Vibeleaderboard
← All Intel
Intel / article

More than just code review

Source
simonwillison.net
Date

Context

Using a coding still requires deciding what should change and establishing that the change works. Simon Willison argues in his August 22 note that reading every generated line is only one way to perform that verification.

Willison's separate guide to test-first development gives a concrete method: write a test for the intended behavior, confirm that it fails, then implement the change and check that it passes. Watching the test fail first establishes that it can detect the missing behavior. This makes the verification question more specific than whether the generated code looks plausible.

Terms in this piece · Glossary
  • AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
  • agent skill — A reusable instruction file that teaches an agent how to do one job well — the procedure, the tools, and what counts as done.
Read the source simonwillison.net
Recommended reads
Comments

Checking sign-in…

Loading comments…