June 2026
- Jun 22, 2026ConceptsGoodhart's Law: When a Measure Becomes a TargetCharles Goodhart noticed in 1975 that the Bank of England's stable money-supply relationships broke down the moment it started targeting them, because targeting changed the system that produced the correlation. Marilyn Strathern gave it the phrasing everyone quotes in 1997. Manheim and Garrabrant later split it into four mechanisms, and Choi, Hecht and Tayler found surrogation deepens exactly when pay is tied to the proxy.
- Jun 22, 2026ConceptsThe Spacing Effect: Why Cramming Fails and Spaced Practice WinsEbbinghaus found in the 1880s that spreading repetitions across days needed fewer total repetitions than massing them, so spacing is not just better, it is cheaper. The 2006 Cepeda and Pashler meta-analysis, pooling 254 studies and more than 14,000 participants, confirmed the advantage is large and gave a practical rule: set the gap between sessions at roughly 10 to 30 percent of how long you need to remember. Covers Bjork's desirable difficulties and the fluency illusion.
- Jun 5, 2026Work SkillsProductivity Evidence Map: What the Research Actually SupportsTask-switching cost and the benefit of short breaks are two of the better-supported findings in productivity research, yet each was disclosed by only 3 of 7 usable top search results, 42.9 percent, less often than the Pomodoro 25/5 ratio at 7 of 10, 70 percent, a claim with no controlled evidence for that specific number. This map rates each popular claim Robust, Supported, or Weak, cites the study behind it, and lists every page checked in the August 2026 measurement.
- Jun 5, 2026ConceptsCognitive Bias Evidence Map: Robust vs. Failed-to-Replicate BiasesNot every famous bias is equally well supported. This map rates ten effects by the strength of the record behind them, separating the ones tested at large scale and failed from the ones never tested at large scale at all, like the planning fallacy. The sharpest case is ego depletion: a pre-registered replication across 23 labs with 2,141 participants returned d = 0.04, a confidence interval straddling zero.
- Jun 5, 2026ConceptsLearning Science Evidence Map: What the Research Actually ShowsAn original measurement sits alongside the research ratings: of the usable top search results for learning styles, only 2 of 7 mentioned that the meshing hypothesis has repeatedly failed controlled tests, the lowest disclosure rate measured, while every result checked for highlighting and re-reading noted their low utility. Eight techniques rated on evidence strength and effect type, each tied to a named peer-reviewed source.
- Jun 1, 2026TechnologyAI for Knowledge Work: Practical Workflows, Prompts, and LimitsLarge language models generate plausible continuations rather than retrieve verified facts, which explains both their strengths and their failures. They are reliable at transforming text you provide, drafting, explaining and brainstorming, and unreliable at citations, statistics, dates and quotes, which arrive fabricated with full confidence. So supply the source material, ask for the passage behind each claim, and verify anything consequential.
- Jun 1, 2026ConceptsBetter Decisions Under Uncertainty: A Practical Guide and ToolkitA sound decision can still produce a bad outcome, and luck can hide weak reasoning, so the only thing you can improve is the process. The sharpest figure here is a calibration one: when people say they are 90 percent confident, studies consistently find them right roughly 70 percent of the time. From there it works through availability, sunk cost and anchoring, Gary Klein's premortem protocol, Bezos's two-way door test, and a ten-item checklist.
No guides match that search. Try a shorter phrase, or clear the search.