Postgres 19 delayed to November 19: what the release team said, and what AI code review had to do with it

PostgreSQL 19 is delayed. On October 10, 2026, the PostgreSQL release team moved the database's general availability (GA) from October 29 to November 19, 2026, and added a second release candidate for November 12. The Register reported the Postgres 19 delay on October 16 under the headline "Postgres 19 delayed after AI reviewers find more bugs than humans can fix." The official announcement doesn't mention AI. The AI connection comes from a PostgreSQL committer who says LLM-assisted code review, one of the coding-agent era's new tools, surfaced defects late in the cycle faster than developers could fix them. Both things can be true, but they aren't the same claim.

This story is based on the PostgreSQL release team's messages to the pgsql-hackers mailing list, the PostgreSQL project's Beta 4 and RC 1 announcements and roadmap, Peter Eisentraut's October 9 blog post, and The Register's coverage, checked October 18, 2026.

The short version

• New date: PostgreSQL 19 GA is scheduled for November 19, 2026, moved from October 29.
• Why, officially: "the volume of fixes that have been added to the back branches." The release team wants those fixes in an RC2 on November 12 first.
• Features pulled: about 8 to 10 significant features were reverted this cycle, by one committer's count. They include SQL/PGQ property graph queries and online enabling of data checksums.
• The AI part: committer Peter Eisentraut says LLM-assisted review means issues "get found much faster," producing long lists of small defects that take time to fix. He names no tools and gives no total count.
• What's not established: that AI reviewers caused the November slip, or how many bugs AI found.

When is PostgreSQL 19 coming out?

The current plan is November 19, 2026. The schedule changed twice in the past month:

Last verified: October 18, 2026

The October 10 message, titled "Updated PostgreSQL RC / GA dates," puts GA "one week after the planned scheduled update release," the project's regular minor-release date. The September 28 message explained that the project tries to ship a major version a couple of weeks before a November update release, so fixes can follow quickly. The PostgreSQL roadmap page still says the 19 release "is planned for October 2026."

The release team's full explanation: "Given the volume of fixes that have been added to the back branches, the release team wanted to make those available in a PostgreSQL 19 RC2 that would be scheduled for November 12 before proceeding with the GA. We know this has been an atypical major release cycle, but we want to ensure we keep delivering a PostgreSQL that's stable and reliable, and we've made these decisions around that goal."

The RC 1 announcement gives the same reason in other words: the November 12 update release "will include a large number of fixes," so the release team decided to include them in PostgreSQL 19 before GA.

Which features were cut from PostgreSQL 19?

The Beta 4 announcement on September 24 listed these reverts and removals:

• SQL/PGQ (property graph query) support
• online enabling and disabling of data checksums
• temporal updates and deletes with the FOR PORTION OF clause
• ALTER TABLE ... MERGE PARTITIONS and SPLIT PARTITIONS
• the pg_get_role_ddl(), pg_get_tablespace_ddl() and pg_get_database_ddl() functions
• a change that forced LC_COLLATE to C in the postmaster process

SQL/PGQ was the most visible. The Register reported on September 15 that developers had pulled it, and quoted PostgreSQL contributor Tom Lane warning that shipping it in v19 risked "post-release bug discoveries that are unfixable until v20."

Features still in the release include REPACK (CONCURRENTLY), which the PostgreSQL 19 release notes say reclaims disk space without blocking reads and writes to the table, pg_plan_advice and parallel autovacuum. The RC 1 announcement lists several fixes for each since Beta 4.

What did AI code review have to do with it?

The AI connection comes from Peter Eisentraut, a long-time PostgreSQL committer, in an October 9 post titled "The reverts will continue until morale improves." He was personally involved in some reverted features. His account:

• The count. "About 8 to 10 significant features, depending on how you count, have been reverted," which he calls "an unusually high number." Maybe one per cycle could be expected.
• LLM-assisted review. Faster review means issues "get found much faster." A feature can end up facing a long tail of small defects late in the beta. In his example, a developer ends up "staring at a list of like forty defect reports," each needing "only a three-line fix on average" but carrying a fixed overhead. That is an illustration, not a count of PostgreSQL 19 bugs.
• A higher bar. Most of the affected code was written before LLM review was useful, so "the effective quality bar has been raised after the time the feature was written and committed."
• Not vibe coding. He says that "at least for the most part, as far as I can tell, this situation is not due to 'vibe coding'," because the code was written before that was possible.
• Other causes. He lists code that "was just not good enough," features that were too hard to finish in time, architectural defects, and what he calls the "Vulnpocalypse": senior developers busy with security fixes from April to August, which cut beta-period review. He says the cause differs by feature.

Eisentraut doesn't say which LLM tools were used, who ran them, or how many defects they found. He says he is confident that "many of the reverted features will come back before too long," and that PostgreSQL 19 "will be very robust, after all the additional scrutiny."

What do the headlines overstate?

• "AI reviewers find more bugs than humans can fix." That's The Register's framing of Eisentraut's argument. It's a fair summary of his view of the reverts, but the release team's delay notice cites back-branch fix volume and doesn't mention AI.
• "AI review finds 40 bugs." Some coverage turned Eisentraut's illustrative example into a count. The PostgreSQL project hasn't published an AI-found bug total.
• The reverts and the delay are separate events. Most reverts were announced with Beta 4 on September 24, before the October 29 GA date was even set. The November slip came later.
• "Delayed" relative to what. The project had said October. GA is now November 19, three weeks after the October 29 date.

Why it matters for developers

PostgreSQL is one of the most scrutinized open-source codebases, and, by one committer's account, this is a concrete case of AI-assisted review changing a release, not just a pull request. It shows a cost of better automated review that vendors rarely mention: findings arrive faster than maintainers can triage and fix them. Anthropic's Claude Code team describes a related problem in its own process, where AI reviewers kept re-flagging old issues; see how the Claude Code team does code review.

The case also complicates the debate over AI in open source. Some projects ban AI-assisted contributions, while vendors offer maintainers free AI scanning, such as Anthropic's OSS Scanner and GitHub's AI Scan for pull requests. Eisentraut's point is that AI review raised the bar for code written before such tools existed.

If you run PostgreSQL, the practical change is the date: plan PostgreSQL 19 upgrades around a November 19 GA, and check the release notes for the features that moved to PostgreSQL 20.

What's confirmed, reported and uncertain

• Confirmed (PostgreSQL project):
  • GA moved from October 29 to November 19, 2026 (pgsql-hackers, October 10)
  • RC1 released October 15; RC2 planned for November 12
  • the reverted features listed in the Beta 4 announcement
  • the release team's stated reason: back-branch fix volume
• Reported (Eisentraut's blog and The Register):
  • about 8 to 10 significant features reverted
  • LLM-assisted review surfacing defects faster than they could be fixed
  • the "Vulnpocalypse" workload from April to August
• Uncertain:
  • which AI tools were used, and by whom
  • how many defects AI review found in PostgreSQL 19
  • whether November 19 holds; the RC 1 announcement leaves room for another delay or RC