The MURE LOG #9: context is the product

The data systems log, curated by humans

May 25, 2026

Hey, welcome back!

Turns out the real differentiator is your company and your data. This drop covers persistent context, knowledge systems, and the unglamorous data work that separates AI that looks impressive from AI that actually delivers.

Our feed

01 / SQLite in Production: Lessons from Running a Store on a Single File - ultrathink.art
They run a production e-commerce store on SQLite. A single file handles everything.

02 / How We Built an AI Second Brain for 60K Knowledge Workers - Analytics at Meta
What if an AI agent had persistent, structured access to everything a person is working on, and carried that context across every interaction? Company brain.

03 / Interaction Models: A Scalable Approach to Human-AI Collaboration - Thinking Machines
Models that handle interaction natively rather than through external scaffolding.

04 / I built a company brain to run my YC startup - Claire Gouze
How to create a Git repo that keeps learning your company knowledge. Company brain, second brain.

05 / The Boring Work That Makes AI Analytics Actually Work - Opeyemi Fabiyi
Context is everything. The teams winning with AI analytics are the ones doing the boring work.

06 / Expanded interoperability with Unity Catalog Open APIs - Alex Jiang and Tathagata Das
External engines can r/w managed and external Delta tables, using credential vending.

07 / Internal vs. External Storage? What’s the Limit of External Tables - Simon Späti
Great write up. An inside look at external tables, their 25-year history and evolution.

08 / Stop upgrading your LLM. Start fixing your data - Ben Lorica
Data is everything. “Agents need governed, scoped access to data, with proper permissions and audit trails built in from the start. Without that foundation, even a well-integrated, well-trained agent is a liability waiting to surface.”

What's new

Anthropic - acquired Stainless
Apache Iceberg - Iceberg Summit 2026 videos are online
Columnar - ADBC driver for Quack
DuckDB - Quack, the DuckDB´s client-server protocol
Google - Cloud Next '26 recap
Google - data agent kit, collection of data engineering and data science skills, tools and plugins
Microsoft - Fabric April 2026 feature summary
Onehouse - OpenXData videos are online
Snowflake - dbt Fusion is now available on Snowflake

Tools & demos

Bird bench - https://bird-bench.github.io, a bench for large-scale database grounded Text-to-SQLs
Coral - a SQL interface over APIs, files, and live sources
dbt Labs - ade-bench, a framework for evaluating AI agents on data analyst tasks
Harbor - terminal-bench, a benchmark for LLMs in the terminal
Mure Data - agentic-data-tools, a data skills manager and index
Nao - sylph, OSS company brain
Strukto AI - mirage, virtual filesystem for AI agents