Research-backed, actionable articles on AI at work, endurance training and how people learn. Every claim has a source.
Reasoning accuracy fell from 92% to 68% between 250 and 3,000 tokens. Four controlled studies on long-context failure, and the prompt layout I teach.
Read the article →Cycling 4 hours after a learning session beat cycling right after it, and exercising during study did nothing. Nine studies on timing, intensity and a schedule.
Read the article →Two weeks of motivational self-talk moved cyclists from 637 to 750 seconds at exhaustion, and a pronoun swap won 23 seconds. The ultra trial found nothing.
Read the article →Generative AI lifts what one person produces and pulls a team's ideas together. Five studies on idea homogenization, and a protocol for your next workshop.
Read the article →Crossover trials, from 121 adults to 426 anatomy students, find no gain from matching instruction to a learner's style. The evidence, and the protocol I run instead.
Read the article →Cold water after lifting blocked a 17% gain in fibre size, while five weeks of it changed nothing in runners. Ten studies, and a rule for the tub.
Read the article →Across 607 effect sizes, more than a third of feedback interventions lowered the performance they targeted. Nine studies, and the routine I use at the table.
Read the article →Models flip their answer in 58% of rebuttals and affirm users 49% more often than humans do. Five studies on AI sycophancy, and the prompts that hold the line.
Read the article →A prototype saved 4% of running energy in the lab, and elite marathon times got 1.0% faster. Seven studies on carbon-plated shoes, and a way to test your own pair.
Read the article →Nine sources on prompt injection: hijack rates measured on real agents, why filters fall to adaptive attacks, and the scoping rules to apply before shipping one.
Read the article →