Category: Working With AI

Where AI Actually Saves Time in Research, and Where It Costs You

August 31, 2026 No Comments

A working rule from a PhD student: use AI where being wrong is cheap and detectable, avoid it where being wrong is expensive or invisible.

GPT-5.6 model routing: Sol, Terra and Luna are three tiers but not a ladder, with a 25x price gap, a 0.1 point difference on Agents Last Exam and a 48 point gap on long-context recall

Not a Ladder: What OpenAI’s Own Numbers Say About Sol, Terra and Luna

August 13, 2026 No Comments

OpenAI’s published tables put Terra and Luna within a tenth of a point on professional agentic work and 48 points apart on long-context recall. Route by failure mode, not difficulty.