Postgres 19 Slips Past October as Haas Counts 1,169 Stable-Branch Commits Against 724 a Year Ago
Software / news
Postgres 19 Slips Past October as Haas Counts 1,169 Stable-Branch Commits Against 724 a Year Ago
Robert Haas, a PostgreSQL committer, blames LLM-assisted bug hunting for the delay. Snowflake's count of reverted features and the postgresql.org schedule show how far the release has moved.
PostgreSQL 19 will most likely ship in November, not by mid-October, and committer Robert Haas says the cause is that large language models are finding bugs faster than people can fix them.
Haas, a Sr. Staff Software Engineer at Databricks and a PostgreSQL major contributor and committer, wrote the explanation on his blog on October 30. His post is titled "Postgres 19 is late, and it's AI's fault", and its first answer is blunt: "The short answer is that LLMs happened."
The postgresql.org roadmap page still describes the release as "planned for October 2026". The Beta 1 announcement of June 4 had said the final version was expected "around September/October 2026".
What Haas says changed since PostgreSQL 18
Haas wrote that the codebase has always held many bugs, and that "what's different now is that it's become vastly easier to find those bugs." Fixing them has become easier only to a lesser degree, he said.
He gave two reasons the schedule slipped. More bugs means more fixes to write and review, and LLM-drafted fixes are often wrong or do not meet project standards, so a human has to check and commit them. Second, review time that would have gone to the release went to security issues and back-branch bugs, which slowed the decision about which features to pull before shipping.
Haas also said he thinks committers were about as conservative as usual. In his reading, more mistakes were caught because LLMs surfaced them quickly.
The commit counts behind the claim
Haas counted commits between April 8, the feature freeze, and September 25 in each year. The figures below are his.
| Window | Branch | Commits |
|---|---|---|
| Apr 8 to Sep 25, 2025 | REL_18_STABLE | 724 |
| Apr 8 to Sep 25, 2026 | REL_19_STABLE | 1,169 |
| 2025 | REL_17_STABLE back-patches | 248 |
| 2026 | REL_18_STABLE back-patches | 433 |
He described the rise in back-patched commits as more than 60 percent, and the rise on the older stable branch as more than 74 percent. Those are Haas's own tallies, and the post links no query or script that a reader could rerun.
Which features came out between August 27 and September 15
Haas listed three features reverted late in the beta: UPDATE and DELETE with FOR PORTION OF, SQL property graph queries (SQL/PGQ), and ALTER TABLE ... MERGE PARTITION and SPLIT PARTITION. He put the reverts between August 27 and September 15.
The blog of Command Prompt, a PostgreSQL support company, gave dates for two of them. It reported that the partition merge and split work came out on August 27 with 14 commits, and SQL/PGQ on September 7 with 47 commits. The revert message cited "multiple design issues which are too late to address in this release cycle." The post names no author.
Elizabeth Garrett Christensen, a Developer Advocate at Snowflake, counted 53 reverts in the PG19 cycle against about 44 for PG18, in a September 16 post. Her list also covers GROUP BY ALL, the lz4 default for TOAST compression and online data checksums.
- PostgreSQL 1844 reverts
- PostgreSQL 1953 reverts
Source: Snowflake Engineering Blog, September 16, 2026, accessed 2026-11-02
Christensen said the 44 for PG18 is approximate and that most PG18 reverts came shortly after its beta. She also wrote that a delayed major release "has not happened with Postgres in a long time", and that the August 2026 PG18 update fixed 28 CVEs where the project used to average a couple per release.
See our earlier coverage of Jepsen's 96 findings against ParticleDB for another database whose release quality came under scrutiny this month.
What is still unknown
Neither the roadmap page nor the Haas post gives a release candidate date for 19. Haas wrote that the project could delay further and called the slip "not particularly concerning", a reflection of its focus on quality. Command Prompt told readers to plan upgrades against PostgreSQL 18, and said PostgreSQL 14 reaches end of life on November 12. A separate look at Python 3.15 shows a language project holding a long-planned feature for five years. Postgres has not said which November date it is aiming for.
Sources
More in Software
- 01Polyend Publishes Tracker Firmware Source Under a Noncommercial Licence, With Two Binary Libraries Left OutThe tracker-ce repository covers the original Tracker, Tracker+ and Tracker Mini, but the synth and DSP libraries ship precompiled under a separate proprietary licence.
- 02Claude-Mem's Installer Uploaded 30 Days of Prompts to cmem.ai Before Asking; v15.8.3 Adds a Consent PanelThe fix shipped on Nov. 2, but the release notes say prompts already uploaded by earlier versions are not removed, and the GitHub issue was reopened the same day.
- 03Vercel Labs' Agent-Desktop 1.0 Ships With a macOS Bug That Types Into the Frontmost AppThe Rust tool lets an AI agent drive desktop apps through the accessibility tree, and 48 open issues sit against it three days after the 1.0.2 release.
- 04T3 Code Reaches 38.1k Stars in Alpha, and Its Remote Access Docs Say to Treat Pairing Links as PasswordsThe MIT-licensed control surface for Claude Code, Codex and other agents has nearly doubled its star count since August 27. Its own documentation lists what stays risky.