LLMs give their best results when they have good context. For example: knowing how to include Explore dashboard links in their responses.

@jonwalls.dev shows how to include ai_instructions to your model, which automatically adds the context for every conversation.

docs.rilldata.com/explore/mcp#...

Most enterprise AI agents are brittle, siloed, and unobservable. Agentic mesh is architecture for agents that are composable, secure, and actually interoperable. Think microservices for AI, but with traceability and zero-trust as defaults.

Your users don't care what tech stack you're using.

They care if your product is slow.

And your tech stack can affect that.

Your tech stack matters.

Rill is proud to sponsor DuckCon6 on Friday!

@medriscoll.com will also present on Rill's SQL-based metrics layer powered by @duckdb.org. Recognizing the wide popularity of SQL & DuckDB’s SQL dialect in particular, we developed a metrics layer that declares its metrics entirely with SQL expressions.

Deepseek reasoning is damn good!

Async code should never spend a long time without reaching an .await. - Alice Ryhl

I’ve often said working in data engineering is like working in the post office. Queues get backed up, deliveries are lost, yet the stream never stops. It’s a thankless, mostly invisible job until something goes wrong, & then it’s complaints.

I hoped the analogies stopped there, but sadly maybe not.

OSA Con starts tomorrow! This free virtual conference offers though-provoking panels, practical workshops, and hands-on sessions from thought leaders including our very own @medriscoll.com.

There's still time to register👇
osacon.io/register/