How I hit 95+ mobile PageSpeed on a data-heavy climate tool using AI pair-programming
Hey makers! 👋
​I’m launching a climate data tool tomorrow (Atlas do Clima) focused on historical rainfall norms in Brazil. One thing that surprised me during the process: we managed to hit 95 Performance (mobile), 94 Accessibility, 100 Best Practices, and 100 SEO on Google PageSpeed Insights.
​The catch? I didn't spend months micro-optimizing code manually.
​Whenever PageSpeed flagged a bottleneck (render-blocking scripts, unoptimized assets, or DOM overhead), I fed the diagnostics directly into Gemini, iterated on the suggestions, and refined the scripts until the scores turned green.
​For indie builders working on data-heavy apps:
​How are you using LLMs in your optimization workflow?
​Do you still manually audit your Core Web Vitals, or do you rely on AI to find the quick wins?
​Would love to hear how other non-traditional devs are approaching front-end performance!
Replies