scout https://scout.jonno.nz One short digest a day: the handful of new papers that actually matter if you build with LLMs, plus an interactive explainer of the best one. en Fri, 10 Jul 2026 04:15:18 GMT Spec checklists beat bigger test budgets, plus cost and reliability patterns for production agents https://scout.jonno.nz/p/2026-07-09/ https://scout.jonno.nz/p/2026-07-09/ Thu, 09 Jul 2026 08:00:00 +1200 Grounding the LLM test-writer in the task spec — one prompt line — beats doubling the test budget (+38pp, cross-vendor). Six more papers on running agents cheaper and catching silent failures: capacity placement, workflow crystallization, trace triage, write gates, and judge audits. Pick: Specification Grounding Drives Test Effectiveness for LLM Code.