-llms-by-30/?eicker.news #tech #media #news

Text-to-LoRA: Hypernetwork that generates task-specific LLM adapters (LoRAs)

https://github.com/SakanaAI/text-to-lora

#HackerNews #TextToLoRA #Hypernetwork #LLMAdapters #AIResearch #SakanaAI

Continuous Thought Machine: как Sakana AI научила модель думать тиками

Аналитический центр red_mad_robot продолжает следить за архитектурными прорывами в мире AI. В этот раз — экспериментальная модель от команды Sakana AI , которая предлагает мыслить не в терминах слоёв, а в терминах времени. Их Continuous Thought Machine (CTM) — попытка встроить в нейросеть внутреннюю динамику, вдохновлённую человеческим мозгом. Разбираем, как устроена архитектура, что такое «внутренние тики» и зачем нейросети синхронизировать собственные мысли — на примерах из CV, сортировки, Q&A и RL.

https://habr.com/ru/companies/redmadrobot/articles/916364/

#ai #sakana #ctm #ml #lstm #sakanaai #architecture #cifar10 #nlm

Continuous Thought Machine: как Sakana AI научила модель думать тиками

Аналитический центр red_mad_robot продолжает следить за архитектурными прорывами в мире AI. В этот раз — экспериментальная модель от команды Sakana AI , которая предлагает мыслить не в терминах слоёв, а в терминах времени. Их Continuous Thought Machine (CTM) — попытка встроить в нейросеть внутреннюю динамику, вдохновлённую человеческим мозгом. Разбираем, как устроена архитектура, что такое «внутренние тики» и зачем нейросети синхронизировать собственные мысли — на примерах из CV, сортировки, Q&A и RL.

https://habr.com/ru/companies/redmadrobot/articles/916364/

#ai #sakana #ctm #ml #lstm #sakanaai #architecture #cifar10 #nlm

Continuous Thought Machine: как Sakana AI научила модель думать тиками

Аналитический центр red_mad_robot продолжает следить за архитектурными прорывами в мире AI. В этот раз — экспериментальная модель от команды Sakana AI , которая предлагает мыслить не в терминах слоёв, а в терминах времени. Их Continuous Thought Machine (CTM) — попытка встроить в нейросеть внутреннюю динамику, вдохновлённую человеческим мозгом. Разбираем, как устроена архитектура, что такое «внутренние тики» и зачем нейросети синхронизировать собственные мысли — на примерах из CV, сортировки, Q&A и RL.

https://habr.com/ru/companies/redmadrobot/articles/916364/

#ai #sakana #ctm #ml #lstm #sakanaai #architecture #cifar10 #nlm

Continuous Thought Machine: как Sakana AI научила модель думать тиками

Аналитический центр red_mad_robot продолжает следить за архитектурными прорывами в мире AI. В этот раз — экспериментальная модель от команды Sakana AI , которая предлагает мыслить не в терминах слоёв, а в терминах времени. Их Continuous Thought Machine (CTM) — попытка встроить в нейросеть внутреннюю динамику, вдохновлённую человеческим мозгом. Разбираем, как устроена архитектура, что такое «внутренние тики» и зачем нейросети синхронизировать собственные мысли — на примерах из CV, сортировки, Q&A и RL.

https://habr.com/ru/companies/redmadrobot/articles/916364/

#ai #sakana #ctm #ml #lstm #sakanaai #architecture #cifar10 #nlm

Sakana AI Darwin Gödel Maschine - Selbstverbessernde KI?

• KI schreibt eigenen Code
• Evolution statt einfacher Updates
• 50 % Erfolgsquote bei Benchmarks

#ai #ki #artificialintelligence #sakanaai #selfimprovingai #evolutionäreki

Jetzt LIKEN, teilen, LESEN und FOLGEN! Schreib uns in den Kommentaren!

https://kinews24.de/sakana-ai-darwin-goedel-maschine-dgm-ki/

Sakana AI Darwin Gödel Maschine - Selbstverbessernde KI?

• KI schreibt eigenen Code
• Evolution statt einfacher Updates
• 50 % Erfolgsquote bei Benchmarks

#ai #ki #artificialintelligence #sakanaai #selfimprovingai #evolutionäreki

Jetzt LIKEN, teilen, LESEN und FOLGEN! Schreib uns in den Kommentaren!

https://kinews24.de/sakana-ai-darwin-goedel-maschine-dgm-ki/

Sakana AI Darwin Gödel Maschine - Selbstverbessernde KI?

• KI schreibt eigenen Code
• Evolution statt einfacher Updates
• 50 % Erfolgsquote bei Benchmarks

#ai #ki #artificialintelligence #sakanaai #selfimprovingai #evolutionäreki

Jetzt LIKEN, teilen, LESEN und FOLGEN! Schreib uns in den Kommentaren!

https://kinews24.de/sakana-ai-darwin-goedel-maschine-dgm-ki/

Sakana AI Darwin Gödel Maschine - Selbstverbessernde KI?

• KI schreibt eigenen Code
• Evolution statt einfacher Updates
• 50 % Erfolgsquote bei Benchmarks

#ai #ki #artificialintelligence #sakanaai #selfimprovingai #evolutionäreki

Jetzt LIKEN, teilen, LESEN und FOLGEN! Schreib uns in den Kommentaren!

https://kinews24.de/sakana-ai-darwin-goedel-maschine-dgm-ki/

Sakana AI Darwin Gödel Maschine - Selbstverbessernde KI?

• KI schreibt eigenen Code
• Evolution statt einfacher Updates
• 50 % Erfolgsquote bei Benchmarks

#ai #ki #artificialintelligence #sakanaai #selfimprovingai #evolutionäreki

Jetzt LIKEN, teilen, LESEN und FOLGEN! Schreib uns in den Kommentaren!

https://kinews24.de/sakana-ai-darwin-goedel-maschine-dgm-ki/

Sakana AI has signed a ¥5 billion ($34 million) deal with Mitsubishi UFJ to automate the creation of banking documents. https://www.japantimes.co.jp/business/2025/05/19/companies/mufg-sakana-ai-credit-approval-documents/?utm_medium=Social&utm_source=mastodon #business #companies #mufg #banks #sakanaai #ai #davidha #tech #entrepreneurs #startups

Japan should produce its own artificial intelligence technology for defense purposes as cutting-edge solutions could become a tool for political leverage, the CEO of a much-hyped Japanese AI startup said Tuesday. https://www.japantimes.co.jp/business/2025/05/13/companies/sakana-ai-defense/?utm_medium=Social&utm_source=mastodon #business #companies #sakanaai #davidha #ai #defense

Sakana AI: KI veröffentlicht erste Peer-Review-Studie OHNE menschliche Hilfe!

Autonome Forschung​
Keine menschliche Hilfe​
Wissenschaftlicher Durchbruch​

Jetzt lesen und folgen!

#ai #ki #artificialintelligence #kuenstlicheintelligenz #SakanaAI #PeerReview

https://kinews24.de/sakana-ai-ai-scientist-v2-peer-review/

Sakana AI korrigiert KI-Fehlinformationen

- Falsche KI-Behauptung entlarvt​
- Schnelle Reaktion von Sakana AI​
- Vertrauen in KI-Technologie gestärkt​

#ai #ki #artificialintelligence #sakanaai #fehlinformation #cudaengineering

https://kinews24.de/sakana-ai-korrigiert-falsche-ki-behauptung/

Google Titans, Sakana AI und Microsoft Majorana-1: Die Zukunft der KI und Quantenphysik

- Google Titans: Revolutioniert KI mit beispielloser Rechenleistung.​

- Sakana AI: Innovative Algorithmen für effizienteres maschinelles Lernen.​

- Microsoft Majorana-1: Durchbruch im Quantencomputing mit topologischen Qubits.​

#ai #ki #artificialintelligence #kuenstlicheintelligenz #googletitans #sakanaai #microsoftmajorana1

https://kinews24.de/google-titans-sakana-ai-microsoft-majorana-1/

KI-Agenten: Woche 7 – Top-Highlights

- Sakana AI beschleunigt Machine Learning: Neues System steigert die Geschwindigkeit um 100-Fache

- Lufthansa Industry Solutions veröffentlicht Whitepaper: „Smarte Assistenten“ zeigt, wie Unternehmen Chatbots effizient und rechtssicher einsetzen können.

#ai #ki #artificialintelligence #sakanaai #lufthansaindustrysolutions #chatbots #machinelearning #kuenstlicheintelligenz #reinforcementlearning #aiagents #kiagenten

Jetzt lesen!

https://kinews24.de/2025-woche-7-ki-agenten/

Sakana AI の間違いを徹底的に査読してみた (Colabコード付き)｜しらいはかせ(AI研究/Hacker作家)

Link

Sakana AI enthüllt bahnbrechende KI-Innovationen

Speicheroptimierung für LLMs
Automatisierte ALife-Forschung
Selbstadaptive KI-Systeme

#ai #ki #artificialintelligence #sakanaai #innovation #kuenstlicheintelligenz

https://kinews24.de/sakana-ai-science-highlights/

Sakana AI revolutioniert Machine Learning mit dem AI CUDA Engineer!

Sakana AI ist ein japanisches AI Science-Startup und hat bereits mehrere bahnbrechende Paper veröffentlicht.

Der AI CUDA Engineer topt ALLES, was Sakana bisher erstellt hat!

- 100-fache (!) Beschleunigung​
- Agentisches KI-System​http://kinews24.de
- Bahnbrechende Technologie​

ai #ki #artificialintelligence #kuenstlicheintelligenz #sakanaai #machinelearning #AICUDAEngineer

https://kinews24.de/sakana-ai-ai-cuda-engineer/

Sakana AI präsentiert Transformer 2: Selbstadaptive KI-Revolution

Neue Ära für Large Language Models (LLMs)
Ermöglicht dynamische Anpassung an Nutzereingaben
Verbesserte Effizienz und Genauigkeit in der KI-Kommunikation

#AI #KI #ArtificialIntelligence #KuenstlicheIntelligenz #SakanaAI #Transformer2

https://kinews24.de/sakana-ai-transformer-2/

Sakana AI präsentiert Transformer 2:

Revolutionäre selbstadaptive KI
Optimierung von Large Language Models
Neue Ära der künstlichen Intelligenz

#AI #KI #ArtificialIntelligence #KuenstlicheIntelligenz #SakanaAI #Transformer2

https://kinews24.de/sakana-ai-transformer-2/

Researchers at Sakana AI, a Tokyo-based AI startup, have introduced a memory optimization that enhances the efficiency of Transformer-based models, including large language models. #AI #MachineLearning #Transformers #SakanaAI #LLM #DeepLearning #NLP

https://winbuzzer.com/2024/12/17/sakana-ai-presents-memory-system-that-boosts-llm-efficiency-by-75-xcxwbn/

Sakana AI sees the potential for more strategic investors in Japan as rising U.S.-China tensions spur the country to boost its own artificial intelligence ecosystem. https://www.japantimes.co.jp/business/2024/09/09/companies/sakana-ai-partnerships/ #business #companies #sakanaai #startups #nvidia #ai

Japan’s Sakana AI partners Nvidia for research, raises $100M - The Tokyo-based startup raised $100 million in a Series A funding round ... - https://cointelegraph.com/news/sakana-ai-nvidia-partnership-boosts-japans-ai-industry #artificialintelligence #sakanaai #nvidia #japan

Good discussion around Sakana AI in this article,
https://arstechnica.com/information-technology/2024/08/research-ai-model-unexpectedly-modified-its-own-code-to-extend-runtime/
#AI #SakanaAI

Speaking of Zelkova trees, here is an image made with Evo-Ukiyoe, an AI image generator by Sakana AI. My prompt was something like "a Zelkova tree sprouting hamburgers as fruit" (in Japanese). This is so stupid I love it :-) https://huggingface.co/spaces/SakanaAI/Evo-Ukiyoe
#AI #Ukiyoe #SakanaAI #huggingface #art #Zelkova #hamburger

Research AI model unexpectedly modified its own code to extend runtime - Enlarge (credit: Moor Studio via Getty Images)

On Tuesday, Tok... - https://arstechnica.com/?p=2043110 #largelanguagemodels #theaiscientist #aiskepticism #hackernews #agenticai #aiagents #aimodels #sakanaai #davidha #biz⁢ #japan #tokyo #ai