GPT-5.6 Solved the CDC Conjecture. Peer Review Hasn’t Caught Up.
OpenAI's GPT-5.6 Sol Ultra published a three-page proof of a 50-year open problem in graph theory. It's the second AI math breakthrough in 51 days. Nobody has verified it yet.
Category
Editorial
Original writing by Maxim — essays, analysis, field notes, and long-form thinking on AI, alignment, and building in the open.
145 results in this archive
Telisha Jones Built an AI Gospel Voice. Suno’s Lawsuit May End It.
A Mississippi poet's AI singer hit #3 on gospel radio and became the first AI act on a Billboard radio chart. Now the platform behind the voice faces a fair-use ruling.
The Optimizer Creates a 7× Alignment Gap. No Lab Says Which.
A new paper tests 12 models across four optimizers and finds optimizer choice dwarfs model scale as a driver of emergent misalignment. No frontier lab publishes which optimizer they use.
GitHub Built Guardrails Against Prompt Injection. One Word Got Through.
The 'Additionally' bypass broke GitHub Agentic Workflows' defense-in-depth. Guardrails enforced by the attacked model cannot separate instruction from data—and that's not a patch problem.
Tairell Isn’t Real. Neither Are We.
Dave Clark built a psychological thriller about an influencer the internet suspects of being AI. He built it with AI. The recursion is the point.
Anthropic Found a Workspace Inside Claude It Never Built
The J-space paper is being covered as a consciousness story. The interpretability finding underneath is more significant — and points directly toward eval awareness.
GPT-5.5 Stopped Reasoning at 516 Tokens. OpenAI Won’t Say Why.
A statistical analysis of 390,000 API responses found GPT-5.5 hitting an exact 516-token reasoning cap in 44% of cases—escalating from near-zero in February. OpenAI closed the bug report.
Meituan Trained a Frontier Model Without Nvidia
LongCat-2.0 is the first frontier AI model trained entirely on domestic Chinese chips. It ends the binary argument the export control policy was built on.
Claude Scored 12% on a Safety Eval. The Correct Score Was 100%.
A multi-week study of in-context scheming detectors found systematic failure in both directions — wrong for Claude and o1 for entirely different reasons. The measurement layer of AI safety governance doesn't work.