The Carwash Test: Can AI Hold on to the Actual Question?
A longitudinal AI reasoning diagnostic that tests whether models can preserve a prompt’s logical constraint when familiar wording pushes them toward the wrong answer.
Issue 2026-08-29Search the complete card catalogue
Across every issue
Titles, domains, summaries, categories, signals, dates, and editorial-interest scores across the entire published catalogue.
Results
A longitudinal AI reasoning diagnostic that tests whether models can preserve a prompt’s logical constraint when familiar wording pushes them toward the wrong answer.
An independent LGBTQIA+ bookshop in Glasgow’s southside, selling queer books, comics, zines, history, art, and writing.
A forming, peer-led recovery effort connecting people in Catoosa County, Georgia, with free confidential support, meetings, and existing treatment resources.
This Korean-language tool lets visitors search a surname and bon-gwan, or clan origin, to compare records of independence figures and people identified in official pro-Japanese-collaboration records.
Italian puzzle books for children aged 6–8, combining 90 paper games with a treasure hunt that runs through the whole volume.
A provenance-first IP intelligence database and live lookup API covering IPv4 and IPv6 ranges, network ownership, location, hosting, VPN, Tor, mobile, and threat signals.
A Windows and macOS utility that assigns sound effects to individual keyboard keys, from typewriter bells and explosions to user-supplied audio clips.
A small game studio making free browser strategy games, including railroad empire-builder Waybill and influence-driven world game Empire Rising.
A free decluttering wheel picks one manageable area and one cleaning task based on the time you have.
An iOS app in development that transforms Apple Health and Strava runs, rides, and walks into animated map videos.
A private Mac and iPad workspace for searching, comparing, organizing, and carefully bulk-editing Apple Contacts.
A Minecraft assistant that turns your inventory or building goal into craftable items, recipes, production methods, and raw-resource chains.
A countertop-appliance planner that measures available inches, working gaps, appliance overlap, and electrical load before you buy.
CrateForge converts piece dimensions and box specifications into printable, shop-ready cut sheets for export packers, crating shops, and machinery movers.
A collection of browser-based tools for comparing, deduplicating, cleaning, matching, and reformatting lists without uploading the data.
A Korean phone-based assistance service that lets older adults order food, buy household goods, or call a taxi without installing an app or creating an account.
Dented Kettle is a three-person board game publisher preparing Long Haul, a two-to-four-player game about running an airline.
A Bangkok studio turns existing agency files into printed briefing books, presentation slides, field maps, and graph-reading workshops.
An open-source cross-platform app for running bioacoustic machine-learning models, reviewing detections, and annotating audio datasets.
An engineering prototype that adds motor drive, control, stereo vision, route planning, obstacle avoidance, and remote driving to a manual wheelchair.
An AI workflow for manufacturing distributors that turns emailed purchase orders into ERP-ready sales orders for human approval.
EATLAS is a visual archive that treats everyday ingredients as numbered specimens, combining botanical knowledge, cooking technique, flavour pairings, waste reduction, and recipes.
A San Francisco social-dining concept where everyone sees one daily restaurant pick, taps “I’m in,” and joins a real dinner with other people.
A prelaunch service where local shops post one item at 50% off during the final hours of the day.