How We Ship 28 Updates a Day

We track how many changes we deploy each day. The current average is around 28. That's not a number we optimize for — it's a byproduct of a process designed around fast feedback and small, safe changes.

Here's how it works.

Small Changes, Not Big Releases

The traditional release model batches work into periodic releases. Features are developed over weeks, integrated, tested as a batch, and shipped together. The theory is that fewer, larger releases are less risky.

The practice is the opposite. Large batches mean large diffs. Large diffs mean hard-to-isolate problems. Hard-to-isolate problems mean longer debugging, more risk, and slower resolution.

We work in the opposite direction: changes are as small as possible. A bug fix ships the day it's found. A feature is split into the smallest independent increments that can each be deployed without breaking anything.

Automated Safety Nets

Shipping frequently is only safe if you have automated checks that catch problems before they reach users. Every change runs through automated tests — the 92+ we added this cycle, plus hundreds more — before deployment. Static analysis catches common issues. Integration checks verify the change works with the rest of the system.

If any check fails, the deployment stops. The change is fixed, the checks run again, and then it ships. No change reaches production with a failing check.

Fast Rollback

Sometimes something gets through that shouldn't. The answer isn't preventing all deployments — it's making rollback fast and automatic. An unexpected spike in errors triggers an automatic rollback within minutes. Problems are contained quickly; fixes are deployed just as quickly.

The Feedback Loop

Frequent deployments create a fast feedback loop. A user reports a problem today; a fix ships today or tomorrow. Feature feedback arrives and can be incorporated quickly. The gap between "we identified an improvement" and "users have the improvement" is measured in hours or days, not weeks.

What This Means for Users

28 deployments a day means the platform you use today is slightly better than the platform you used yesterday. Problems get fixed fast. Improvements compound quickly. The platform you use in a month is noticeably better than the one you use today, because small improvements arrive continuously rather than in infrequent batches.

Velocity isn't just about moving fast. It's about moving fast toward better.