ContextArticle

The tool's explanations should be treated with caution since the author can't personally verify their correctness, even though they seem plausible to him.

The author flags that he doesn't know enough about SQLite query plans to confirm the tool's output is correct, though it looks reasonable to him. ✦ AI generated

Simon Willison · Simon Willison's Weblog · 2026-07-18 · original ↗

Share card ↓Export .md Export .txt

Approach with caution, since I don't know enough about SQLite query plans to verify the results myself, but it seems cromulent enough to me.

Read full article ↗excerpt · fair-use quotation

Related moments

Mechanism◆Audio · 11:32 · 15m

The value of a conceptual breakthrough like Galois theory is verified on a timescale of a century, through a chain of human judgment rather than any immediate reward — meaning current RLVR-style reward functions are structurally unable to recognize the 'Galois instinct'.

Grant Sanderson · Dwarkesh Podcast

Mechanism◆Article · 223 words

Replit attacks the assumption that autonomous code genuinely works by building a self-testing reflection loop that drives a real browser to catch the 'Potemkin interface' — code that looks complete but fails when actually used.

Article author · ByteByteGo Newsletter

ClaimAudio · 8:35 · 1m

The ambition of what I even consider building has gone up a lot because of coding agents — you can push agents to end-to-end implement, verify, and maintain stuff, making you much more ambitious and your output more robust.

Simon Last · No Priors