The man paid to prove Fermat by hand says Claude did it in 11 days
THE SO WHAT
Claude spinning up dozens of agents to generate 13 million lines of Lean and 30,300 intermediate proofs in 11 days is a preview of AI as a force-multiplier for formal methods, not just code autocomplete. For teams in safety-critical domains, this points toward feasible large-scale verification workflows that were previously cost-prohibitive.
READ THE SOURCE
MORE FROM THE WIRE
Applied AIOpenAI, Microsoft face copyright lawsuit from local newspapers
Copyright risk is moving from abstract to line-item as local publishers join the litigation wave against model trainers. Any team training or fine-tuning on web-scale data should be logging provenance and licensing assumptions now — retroactive cleanup will only get more expensive.
Applied AIxAI Loses Legal Battle Against Minnesota Nudification Law, War Rages on
State-level nudification laws surviving first contact with an AI lab means non-consensual sexual imagery is becoming a hard legal boundary, not just a policy line. If you ship generative models or image tooling into the US, you now need state-by-state compliance and logging strategies, not a single global content policy.
Applied AIEU tech chief says AI guardrails are coming to the US anyway
Convergence on AI guardrails via Brussels-first regulation and US courts-state law second means multinationals should plan for a de facto shared baseline, not two incompatible regimes. If your AI roadmap assumes the US will stay materially looser than the EU, revisit data use, logging, and explainability standards now.
Applied AIKenya bet its graduates on gig work in 2022. ChatGPT launched that year
Kenya’s national bet on online outsourcing colliding with ChatGPT-era automation is a live case study in how quickly AI can erode newly built labor markets. If your business leans on low-cost digital piecework, treat that dependency as duration-limited and start mapping which tasks will be first to compress.