Codebases grow over time, and as they do, they tend to get worse to work with. This usually isn't anyone's particular fault. People come and go, features get shipped, dependencies get added. It feels like a death by a thousand cuts.
At Sentry, we noticed this too: Long CI pipelines, hundreds of components, bad abstractions, huge dependency graphs and import cycles. The codebase is 10+ years old! And as AI comes along and makes adding new code trivial, all of a sudden all the symptoms amplify.
In this talk, I will showcase the countermeasures we're applying at Sentry to stop this from getting worse. We'll look at how we can make it easier to do the right thing in a large codebase, how fast feedback loops fit into that, and why humans and agents might not be so different when it comes to what they need to work effectively…
This talk has been presented at React Advanced 2026, check out the latest edition of this React Conference.


























