Nolan Lawson has been telling developers to use the platform for years. Now he's writing the best version of the counterargument, and the result is more honest and more useful than the original mantra ever was. The essay works because it refuses to flatten a genuinely complex behavior into laziness or ignorance. The historical case is airtight: browsers spent a decade trailing the ecosystem built on top of them. jQuery wasn't a crutch — it was load-bearing infrastructure while IE6 slowly died. Developers who learned to roll their own during that era didn't suddenly rewire their instincts when evergreen browsers arrived. Muscle memory is real, and the web platform's reliability was earned late. But the essay's sharpest move is identifying the IKEA effect in code. Building a modal dialog from scratch — position:absolute, z-index hacking, focus traps, Esc handling — is objectively worse than using . It's also more fun, more educational, and produces something the developer feels ownership over. Lawson names this without condescension: the same people who now sit in W3C meetings got there by building polyfills and shims that shouldn't have needed to exist. The path to platform expertise runs through platform ignorance. The ClickHouse anecdote is the essay's structural proof-of-concept. Two experienced developers independently built slower, clunkier versions of what the platform already did — one via compression, one via a separate key-value store — because neither had read the docs thoroughly enough. The platform won by doing less. This pattern recurs across iOS, Android, game engines, and every other layer developers build on top of. The documentation gap matters more than Lawson explicitly weights it. Dragula's beautiful landing page versus MDN's dry reference for the Drag and Drop API isn't just an aesthetic comparison — it's a funnel problem. Developers discover solutions through search, and for years the top results pointed to libraries, not platform APIs. The ecosystem's marketing was better than the platform's. The AI coda is the weakest section — Lawson admits as much — but the framing is right. LLMs could be the ultimate 'grizzled senior engineer,' replacing baroque custom code with single platform calls. Or they could be the ultimate IKEA-effect amplifier, generating mountains of non-idiomatic code that no one feels ownership over and therefore no one audits. Both outcomes are already happening. The question is which scales faster. What makes this essay land is its refusal to pick a winner. 'Use the platform' remains correct as engineering advice. It remains insufficient as a theory of developer behavior. Lawson holds both without collapsing into either, and that's harder than it looks.
