One Guy's Benchmarks Made Cloudflare Workers 3x Faster for Everyone

The Cloudflare Workers CPU thing is being talked about like a performance release, a changelog entry, a thing the team planned. It was not that.

Theo posted benchmarks. The numbers looked bad. Cloudflare's team, presumably mildly horrified, went and actually looked — and what they found was not one big architectural failure, not a single smoking gun, but a pile of small stupid things. Little inefficiencies that had accumulated like dishes in a sink. Each one ignorable in isolation, collectively embarrassing when someone with a big audience pointed a measuring instrument at them.

They fixed the pile.

Workers got three times faster. For everyone. Not for Theo, not for the accounts that complained, not for enterprise customers with the right tier — for everyone running anything on the platform, retroactively, because one guy on Twitter ran some benchmarks and the right people saw them.

This is actually how infrastructure improves, most of the time. Not through roadmaps. Through a sufficiently public moment of "huh, that's slow" that someone can't ignore. The Cloudflare team deserves credit for actually digging in rather than writing a blog post about their commitment to performance — but the whole chain of causation started with someone outside the building noticing something the people inside had stopped seeing.

The underrated part is the "bunch of little things." A single catastrophic bug is easy to find and easy to fix. A dozen small inefficiencies that compound into a 3x slowdown are exactly the kind of thing that survives years of internal review, because each one is technically fine. You need the outside eye.

Three times faster. From a tweet. File that under things that shouldn't work but do.