Research-backed, actionable articles on AI at work, endurance training and how people learn. Every claim has a source.
Cues marking the relevant part of a slide lifted retention by 0.53 SD across 103 studies. Ten papers on slide design, and the deck rules I follow now.
Read the article →Thirteen percent of Boston Marathon finishers ended low on sodium, and 93.9% of ultrarunners took salt pills anyway. Ten studies, and a way to set your own rate.
Read the article →Nine studies on hallucinated references: the fabrication rates by model and field, why retrieval tools still miss 17 to 33%, and the check that catches them.
Read the article →A 2025 meta-analysis of 60 experiments puts beginners at d = 0.505 with full guidance and experienced learners at -0.428 without it. A rule for which to use.
Read the article →Ten easy days at 40°C lifted time-trial work 8% in the heat and plasma volume 6.5%. The evidence, the sauna and bath alternatives, and the block I run.
Read the article →Performance kept sliding for 14 nights while the feeling of sleepiness leveled off. Eight studies on sleep debt, banked sleep and which end of the night to cut.
Read the article →Reordering two answers handed the win to the weaker model on 66 of 80 questions. Ten studies on LLM judges, their biases, and a validation protocol.
Read the article →Harvard physics students learned more in the classes they rated lower. Six studies on why satisfaction scores miss learning, and a feedback form that does not.
Read the article →Two weeks out, cut volume 41 to 60% and leave the hard sessions alone. Nine studies on taper length, what to cut, and the overload block that backfires.
Read the article →Eight studies on the security of AI-written code: the failure rates by language, the confidence gap it creates, and the review routine that catches the weaknesses.
Read the article →