rawquery: Average Is All You Need

Sid Shah thinks you shouldn't need a data team to answer basic questions about your business. His new platform, rawquery, is built specifically for LLM agents to operate on. You connect your data sources like Stripe and HubSpot, then tell Claude Code or Cursor what you want to know in plain English. The agent writes the SQL, runs the queries, builds the charts. You focus on asking the right questions.

The demo on rawquery's blog walks through a real scenario. You have transactional data from Stripe and email campaign data from HubSpot. You want to know if a mailing campaign actually increased revenue or customer basket size. Instead of building what Shah calls "attribution model wankery" with debates over last-click versus first-click versus weighted attribution, you just describe the joins and correlations you want. Claude writes the SQL, you get the table showing that the email cohort had a 46% higher average basket. No SQL written by you, no data team needed.

Shah's broader thesis, laid out in his post "Average Is All You Need," is that LLMs are making average-quality output cheap and fast. Creative fields got hit first. Now software and data are getting the same treatment. And he thinks that's fine. "It is in fact amazing that anyone can now create average things whereas before they had to fight hard for sub-par," he writes. The argument is that most people already have good intuition about what's in their organization's data. They just lack the technical skills to extract it. LLMs handle the mechanics. Humans handle the thinking. However, research has found that cognitive surrender is common, where users uncritically accept AI outputs.

This is a practical pitch for agent-first infrastructure. Rawquery isn't trying to replace complex analysis or deep statistical work. It's targeting the vast middle of data questions that would otherwise require writing SQL, understanding sync strategies, or building dashboards nobody reads. For teams without data resources, average answers to most questions beat perfect answers to no questions.