#660 — August 5, 2026
Web Version
Lots of videos from this year's PGConf.dev are now on YouTube and I've rounded up some highlights at the end of the issue!__Your editor, Peter Cooper
Postgres Weekly
COUNT(DISTINCT) Too Slow? A Guide to Fast Approximations — A thorough, hands-on look at Postgres's powers of approximation and sampling, covering the pros, cons, and performance profiles of approaches including TABLESAMPLE, hll, and Apache DataSketches, on a 10 million row test table.
Elizabeth Garrett Christensen (Snowflake)
Model the Cost of an Oracle-To-Postgres Move — Weighing a move off Oracle? EDB's free calculator lets you model the cost by database tier and support level. Pressure-test the migration in a few minutes, before you commit to anything.
EDB sponsor
SQL Improvements in Postgres 11–18: A Personal Selection — Eight releases' worth of SQL features, picked by a longtime contributor and grouped by topic. The part on modelling band membership over time with tstzmultirange and range_agg() alone is worth a look, but there's much to enjoy.
Dimitri Fontaine
Unveiling a 13-Year-Old Postgres Bug in Cascading Replication — A guard added to streaming replication in Postgres 9.3 could lock a cascading standby out of streaming from its upstream after falling back to archive recovery. It's been fixed and will land in the next minor releases: 18.5, 17.11, and so on.
Gabriele Bartolini (EDB)
🤖 'Turning Claude into Postgres So I Can Raise a Series A' — What happens when you stick an LLM behind the Postgres wire protocol and just let it deal with queries however it wants? This, apparently. Yes, it's a bit of fun, but he really did implement actual storage APIs for it to use! 😅
Jacob Jackson
Waiting for Postgres 19: SQL Property Graph Queries (SQL/PGQ) — If you've wondered what SQL property graph queries buy you over just using normal joins, you're not alone!
Hubert depesz Lubaczewski
📄 Andy Pavlo Joins ClickHouse to Establish ClickHouse Labs – "One larger question we will investigate is how DBMSs like ClickHouse and PostgreSQL fit into emerging AI and agentic technologies." Andy Pavlo (ClickHouse)
📄 Hybrid Search Patterns with Postgres and pgvector – What iterative index scans bring to the party. Christopher Winslett
📄 TDE (Transparent Data Encryption) Performance in Postgres Zsolt Parragi (Percona)
📺 THE PGCONF.DEV VIDEOS:
PGConf.dev 2026 took place in Vancouver this May, and there are a whopping 47 videos in this YouTube playlist to enjoy, if you didn't attend. Some highlights:
▶️ 30 Years of PostgreSQL Retrospective with Momjian, Wieck, Chen, Lockhart, Lane and Mikheev – A panel (above) of Postgres pioneers revisits technical decisions, defining moments, and lessons from over the years. Moderated by Melanie Plageman and Jonathan Katz.
▶️ What's Missing in Postgres? with Bruce Momjian – A look at 'missing' features including sharding, TDE, global indexes, and multi-master replication. The slides are available, if you prefer.
▶️ Building the Next Generation of Postgres Contributors with Claire Giordano – The Talking Postgres host reflects on what it takes to get more people involved with the project.
▶️ Real-Time Postgres Patch Idea Evaluation with Freund, Linnakangas, Lane, and Haas – Leading committers evaluate 'surprise patch ideas' live, with Robert Haas moderating.
▶️ Why is Postgres Terrible? with Christophe Pettus – First admitting Postgres isn't terrible, Christophe nonetheless examines real-world pain points, mitigations, and opportunities for improvement.
▶️ Profiling Postgres Perils with Andres Freund – A tour of misleading benchmarks, profiling traps, and counterintuitive performance results.
Alternatively, check out the playlist for anything that catches your eye – there's a lot more than just the above.
For a few weeks I ran a proper multi-agent orchestrator. The concept was right, and I still think the people building those tools are pointed the correct way. But the bill was absurd, and it took me a while to work out why.
It wasn't the coding. It was the talking about the coding.
Every turn, age
You sit down Friday with one messy GitHub issue. You want a usable plan before Monday morning. You paste the text into a chat agent.
The reply looks polished, complete, and very sure. It adds Kafka, Redis, and a new auth service. The issue never named those systems.
You do not have a brownfield bu
Il problema: SQL non e uno standard unico
Sulla carta, SQL e uno standard. Nella pratica, ogni database parla il suo dialetto. MySQL usa i backtick per quotare gli identificatori, PostgreSQL le virgolette doppie, SQLite le accetta entrambe ma preferisce le virgolette. MySQL ha AUTO_INCREMENT, Post