Découvrez nos réflexions sur le logiciel, l'IA et l'avenir du développement.
Bjorn Krols2026-05-08Shard your E2E tests across runners, merge the reports
Playwright test sharding splits your E2E suite across multiple GitHub Actions runners with isolated databases per shard. Full parallelism, no shared state, one merged report.
Bjorn Krols2026-05-01Mining your review history to teach the agentic reviewer what your team actually cares about
We used the GitHub CLI to pull every review comment from our PR history, extracted the patterns reviewers kept repeating, and wrote them into AGENTS.md. The same file the agent reads when writing code now drives our agentic code review.
Bjorn Krols2026-04-27Making schema conventions a build failure, not a code review comment
We added a GitHub Actions workflow that spins up a real Postgres instance, applies migrations, and verifies naming conventions, indexes, and required fields before any PR can merge.
Bjorn Krols2026-03-27How we turned merge commits into customer-facing release notes
We built a pipeline that reads Linear tickets, git diffs, and commit messages, feeds them to an LLM, and delivers a formatted changelog to every PR and every stakeholder inbox, automatically.
Bjorn Krols2026-03-20Isolated workspaces with git worktrees, bare repos, and Worktrunk
Git worktrees let you work on multiple branches simultaneously without context switching. Combined with bare repos and Worktrunk, they become the infrastructure layer for parallel AI agents.
Bjorn Krols2026-03-12Using Chrome DevTools MCP to close the visual feedback loop
Chrome DevTools MCP gives coding agents direct access to a live browser. The agent navigates, inspects, screenshots, and fixes, without the developer switching windows.
Bjorn Krols2026-03-05How we automated visual proof into every PR
We replaced a manual video upload step with an automated pipeline that gives reviewers, agents, and customers a recording of every front-end change, automatically.
Bjorn Krols2026-02-26A plan for the agentic future
How we're shifting from writing code to orchestrating AI agents, and the five steps to get there.
Dites-nous ce que vous construisez. Nous vous dirons comment nous l'aborderions, ce qu'il faut, et à quelle vitesse nous pouvons avancer.
Nous vous dirons honnêtement si nous sommes le bon choix. Et si ce n'est pas le cas, nous vous orienterons vers quelqu'un qui l'est.