AI made bugs much easier to find, but fixing them still takes humans
by Jonathan Katz and Robert Haas
PostgreSQL 19 is now generally available, and there are some really cool things in it:
We walked you through what's new on PostgreSQL 19 here. In this post, we take a step back; as PostgreSQL contributors, we want to share some backstory on how a major release comes together, and some reflections on PostgreSQL 19 and what made this release different.
Over the past decade, the timing of PostgreSQL major releases has become very predictable: no earlier than the middle of September, and no later than the middle of October. This year was different: PostgreSQL 19 shipped more than a month after that window closed. So, what happened?
To answer that, it helps to know what a normal year looks like.
PostgreSQL started in the 1980s as the POSTGRES project at UC Berkeley, and it has been developed in the open since 1996. No single company owns it. A worldwide community of developers builds it, and that community values reliability and predictability. Part of that predictability is the calendar: one major release a year, at about the same time each year.
Each release takes 14 to 15 months from the first commits to general availability, and that time splits into two phases:
When a feature can't be stabilized in time, the usual practice is to defer it to a future major version and keep the release on schedule.
All software contains bugs, and PostgreSQL's code base is large, complex, and old. Our git history – converted from CVS – now reaches back more than 30 years, and the first code was written a decade before that. There's a huge amount of code, and that code contains a huge number of bugs.
That's always been true, and it hasn't stopped PostgreSQL from being popular or successful. What's different now is that it's become vastly easier to find those bugs. Fixing them has also gotten easier, but not to the same degree.
The change came quickly. Around February 2026, developers started noticing that the latest models were good enough to write real code, that the code going to production wasn't as buggy, and that the models were helping them move faster.
Reviewing code and researching security issues used to take a lot of human expertise. With AI, they take much less, and how much AI can discover on its own has gone up sharply.
You can see the result in PostgreSQL's quarterly update releases. From August 2023 through November 2025, each one fixed between one and four CVEs. In 2026, that went to 5 in February, 11 in May, and 28 in August. Bug fixes per update release followed the same curve, from roughly 50 to 75 per release to about 140 in August 2026.
Not every CVE is equally severe though. Many of the bug fixes are long-tail issues. But the volume is still higher than we’re used to.
The commit history shows the same thing. Between April 8th, 2025 and September 25, 2025, there were 1037 commits to PostgreSQL's master branch and 724 commits to REL_18_STABLE. In 2026, during the same date range, there were 1443 commits to the master branch and 1169 commits to REL_19_STABLE. In other words, if you look at the number of commits that went into the master branch only between feature freeze and September 25th, it actually declined slightly from last year to this year, but the number of back-patched commits jumped by more than 60%.
The commit history shows the same thing. Between April 8th, 2025 and September 25, 2025, there were 1037 commits to PostgreSQL's master branch and 724 commits to REL_18_STABLE. In 2026, during the same date range, there were 1443 commits to the master branch and 1169 commits to REL_19_STABLE. In other words, if you look at the number of commits that went into the master branch only between feature freeze and September 25th, it actually declined slightly from last year to this year, but the number of back-patched commits jumped by more than 60%.
It's possible that some of this jump is the result of committers making more mistakes this year than they did last year, but that’s unlikely. Rather, it seems likely that PostgreSQL committers were about as conservative in 2026 as they were in preceding years, but a lot more of the mistakes they made were quickly found by LLMs.
And it isn't just mistakes in the latest release. In the same time period, during 2025, there were 248 commits to REL_17_STABLE, and during 2026, there were 433 commits to REL_18_STABLE. In other words, the number of commits being back-patched not just to the release in progress but also to at least one prior release jumped by more than 74% in a single year.
The reason for this is simply that a huge number of people on the Internet are now reporting bugs in PostgreSQL that they found using LLMs, and the PostgreSQL committers have been very busy fixing them.
Time that would have gone into reviewing what was going into PostgreSQL 19 went into fixing security issues and other bugs in the back branches instead. A few features that were in progress this cycle weren't ready in time, so the project did what it always does and deferred them to a future release.
The fixes kept coming right up to the end. The November 12 update release for the supported branches included so many fixes that the release team put out a second release candidate so PostgreSQL 19 could ship with them.
In the end, we got PostgreSQL 19 out the door, a few weeks later than usual, and the delay came directly from the project's concern for quality. The way we all develop software has changed, and the PostgreSQL community is adapting.
Nobody knows yet exactly how open source software development will change, but contributors across the PostgreSQL project are working hard to figure it out. LLMs are going to push every PostgreSQL developer, and to some extent every software developer, to raise their game. Every feature we write from now on will be scrutinized by AI, and bugs that would have gone unnoticed forever in an earlier era will now be found, sometimes in a matter of hours.
That's a big challenge, and an even bigger opportunity to make PostgreSQL better than it's ever been. We're excited to be part of it.
Subscribe to our blog and get the latest posts delivered to your inbox.