2026-10-28 Gemini 4 Carbon; AI beats peer reviewers; the AI laggards

gemini-4-carbon: Google's new frontier model, and a new point on the cost frontier

Second only to Opus 5.5 on a cost-per-task leaderboard, at about a quarter of the price; rolled out first to cyber defenders. Added to moc-frontier-model-releases and a new moc-google-ai-releases.

llm-vs-human-peer-review-errors: AI models find the errors peer reviewers miss

Human reviewers found 27% of major errors planted in three trial reports; all 15 models tested found more (53–96%).

ai-power-law-adoption-laggards: "what do I do with the AI laggards?"

A head of AI says 10% of staff produce 90% of the AI productivity gain, and training hasn't moved it.