How to Solve Layer Violations using ArchUnit and ts-arch
Layer violations slip past wiki diagrams under delivery pressure. ArchUnit and ts-arch turn dependency rules into executable tests that fail in CI and agent loops.
agents
5 posts tagged agents
Keep Agent Work Alive After a Crash
Long AI jobs should not vanish when a laptop sleeps or a tab crashes. File-backed progress lets restarts resume and risk drop. Teams skip a cloud workflow bill when durability alone is the goal.
Loop Until the Outcome Is Met, Not the Chat Ends
One-shot AI prompts stop too early. A loop with a clear goal, reusable playbooks, and hard stop rules cuts wasted spend and unfinished work.
How to Solve False Done Claims using Validation Contracts
Vague done claims green-light bugs in AI and human handoffs. Validation contracts turn success into numbered, falsifiable behavior checks with proof paths.
Split AI Coding Work Across Three Roles
Teams split AI coding work into planner, builder, and checker roles. Written gates and one editor at a time reduce mistakes. Done means proof, not chat confidence.