A developer challenges the vibes-based hype around JavaScript frameworks by creating a benchmark to evaluate their true AI cost. Using a Pokémon app built with Next.js, Nuxt, SvelteKit, and others, the test measures development time, tokens, lines of code, and Lighthouse scores. The experiment uses Opus 5.5 agents with official MCP tools, avoiding Tailwind and over-specification to focus on framework performance.
Results show high stochasticity due to LLM randomness, but the goal is evidence-based framework selection in the agentic era.
Source: Hacker News · Summarized by HeadlinesBriefing