AI Bug-Finders Are Delaying Postgres 19
Original: Postgres 19 is late because so many more bugs are being found now
Why This Matters
Shows how AI tooling is reshaping open-source release cadence at scale.
PostgreSQL 19 will miss its usual September–October window, likely releasing in November 2026. LLMs accelerated bug discovery, causing back-patch commits to jump 60%+ and forcing late reverts of three major features.
For over a decade, PostgreSQL major releases landed predictably between mid-September and mid-October. Not this year. PostgreSQL 19 is heading for a November release at the earliest, with postponement still on the table — and contributor Robert Haas, Sr. Staff Engineer at Databricks, places the blame squarely on LLMs.
The numbers are striking. Between feature freeze (April 8) and September 25, back-patch commits to the current stable branch jumped from 724 in 2025 to 1,169 in 2026 — a 61% increase. Commits patching the *prior* stable release rose 74% in the same window. Meanwhile, commits to the master branch barely budged.
Haas argues the codebase hasn't gotten buggier; it's just far easier to find existing bugs now. A wave of users are running PostgreSQL code through LLMs and filing reports on what comes back. Committers are swamped. LLMs can draft fixes, but human review is still required — and the human reviewer headcount hasn't scaled with the bug report volume.
The crunch had real consequences: three substantial features — UPDATE/DELETE FOR PORTION OF, SQL/PGQ, and ALTER TABLE MERGE/SPLIT PARTITION — were all reverted between August 27 and September 15, unusually late in the beta cycle. Review bandwidth that would normally have caught these issues earlier was consumed by back-branch bug fixes instead.