THE RAVIT SHOW  //  POSTGRES EDITION

POSETTE 2026 Wrap: The event that quietly redefined how we talk about Postgres and AI

Event dates: June 16 to 18, 2026  //  Format: virtual, free  //  Organized by the Postgres team at Microsoft, in partnership with AMD

POSETTE 2026 is done. June 16 to 18. Three days. 44 talks. 50 speakers. Four livestreams. All free. All online. All now sitting on YouTube with captions in 16 languages. This is brought to you by Microsoft in partnership with AMD!!!!

And if you missed it, you missed one of the most technically dense Postgres events of the year.

Not because of hype. Because of what emerged across the sessions. This year the conversation shifted. Postgres stopped being framed as the reliable relational database and started being framed as the retrieval backbone of production AI systems. That is a big shift, and POSETTE 2026 is where it happened in the open.

Here is the wrap.

Five years in, POSETTE has grown into the single most concentrated Postgres event on the calendar. Free to attend. Free to rewatch. No registration wall.

Delivered by Affan Dar (VP Engineering for Postgres at Microsoft) and Charles Feddersen (Director of Product for Postgres). This is the state of the union. How Microsoft is contributing upstream to the open source project, and how those contributions land in Azure Database for PostgreSQL and Azure HorizonDB. If you want the strategic picture of where cloud Postgres is going, start here.

A live panel with Heikki Linnakangas, Melanie Plageman, Thomas Munro, and Alvaro Herrera. Four of the people actually shipping Postgres code, on camera, talking about what made it into Postgres 19 and what did not. This is the closest you get to sitting in a hackers meeting without being one.

The strongest signal across the event. RAG pipelines, pgvector, agent memory, hybrid retrieval. Multiple talks made the case that when you care about accuracy, freshness, and traceability, you want your vector search living next to your structured records, not in a separate system that has to be synced.

Pamela Fox's work on Model Context Protocol surfaces a design question most teams are ignoring: when you expose Postgres as a set of tools to an LLM, the shape of those tools decides whether the agent behaves well. This is the layer between agents and data that almost nobody has named yet, and POSETTE put it on the map.

Temporal keys. NOT ENFORCED constraints. NOT NULL promoted to a first-class constraint. Better constraint handling for partitioned tables. Plus meaningful upgrades in the WAL path. The 17 to 18 to 19 progression is delivering more application-level value per release than at any point in recent memory.

The talk on building safety tooling for risk-free AI tuning of Postgres captured it best: fast cars need fast brakes. As agents start to write and issue queries dynamically, the operational surface grows. What stops them from a destructive statement. What stops them from hammering a hot index. This is the next frontier, and POSETTE started the conversation.

If you only have an afternoon, this is the order to watch:

1. Driving Postgres forward at Microsoft. Keynote 1 for the strategic view.

2. Postgres 19 Hackers Panel. Keynote 2 for the technical roadmap.

3. From Queries to Agents: The Next Era of Data Retrieval on PostgreSQL. The clearest articulation of the AI retrieval thesis.

4. Production RAG at Scale with Azure Database for PostgreSQL. The production playbook.

5. PostgreSQL 17 vs 18: Side-by-Side Performance Wins in Real-World Queries. The upgrade math, benchmarked.

6. Building safety tooling for risk-free AI tuning of Postgres. The guardrails talk everyone will be quoting in six months.

Bonus picks for developers: Chun Lin Goh's session on using Postgres as your message queue is a genuinely useful 'you probably don't need Kafka for that' reframe. Josef Machytka on Apache AGE for graph queries inside Postgres is another sleeper.

If you take one thing from POSETTE 2026, take this: Postgres is no longer just the database at the end of a request. It is becoming the negotiated layer between AI systems and your data. That shift is not marketing. It is happening in production, at real companies, and this event captured it before the rest of the industry caught up.

Watch the talks. Then plan your next architecture accordingly.

See you at POSETTE 2027.

The Ravit Show is a Data and AI media platform with 1M plus community members and 137K plus newsletter subscribers. If this wrap was useful, forward it to one person who works with Postgres. That is how the community grows.

Follow The Ravit Show on LinkedIn. Subscribe to the newsletter for weekly Data and AI briefings.

The brands winning on Amazon stopped guessing. They know which external channels pull weight — and which ones don't. Levanta's free playbook breaks down 7 strategies driving millions in off-Amazon revenue.

→ Grab the FREE playbook

🔍 Stay Ahead in AI & Data! Join 137K+ Data & AI professionals who stay updated with the latest trends, insights, and innovations.

📢 Want to sponsor or support this newsletter? Reach out and let's collaborate! 🚀

Best,
Ravit Jain
Founder & Host of The Ravit Show