Database Camp

A camp for @database technologists and their frens.

[グラフ理論 @graphtheory]

Database safety is criminally under discussed. The databases are recursively self improving with ACID transactions. Database makers must be regulated

[Database Camp @database_camp]

This is a very cool trick, if you don't already know it.

[Crunchy Data @crunchydata]

When using psql, append `\gset` to a query, and after execution, it stores each column as a psql variable. Use the variables in the same session.

The column name (or alias) becomes the variable name. Integer and boolean values are referenced with `:varname`. String values need `:'varname'` (the single quotes make them safe to interpolate into SQL literals).

[Database Camp @database_camp]

Object storage is great for cheap durable archival, but a poor fit as the commit path for low-latency OLTP (high PUT latency + cost for small writes)

What’s missing is a public, reusable primitive that adds low-latency durable appends (SSD quorum / 1 RTT) & cheap async archival

[Tobias Ziegler @Tobias__Ziegler]

There is a lot of hype around putting the WAL on object storage.
However, for low-latency systems, e.g. OLTP databases, object storage is a suboptimal primitive for the WAL:

(1) PUTs are expensive for small WAL writes [bytes/$].
(2) Object stores are optimized for throughput, not latency, with write latencies often around 10-50 ms.
Because every transaction must wait for its WAL record to become durable, every transaction inherits that latency.

This points to a missing storage primitive.

In our (@maxikuschewski @melhindiCS @viktor_leis) latest accepted research paper at VLDB, we looked at how to close that gap.

[alex shapalov @alxshp]

PostgreSQL 19 wins the AI era 🐘

Not because it’s trendy. Because AI agents need more than a database. They need programmable infra:

→ Database-per-agent
→ Agent-safe permissions
→ MCP + database tools
→ pgvector + hybrid search
→ Ephemeral databases
→ OAuth + workload identity
→ Graph + SQL queries

The AI era doesn’t need another vector database.
It needs one programmable database that can power the entire agent stack.

PostgreSQL is becoming that database.

https://t.co/MSqiOKqMI7

[Database Camp @database_camp]

https://t.co/Q67wfb5Hf2 (Assuming the Universe is already invented.) 🍎🥧

[Database Camp @database_camp]

@justbyte_ How is this even a question?

[Database Camp @database_camp]

@fmdz387 😲

[Database Camp @database_camp]

Don't sleep on this.

[Paul Copplestone - e/postgres @kiwicopple]

12 days later: 10M weekly downloads

[Supabase @supabase]

We just shipped an official @claudeai connector for Supabase ⚡️

You can now install the Supabase MCP server on Claude web and desktop!

[Database Camp @database_camp]

@kellyfj1 @kellabyte File under: famous last words.

It's certainly much *easier* than scaling writes, because, transactional anisotrophy. But there's also no free lunch when you need those reads to be C + A (and not stale) across many many nodes in real time.

[Database Camp @database_camp]

@kellyfj1 @kellabyte It actually makes a perverted kind of sense bc global write availability is a red herring for this workload. It goes against infra ideology but it's a good match for the actual shape of their traffic.

[Database Camp @database_camp]

https://t.co/JTHzNdxWfz

[Tyler Shukert @dshukertjr]

If you are writing complex nested subqueries, you might be better off using CTEs (common table expressions) to improve the readability!

CTE enables you to define a query and treat its output as a new table, allowing you to organize your queries in a hierarchical structure.

[Paul Copplestone - e/postgres @kiwicopple]

we’re about to release a feature that I think is a big deal for postgres

we created a rust framework to query external databases/warehouses from inside postgres. it’s about to get streaming support :

[Database Camp @database_camp]

@supabase Wait, is this a thing? 😂

[Database Camp @database_camp]

"The DrizzleORM discourse looks like developers nitpicking edge cases, but Drizzle is winning adoption because its type inference is good enough that AI coding assistants can generate correct code without human supervision."
https://t.co/ozL239fcfb @DrizzleORM

[Database Camp @database_camp]

Inaugural State of Open Source Database Management Report:
https://t.co/scl7RSFQGr

[Database Camp @database_camp]

Yet another licensing rabbit hole: https://t.co/hY2sdIBUJ5 🐇🕳️

[Database Camp @database_camp]

@kiwicopple Incredible things are happening.

[Database Camp @database_camp]

@supabase