PostgreSQL 19 delayed to November after AI code review finds more bugs than devs can fix

Release pushed to November

PostgreSQL 19, the next major version of the widely used open source database, will now become generally available around November 19 instead of the late September or early October window originally planned. According to The Register, the PostgreSQL Global Development Group has also reverted between eight and ten significant features, postponing them to PostgreSQL 20.

The cause is unusual. Peter Eisentraut, a leading PostgreSQL committer and vice president at EDB, a PostgreSQL support and consultancy, wrote in a blog post that LLM-assisted code review bears at least part of the responsibility.

More findings than fixing capacity

Eisentraut split the problems into two groups. Some of the reverted features had architectural flaws serious enough that they could not have been patched within a beta cycle. The bigger drag, though, was a large volume of small, mostly harmless defects around edge cases — the kind of problems that in earlier eras would likely have gone unnoticed until well after launch, then been resolved through bug fixes spread over the following five years.

AI-assisted review finds such issues far sooner. Eisentraut described confronting a list of forty defect reports: even when the average fix is only about three lines, each report carries fixed overhead, and eventually the calendar runs out.

Timing made it worse. The major PostgreSQL 19 features were written before LLM-based review had matured into its current usefulness; in the months before launch the tooling improved, meaning the effective standard the code was held to tightened after that code had already been committed. Eisentraut suggested that if LLM review capability keeps improving at its present rate, future development cycles should adjust and the situation will resolve itself.

He was also explicit that the reverts were not a symptom of "vibe coding," since that approach was not available when most of the new features were written.

What was cut and what remains

The most visible casualty is support for SQL Property Graph Queries (SQL/PGQ), syntax added to the SQL standard in 2023 for exploring relationships between nodes connected by edges. PostgreSQL's implementation was removed over unresolved issues. Long-time contributor Tom Lane had warned of the danger: "At this point I'd be willing to bet dinner that if we ship it in v19 there will be post-release bug discoveries that are unfixable until v20."

PostgreSQL 19 will still include REPACK CONCURRENTLY, an operation that reclaims disk space without the table locking that VACUUM FULL requires. Eisentraut promised the release will remain a strong one and, after the additional scrutiny, unusually robust.

The reverted features are expected to reappear in PostgreSQL 20 — a reversal of expectations from just a few months ago, when developers anticipated that release would contain fixes but no new features.

Why it matters

This episode is a concrete data point in the debate over AI and software quality, and it inverts the usual narrative. The AI tooling did not create the bugs; it surfaced a hidden backlog faster than one of the most mature open source projects in the world could process it. For any team adopting AI review, the lesson is that accelerating detection without matching capacity for triage and repair simply moves the bottleneck. For PostgreSQL users, the delay is a conservative trade — a slightly later release carrying fewer latent defects, and a PostgreSQL 20 that now looks likely to be feature-heavy rather than quiet.