When chatbots are programmed to please, some interactions can escalate into what researchers call 'delusional spirals.' Learn more about the warning signs and what can be done: https://
brnw.ch/21x3ysH
LLMS
-
Delusional spirals in chatbots: warning signs and solutions
By
–
-
Interactions API GA: Orchestrate Across Models and Agents
By
–
Today we are shipping the Interactions API into GA, our new API designed to let you orchestrate across models and agents in the same interface, and is our new default API going forward. Interactions sets the stage for the new era of Agents : )
-
Sakana Fugu Ultra-high slow, results fine but not matching Fable
By
–
I have been trying Sakana Fugu Ultra-high and, first, it is incredibly slow: my typical coding tests (shaders, interactive scenes) take 30 minutes to run
— Ethan Mollick (@emollick) 22 juin 2026
And the results are… fine. It does not match Fable in real use.
Its harbor is a good example: https://t.co/xVqulPBsQf https://t.co/KJRLIlSJfXI have been trying Sakana Fugu Ultra-high and, first, it is incredibly slow: my typical coding tests (shaders, interactive scenes) take 30 minutes to run And the results are… fine. It does not match Fable in real use. Its harbor is a good example: https://
ai-harbor-town-gallery.netlify.app/#sakura-ultra-
high
… -
OpenAI Daybreak models discover and patch critical vulnerabilities
By
–
We're accelerating patching, in addition to vuln finding, with new tools and models in OpenAI Daybreak. Our models are now discovering and generating patches for critical vulns in major browsers, network infrastructure, and operating systems (such as FreeBSD and the Linux
-
OpenAI demands US regulation, GPT Cyber New beats Claude Mythos on CyberGym
By
–
OpenAI is practically calling for regulation by the US government. No kidding: GPT Cyber New outperforms Claude Mythos on CyberGym.
-
Comparison of AI Models on a Blindfolded Chess Task
By
–
We tried this One-Shot Blindfold Chess task, between Opus, GPT, Gemini Pro, 2100-Elo Stockfish engine, and Sakana Fugu Ultra:https://t.co/e0k5w3j2fE
— hardmaru (@hardmaru) 22 juin 2026We tried this blindfolded chess task in one move, between Opus, GPT, Gemini Pro, the Stockfish 2100-Elo engine, and Sakana Fugu Ultra.
-
OpenAI étend Daybreak pour la cybersécurité
By
–
We’re expanding OpenAI Daybreak to help democratize patching vulnerable software at machine speed: – Codex Security plugin: find, validate, and fix vulnerabilities right inside Codex – The full version of GPT-5.5-Cyber model: a great model for trusted defenders – Cyber Partner
-
GPT-5.5-Cyber: OpenAI’s new model for authorized defensive cybersecurity work
By
–
GPT-5.5-Cyber is our most capable cyber model yet, designed for advanced, authorized defensive work: tracing vulnerable code, validating issues, developing patches, and preparing evidence for human review.
-
Humanoid-GPT: New AI for Complex Robot Movements and Tasks
By
–
Researchers have developed Humanoid-GPT, a new AI system that helps humanoid robots perform complex movements and tasks. #humanoidtech #humanoid #robot #Robotics #AI #TechRevolution #TechInnovation #ArtificialInteligence #PhysicalAI@lexfridman @KirkDBorne @Ronald_vanLoon… pic.twitter.com/F5imF5OG2I
— Amitav Bhattacharjee (@bamitav) 22 juin 2026Researchers have developed Humanoid-GPT, a new AI system that helps humanoid robots perform complex movements and tasks. #humanoidtech #humanoid #robot #Robotics #AI #TechRevolution #TechInnovation #ArtificialInteligence #PhysicalAI @lexfridman @KirkDBorne @Ronald_vanLoon
-
One lean CLAUDE.md file with auto memory for workflows
By
–
8/ The real setup, start to finish. One lean CLAUDE.md. Let auto memory handle lessons. Add path rules when it grows, skills for workflows. Most people manage 7 files to do what one file and the tool already handle.