The Measurement Gap
A 10,366-word research essay arguing that AI benchmark scores increasingly measure model-and-harness combinations rather than models alone.
Issue 2026-09-01Search the complete card catalogue
Across every issue
Titles, domains, summaries, categories, signals, dates, and editorial-interest scores across the entire published catalogue.
Results
A 10,366-word research essay arguing that AI benchmark scores increasingly measure model-and-harness combinations rather than models alone.
A source-traced archive of humanoid robot fighting leagues, fighters, editions, rankings, and bout results.
A planned gift product that collects milestone messages as audio and plays them when its physical cube is opened.
An independent directory comparing AI agents reachable through iMessage, WhatsApp, SMS, voice calls, and related channels.
A free Spanish-language collection of interactive tools for designing, analyzing, validating, and reporting medical research.
A Hebrew-language neighborhood marketplace connecting customers with nearby people who offer 3D printing, laser cutting, CNC, woodworking, clothing printing, and design.
Mila is a free Modern Greek learning app with offline dictionary search, daily vocabulary, spaced repetition, example sentences, and lessons from A1 to B1.
A German work platform for planning projects, managing files and teams, commissioning work, and running routine agents in the browser, desktop app, or on your own server.
A browser-based cat factory game growing out of a Discord community, with uranium shifts, fishing trips, ranks, and a world still under construction.
An offline practice site simulating Indian election machines, paperwork, procedures, and polling-day roles.
A professional service that restores antique French silverware and hand-engraves it with names, dates, phrases, or brand messages.
A source-linked status directory for lifts, railways, huts, refuges, routes, glaciers, and access across the Mont Blanc massif.
An Android app for logging daily sleep, mood, energy, thoughts, medication, and related measures, then reviewing or sharing the results.
An offline tool for IDELs that organizes postoperative assessments, tracks change across visits, and prepares a concise handoff.
A Windows VJ deck that turns audio, shaders, video clips, cameras, MIDI, and projection surfaces into live visuals.
A lightweight cookie-consent banner that blocks scripts before they load, supports Google Consent Mode v2, and makes no server calls.
A Spanish Minecraft host publishes node load, caps sales when capacity is full, and lists plans from €4 to €24 per month.
A short psychological-horror game playable in the browser, set behind the counter of a neighborhood Filipino sari-sari store.
A civic archive tracking Beacon, New York, council meetings, agendas, documents, and property-related proceedings.
A free daily 9×9 crossword built from recent world news, with three difficulty levels and links to the reporting behind its answers.
A bilingual AI timeline and short-course site that turns research papers, official announcements, and government records into weekly lessons for middle- and high-school students.
A Newton, Massachusetts family is hiring companions, caregivers, overnight helpers, and a lead supervisor for their 82-year-old mother.
A free address-based report of the aircraft passing overhead, including how often they might be heard and how loud the loudest flight was.
A Delhi student-safety campaign documenting eight demands after the site says a Satya Niketan PG collapsed, with a public checklist and anonymous hazard reporting.