Articles
The archive is getting rebuilt from scratch. What shipped before was mostly docs-paraphrase and LLM-template content. That's gone. What replaces it is a much smaller set of posts, each of which has to clear a four-question bar before I ship it.
psst, you found the easter egg.
Latest
Newest first
We Turned Simulation Off and Two Bugs Showed Up in the First Five Minutes
AI & LLMsA simulate flag hides real path bugs by construction. What broke the day we ran live checkout for real, why we moved from a global mock to role based simulation, and the two envelope rules that keep a recovery from painting a good result red.
No Object Generated: Response Did Not Match Schema
AI & LLMsNine unrelated features broke in one afternoon with the same error string, because generateObject asks the provider to enforce your schema and most providers quietly do not, plus the schema field that was pulling the model into hallucinating dates.
Self Preference Bias Is a 5 to 15 Point Swing, So Our Judge Is a Different Model Family
AI & LLMsTesting an agent whose output changes every run: deterministic structural assertions for the objective half, a cross family LLM judge for the rest, rubrics written as falsifiable claims, and a harness that admits what it did not grade.
Cannot Perform I/O on Behalf of a Different Request
DevOpsA supplier endpoint with a p90 of 15 seconds, four commits to put a timeout on it, one full revert in production, and the same illegal background I/O turning up again a layer down in the cache.
The Reattach Found an Empty Buffer Every Time in Production and Never Once on My Machine
DevOpsAn in process Map that buffered AI chat streams worked perfectly in dev and silently lost every stream in production, because Cloudflare Worker isolates do not share memory, plus the four bugs it took to persist a logged out visitor conversation.
A French Catalog That Still Says Save Passes the Type Check Perfectly
TypeScriptRecord<Locale, typeof en> makes shipping a half translated screen a compile error across 14 languages, and does nothing at all about a locale file that is structurally perfect and still in English.
The bar
4 questions
- 01
Did I pay for this knowledge?
Through an incident, migration, failed experiment, or shipping a product, not by reading a README.
- 02
Can I show numbers, diagrams, or code from real work?
Vendor benchmarks and stock illustrations don't count. Either I measured it or I don't mention it.
- 03
Am I committing to an opinion?
No "it depends" closers. If I can't pick a side, the post isn't ready.
- 04
Would a reader know more about me after reading?
Not just about the topic. About how I think, what I'd do in their shoes, which constraints actually matter.
If any answer is no, the draft is a tutorial in disguise and doesn't ship. It's how the last several hundred retired posts got retired.