Testing
Topic archive • 5 matches
AI-generated: summaries written by AI from the linked sources. How we use AI
2026-11-10
Tips
Prompt Tips
Anthropic documentation, quoted by Simon Willison, says Claude Fable 5.2 responds to prompts about who owns the code it works on. In Anthropic's testing on small coding tasks, when a prompt said nothing about ownership or framed the codebase as someone else's, the model tended to add to the existing structure even when that repeated existing code.
Permalink
2026-11-11
Technology
Based on a transcript excerpt: AI Explained analyzes the Fable 5.2 and Gemini Carbon model releases, drawing on a 264-page system card, a 30-page METR report on AI acceleration, and the creator's own testing. The video covers benchmark results, a 'Max Bench' test designed to catch models that memorized answers, the remote labor index, and claims from Anthropic that the pace of AI improvement will quicken. It also discusses whether benchmark results hold up to scrutiny.
AI News & Analysis • AI Explained
Permalink
2026-11-14
Technology
Google's Gemini 4 Carbon and Anthropic's Fable 5.2 have displaced OpenAI's GPT-6.1 Sol from the top of the MathArena math benchmark, the first time in a while no OpenAI model leads. Gemini 4 Carbon reportedly costs 15 times more than GPT-6.1 Sol and ignored the time limit during testing, almost never giving up on a task.
AI Models • The Decoder
Permalink