Two hundred pieces in
What writing in public taught me about engineering, what I was most wrong about, and why the archive is the point.
#craft
59 pieces tagged craft,
from March 7, 2025 to August 3, 2026.
Two years of agentic coding: what stuck
The workflows that survived contact with real work, the ones that did not, and what the whole thing actually changed.
Deleting code is the highest-value work nobody schedules
Every line you remove is one that cannot break, cannot be misread, and does not need to be maintained.
The performance budget
A number, agreed in advance, that fails the build. The only mechanism that has ever kept a system fast.
What I look for in a codebase in the first hour
A checklist for assessing an unfamiliar codebase quickly — for a job, a due diligence, or a project you inherited.
The unreasonable effectiveness of a changelog
A file that takes ten minutes per release and answers most of the questions your users would otherwise ask you.
Property-based testing deserves its moment
State an invariant, let the machine find the counterexample. A twenty-year-old technique that fits this moment exactly.
The onboarding document that actually works
Most onboarding docs are written by people who already know. Here is what a new engineer needs on day one, in order.
Secrets management, practically
Not a survey of vaults. The specific practices that actually reduce risk, in order of what to do first.
Code review comments that change things
Most review comments are noise or nitpicks. A small taxonomy of the ones that are worth writing.
Configuration is the most under-designed part of your system
It has no type system, no tests, no review, and it causes a disproportionate share of outages.
Randomness, and the bugs you cannot reproduce
The bug that happens once a week and never in staging. A systematic approach to the class of problem everyone handles badly.
everything else tagged #craft 47
- What a staff engineer does all dayessay
- Distributed tracing that people actually useessay
- Refactoring under an agentessay
- Static analysis is finally worth the false positivesessay
- Time zones, and why your calendar code is wrongessay
- Rewrites: when they actually workessay
- What a great bug report containsessay
- Model evaluation for people who shipessay
- Idempotency is the only distributed systems concept you needessay
- Compression is underratedessay
- Run the incident before the incidentessay
- Error messages are a user interfaceessay
- Naming things: the actual rulesessay
- Search is hard and you should probably not build itessay
- Migrations that don't wake anyone upessay
- The three kinds of technical debtessay
- Accessibility is a testing problemessay
- The API versioning decisionessay
- Observability costs more than the thing it observesessay
- The joke RFCs are the best documentationessay
- Why your build is slowessay
- Design docs people actually readessay
- The cost of a meeting, in engineering termsessay
- Schema design is the only design that lastsessay
- GDC and the engineering nobody outside games learns fromessay
- Feature flags and the state space nobody testsessay
- The engineer's guide to saying noessay
- On-call is a design problemessay
- Documentation is a product and you should staff it like oneessay
- Type systems and the cost of being rightessay
- The dependency budgetessay
- How to read a benchmarkessay
- Your monitoring is measuring the wrong ninesessay
- Reading code is a skill and nobody teaches itessay
- Verification is the whole job nowessay
- What I got wrong in 2025essay
- The postmortem that changes somethingessay
- Advent of Code and what a puzzle is foressay
- The technical interview is measuring the wrong thing, againessay
- Estimation is a communication problemessay
- The pull request is the wrong unit of reviewessay
- Caching is the only optimization that reliably worksessay
- Write the README firstessay
- Your logs are a product and you are shipping a bad oneessay
- The test suite is a design documentessay
- Vibe coding is real, and it is not what you thinkessay
- The eight git commands worth actually learningessay