Member-only story
5 things Nolan Lawson’s 60-app agent benchmark says to React developers
React’s cost premium is smaller than the noise, the JavaScript it ships isn’t, and the framework skills barely got used.
Nolan Lawson had Opus 5.5 build the same Pokémon app five times in each of twelve JavaScript frameworks and published the bill: dollars, minutes, tokens, Lighthouse scores, lines of code. Three of the twelve are React frameworks.
I write React and TypeScript, run Claude Code most days, and review code for people with less experience than me, so I read it looking for the React rows.
1. The framework skills and MCP servers barely got used, and use didn’t predict cost
This is the table I’d put in front of anyone choosing a stack “because it’s better for agents,” and in a thread of 391 comments I couldn’t find anyone who talked about it.
Lawson installed each framework’s official AI tooling to level the field, then recorded whether the agents touched it.
Next.js shipped a skill and an MCP server, and the agents read the skill file in 1 of 5 runs and AGENTS.md in 2 of 5. TanStack Start’s agents found its skill files in 2 of 5…
