A bigger context window gave me a worse answer, repeatably
What you leave out of a context pack is a design decision, not waste.
AI-Driven Development
Shipping production software with coding agents — context, evals, guardrails, and what breaks at scale.
36 posts
One warm context beats five cold ones
Per-layer workers each pay full context assembly. Loading once and generating inline amortises it — and the measured saving was smaller than estimated.
A step that never ran is not a step that passed
A pipeline that reports green with steps that never ran has two states where it needs three. Not-run must block completion or it silently becomes success.
Is this derived thing current? Make it a comparison
Staleness checked by timestamp or by memory fails quietly. A digest of the declared inputs, stamped beside the output, makes currency a comparison.
Which artifact is actually the design
When several artifacts describe one screen, only the highest-fidelity one is read by codegen. Editing a lower tier looks like a change and reaches nothing.
The running app and the spec disagree, and only one is documented
Hand-editing generated source creates a divergence from the spec that nothing reports and the next regeneration silently erases. The fix belongs upstream.
Print the list before you touch anything
A batch ceiling refuses work without telling you how much exists. Enumerating first turns an unknown amount of work into a finite list you can decide about.
The orchestrator that wrote the code itself
When the coordinator writes a layer instead of dispatching its generator, the conventions the generator carries are gone and the output still compiles.
Editing the generated thing is how the spec stops meaning anything
A change implemented directly in source skips the spec, so every report downstream is computed without it. The feature works and nothing can see it.
Every generated file should say what made it
A generated artifact carries no record of how it was made once the process ends. A provenance sidecar makes which-path-produced-this answerable by grep.
It returned an identifier and produced nothing
An exhausted quota returned a success with a session id and no artifact, and the message pointed at the prompt. The misdirection cost more than the outage.
Three hundred and three of three hundred and four
A mandatory generation step had produced its fallback on 303 of 304 artifacts, for weeks, because the flag that selected the real path crashed on any real file.