PostgreSQL 19 loses 8 to 10 features, and a developer points to AI code review
One of the developers whose work was pulled says AI review found in weeks the bugs that used to take years to surface.
PostgreSQL 19 will ship on November 19, 2026, a week after the release team had planned, and with roughly 8 to 10 big features pulled out of it. Jonathan Katz announced the new date for the release team on October 10, on the project's developer mailing list.
A day earlier, Peter Eisentraut, one of the developers whose features were pulled, wrote on his own blog about why so many went. One of his reasons is AI. LLM-assisted code review, he says, now finds in weeks the edge-case bugs that used to surface over five years.
- Release candidate 1
- October 15
- Release candidate 2
- November 12
- General availability
- November 19
- Features reverted
- about 8 to 10
What the release team said
Katz's message gives the backlog of fixes as the reason for the week. So many fixes have gone into older branches that the team wants them out in a second release candidate first:
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.
His message doesn't mention AI. That part comes from Eisentraut.
Why the features came out
Eisentraut says one revert per cycle is normal. This cycle had about ten, some late in beta. He lists several causes and starts with the plain one: maybe the code wasn't good enough. Some features had architectural defects that couldn't be fixed quickly during a beta.
Then there's the long tail. He describes it as a pile of small, mostly harmless edge-case bugs that review used to miss for years.
But with LLM-assisted code review, these kinds of things can get found much faster, and then you’re staring at a list of like forty defect reports.
Each fix might be three lines, he writes, but every report carries its own overhead, "and then you run out of time." Most of the reverted features were written before AI review was this useful, so the bar went up after the code was already in. He also says this isn't the result of AI-written code: "This code was written well before that was a possibility."
Security work squeezed the schedule too. He says many senior developers spent April to August fixing security bugs in what he calls the Vulnpocalypse, which is still going on, so the usual beta testing started late.
What went, and what stays
The best-known casualty is SQL/PGQ, the property graph query feature. Daniel Gustafsson confirmed on the general mailing list on October 7 that it "was reverted and will not be in PostgreSQL 19."
Eisentraut thinks LLM-assisted fuzzing should become part of building features, because nobody can list every combination of domains, partitions and triggers a new SQL feature has to survive by hand. He's upbeat on the release itself, and names concurrent REPACK, logical replication of sequences and planner hints as reasons to upgrade. Some features that missed 19 are already queued for PostgreSQL 20.
The second release candidate is due November 12, with the final release a week later.
More on Coding agents
- Anthropic moved its monorepo off git after daily merges rose from 2,500 to 30,000October 13, 2026
- Codex models and plans explained: GPT-6 Astra, Sol and Luna, and which plan gets whichSeptember 23, 2026
- Claude Code vs Codex vs Gemini CLI: which plan you need and which models you getSeptember 23, 2026
- Claude Code reads AGENTS.md as of version 2.1.277, but only when a project has no CLAUDE.mdSeptember 19, 2026

