NS Nikolay Samokhvalov Nikolay Samokhvalov Founder of Postgres.ai, working on tools for safe schema changes and query optimization in Postgres.

“My recommendation usually was wait until the .1 or .2 version, but now I think the better recommendation is don't wait, start testing earlier. Most people have much more capacity thanks to AI. Start testing earlier and then jump to a new version sooner, with proper testing.”

Postgres FM · Hands-On Engineering Podcasts · November 2026

“My recommendation usually was wait until the .1 or .2 version, but now I think the better recommendation is don't wait, start testing earlier. Most people have much more capacity thanks to AI. Start testing earlier and then jump to a new version sooner, with proper testing.” — Nikolay Samokhvalov, Postgres FM

Context

Nikolay Samokhvalov reverses the old advice to let a new Postgres major version settle before adopting it. His reasoning: AI makes thorough upgrade testing cheap enough that teams can move sooner.

Transcript

Postgres FM Around 39:49 into the episode
Kirk Wolak

…So you're actually changing the underlying structure in a very interesting, and it's complicated, right? So this is why it's going to be a hot second before they get to DDL. But I tell you what, once they do, I think a lot of people are going to be really happy. That's a nice hurdle to overcome.

Speaker 4

Good. To cover our favorite features.

Nikolay Samokhvalov

I think recommendations should change as well because testing can be done more actively earlier because of AI, right? So what do you think? How would testing of production systems would look like now when we have AI and we can test thoroughly before actually switching it in production? On just clone, right? You test all your application in full clone environment, checking that everything works. As usual, you need just to like such environments should be detached from the world, not sending any emails, not making any API calls, right? Because if you test your application on actual copy, it can be dangerous. But it's outside of the database, but we like the plans will be the same and so on. Yeah, and

Michael Christofides

synthetic workloads are a bit easier to generate now. So I think one of the hardest things with these is actually having, you know, even if you test on a clone of production, if there's no traffic, you're going to miss a whole category of locking-related issues. That's a great point. It's actually

Nikolay Samokhvalov

the whole idea behind DBLAPForge is that now we can understanding some telemetry from like RCA, from root cost analysis from production problem, we can simulate with synthetic data purely, we can simulate, reproduce, and like this very problem and study it in detail in non-production environment.…

Kirk Wolak

So our approach has been in the past that we pull the full data, we put it in a dev environment first, and then we run the process through a dev environment with a combo of usually a dev person and an IT person that double checks everything that makes sure everything passes, follow up with some…

Speaker names come from our own diarization. The position is estimated from where the line sits in the episode.

More from Postgres FM