Menu

1 / 3

I Built a Claim-Checker for Writing With Codex. It Passed a Wrong Claim.

Reading 0:00
15s threshold

I have had to think carefully about how technical results are phrased, especially when a sentence gets rewritten to sound cleaner. One easy mistake I've seen is turning a 20% reduction in error into a 20% improvement in accuracy. A 20% cut in error shrinks a small remaining number, so accuracy itself can barely move. The number survives the rewrite, but what it describes changes completely.

That kind of mistake is subtle enough to survive a quick read, because nothing in the sentence looks out of place. Recall can quietly become accuracy, a figure can drift by ten points, or an observed association can pick up a causal verb on its way to a tidier sentence. It is the kind of mistake I wanted Evidence Desk to help catch.

Evidence Desk is a browser tool I built with Codex. You give it a draft along with the source files the draft came from, and it breaks the draft into claims.…

Continue reading — create a free account

Join HashtagPLUS to read full articles, follow hashtags, vote, and join the conversation.