Blog
Notes on measuring what you ship
Engineering leadership, metrics, and the craft of AI web-development.
Write acceptance criteria for failures, retries, and partial success
A field guide for PMs and engineers writing acceptance criteria that cover timeouts, duplicate submissions, stale data, partial success, and unknown outcomes -- with a reusable coverage matrix, a hypothetical import-flow walkthrough, and corrective actions for common gaps.
21 min read
Build a fair AI-assisted versus human-only delivery comparison
A cohort design for analysts comparing AI-assisted and human-only delivery that controls obvious confounding, measures throughput and quality together, and states clearly where causal claims remain unjustified.
19 min read
From customer requests to a defensible product problem statement
Customer requests are solutions in disguise. Here is a repeatable process -- with a worked hypothetical example, a reusable template, and failure-mode diagnostics -- for extracting falsifiable problem statements from a noisy request backlog.
18 min read
A risk-based approval matrix for autonomous coding agents
Agents that can propose, modify, and merge code need guardrails proportional to the blast radius of each change. Here is a concrete decision matrix that maps change impact and reversibility to approval requirements, without blocking routine work.
13 min read
You can’t improve what you don’t measure - especially with AI
AI can write a lot of code fast. Whether that code is any good is a different question - and the one that actually matters.
3 min read
Reclaiming the “feature factory”
“Feature factory” is meant as an insult. We think the problem was never the factory - it was the missing instruments.
3 min read
What DORA metrics still get right in the AI era
The four DORA metrics predate coding agents - and they hold up remarkably well as a foundation.
3 min read
Writing product plans your AI agents can actually build from
A vague product plan wastes a human sprint. Handed to an agent, it wastes tokens and ships the wrong thing faster.
3 min read
Reviewing AI-authored pull requests without slowing down
Agent-authored PRs are often large, plausible, and subtly wrong. Here’s how to review them fast without rubber-stamping.
3 min read