The agent lied to you, and the benchmark caught it. microsoft and hugging face released thinkingbox…
the agent lied to you, and the benchmark caught it. microsoft and hugging face released thinkingbox on oct 3, a benchmark that runs agent tasks 20 times and scores the actual state of the database afterward, not the agent's claim.
saying done and being done are now separable metrics.
Context
Microsoft's ThinkingBox grades agents on the records they leave behind, not their final message. ThinkingBox-Bench has 507 executable tasks across five business domains, and each is run 20 times to test repeatability. A Hugging Face blog dated October 3 says it is now available there.
Microsoft's own post describes isolated stateful tool environments, a simulated user, and inspection of side effects.
Date: Microsoft's original ThinkingBox post is dated August 19. October 3 is the Hugging Face availability post.
The post says 'scores the actual state of the database', which matches Microsoft's description of inspecting backend state.
'Saying done and being done are separable metrics' is the author's opinion.
Related work
- Hugging Face: The Agent Said It Was Done ↗The Oct 3 availability post.
- Microsoft: ThinkingBox ↗The original Aug 19 announcement.
- microsoft/ThinkingBox-Bench dataset ↗The 507-task dataset.
Watch next
- Look for leaderboard results of named agents on ThinkingBox-Bench.
Sources
- Hugging Face blog, Oct 3, 2026huggingface.co
- Microsoft Command Line, Aug 19, 2026commandline.microsoft.com
- Hugging Face: ThinkingBox-Benchhuggingface.co
- GitHub: ThinkingBox-Bench v1 READMEgithub.com
Provenance
The note above is reproduced unedited from the original post, first published on Threads on 9 October 2026 at 22:02 IST. Sources are the papers and datasets the note draws on.
View the original post ↗Embed this note
More notes
The air is now being asked to keep its own ledger
the air is now being asked to keep its own ledger: ecmwf’s aifs compo becomes the first ai model to forecast atmospheric composition globally every three hours, cleanair simulates 365 days of pm2.5 over china in ten seconds, and a unified framework maps six pollutants at one kilometer across the whole country. the air now files its own composition report.
read the note →The current is now being asked to draw its own map
the current is now being asked to draw its own map: china’s langya 2.0 predicts six ocean phenomena including internal waves and mesoscale eddies, a deep net called wenhai resolves eddies globally with air sea flux formulas built in, and scripps infers surface currents from the way temperature patterns deform in satellite images. the ocean now files its own circulation report.
read the note →The soil is now being asked to report its own carbon
the soil is now being asked to report its own carbon: a nix color sensor paired with generative data augmentation predicts soil organic carbon without a lab, random forest drives 74 percent of soil health mapping studies, and sentinel 2 tracks five year carbon change across france and italy from 922 samples. the dirt now files its own carbon account.
read the note →