Things You Should Never Do, Revisited
Joel Spolsky told a generation of engineers to never rewrite from scratch. Twenty-six years later, with agents that can read and rewrite a codebase in days, does the advice still hold?
Every essay, in one place. Browse by tag or year, or search the full text - the URL holds your selection, so any view is shareable.
Joel Spolsky told a generation of engineers to never rewrite from scratch. Twenty-six years later, with agents that can read and rewrite a codebase in days, does the advice still hold?
Kimi K3's weights landed right on schedule - into a federal distillation accusation, an industry-wide open letter, and the first autonomous AI cyberattack. Part 3 of the K3 series looks at what came next.
I gave the same game-dev prompt to Kimi K3, GPT-5.6 Sol, Claude Sonnet 5, and Claude Fable 5 - same machine, same rules, native harnesses - then played what came back. Part 2 of the Kimi K3 series.
Kimi K3 launched last week to a flood of hot takes and hyperbole. What the benchmarks, the token economics, and a few identity probes actually say about the newest open-weight moment - part 1 of a three-part series.
A mandatory workshop in February put me back into hands-on engineering and changed how I work. These are the practices I've learned from countless hours with agents since - most of them useful well beyond coding.
In 1972 the HP-35 made the slide rule obsolete almost overnight, but math didn't die. AI is doing the same thing to software engineering, and the fundamentals are getting more valuable, not less.