LLM Eval Tooling: Four Categories, and the Two That Matter
A held-out task set and twenty assertions beat any platform bought without one. Plus the judge biases that invalidate scores.
The apps and dev tools built on top of new models, tested against what they claim.
19 articles
A held-out task set and twenty assertions beat any platform bought without one. Plus the judge biases that invalidate scores.
Two ways to get reliable, machine-readable responses out of a model, and the failure modes each one avoids. Model Drop breaks down what actually matters.
A roundup of inline completion, chat, and agent extensions worth installing, and what each is actually good for. Model Drop breaks down what actually matters.
Four categories, three conditions that justify adoption, and the reason tool design beats orchestration every time.
Most of these aim at the wrong half of review. Knowing what a change breaks beats knowing whether it reads well.
Three categories, distinguished by how much review they generate. Leaderboard position predicts almost nothing about your repo.
Context bills every call. An index bills once. That arithmetic has outlived every context window expansion so far.