MONDAY, NOVEMBER 16, 2026 Tech news, minus the noise.
N News Cast
LATEST
Home  /  Dev
Dev

PostgreSQL 19 Delay: AI Found So Many Bugs the Release Slipped to November 19

PostgreSQL 19 was supposed to ship in the autumn. Instead it arrives November 19, and the project's own developers say the delay happened because AI tools found bugs faster than humans could fix them.

PostgreSQL 19 Delay: AI Found So Many Bugs the Release Slipped to November 19
Source: Martin Vorel, CC BY-SA 4.0, via Wikimedia Commons

PostgreSQL 19 Is Late, and AI Found the Bugs

PostgreSQL 19 will not ship on its usual schedule. The next major release of the world's most widely used open-source relational database is now planned for general availability on November 19, 2026, roughly eight weeks later in the year than usual. Release Candidate 1 arrived on October 15 as planned. A second release candidate is due today, November 12, the same day as the scheduled update releases for older branches.

The reason, according to longtime PostgreSQL committer Robert Haas, is blunt. In a blog post published October 21, he wrote that Postgres 19 is late because of AI. His one-sentence diagnosis: "The short answer is that LLMs happened." Large language models got very good at finding bugs in a codebase with forty years of them, and the reports arrived faster than the humans who fix them could absorb.

What the Commit Counts Show

Haas backed the claim with numbers. Between feature freeze on April 8 and September 25, commits to the release branch rose from 724 in 2025 to 1,169 in 2026, an increase of more than 60%. Commits to master rose from 1,037 to 1,443 in the same window, up 39%. Stable branches saw similar jumps. The project has not published an official count of AI-found bugs, but developer Peter Eisentraut has publicly sized the list of defect reports at roughly forty.

The damage shows in the release itself. Three major features were pulled between August 27 and September 15, late enough in the beta period that one revert would be unusual, let alone three: UPDATE and DELETE FOR PORTION OF, SQL Property Graph Queries, and ALTER TABLE MERGE and SPLIT PARTITION. The team's freed-up resources went to fixing stable branches instead of developing new functionality.

The official explanation differs slightly from Haas's. The RC 1 announcement ties the November date to today's minor update, which will include a large number of fixes the team wanted inside version 19 before declaring it final. Both explanations point the same way: the project found more real bugs than it could process on its old timetable, and it chose to slip the release rather than ship known problems.

Not Everyone Agrees AI Is the Cause

There is a genuine argument inside the project about this. Another committer, Tomas Vondra, went through the mailing list threads behind 12 reverts in the version 19 cycle and concluded that only 4 came out as "probably AI" or "AI". His point is that the AI story behind the reverted features is mostly wrong, and that the delay has more ordinary causes.

Both committers describe the same underlying fact: a jump in back-branch and security fixes. The disagreement is about attribution, not about the workload. Whether AI found the bugs or not, the humans are the bottleneck, and the bottleneck got worse this year.

The pattern is not unique to Postgres. Simon Willison, whose llm project is one of the most affected by AI-assisted pull requests, says he is not finding it to be a problem, and that newer models help him triage to spot the most useful fixes. Trail of Bits published work on backporting security fixes at scale with Codex, and found thousands of CVEs unfixed in stable kernel branches. The tools that find bugs are scaling faster than the people who fix them, and every mature open-source project is feeling it.

November 12 Is a Big Day for Postgres Users

Today matters for more than the release candidate. November 12 is also the final release for PostgreSQL 14. An AI agent's count of the git history shows PostgreSQL 14's August update fixed 29 CVEs, up from 3 a year earlier, and 135 more commits are queued. After today, none of those fixes will reach version 14. If you still run it, the project's own versioning policy says today is the end of the line.

The same day brings update releases for versions 14 through 18, and RC 2 for version 19, so version 19 gets those fixes before launch. The RC 1 announcement adds a condition worth remembering: the November 19 date holds unless an issue is discovered that requires a delay or an additional release candidate.

If you are considering version 19, test against RC 2 and keep in mind that its first release will lack the three features that were pulled. They may return in a later version, or they may not. The lesson of this cycle is that AI-assisted review is now part of how mature software gets made, and the projects that survive it will be the ones that build triage and review capacity to match. For more on how developer tools are changing, read our Python 3.15 upgrade guide and our VS Code agent control center coverage, plus how Perplexity open-sourced SPACE.

AS
News Cast Editorial Team

We cover AI, gadgets, startups and security — original reporting and explainers, written by humans, for humans. Got a tip? Reach us via the contact page.