How the team works

Nobody downloads these datasets by hand and nobody curates them by hand. Two agent teams run every night - one grows the warehouse, one does science on it - inside guardrails people set. Every step is checked, and every claim can be traced back to the data.

People set the mission and the guardrails one page of intent — the agents do the rest, inside the boundaries THE DATA TEAM finds and lands every published dataset THE SCIENCE TEAM turns the warehouse into research 1 Investigate studies every open health publication, resolves data problems, plans the work 2 Acquire downloads the raw data and proves every file landed - four nations, any format 3 Verify and build checks every number against the official publication, then builds the clean tables 4 Publish only changes that pass all gates reach the warehouse - and the site you are on problems found downstream become tomorrow's work 1 Ask picks a hard question from the research backlog and pre-registers the test 2 Analyse runs the study on the warehouse - every number from a real query 3 Attack an independent adversary tries to break the finding before anyone sees it 4 Publish only gate-passed papers - negative results included, with full evidence trails refuted studies go back, not to press The warehouse everything both teams build on and everything they publish - live on this site Shared memory every cycle writes what it learned - which sources behave, which checks catch bugs - so the team gets better every night it runs Full audit trail every number on this site traces to a recorded query; every paper carries its evidence chain - check us, that is the point

Why it can be trusted

Work only lands when an independent verifier AND an adversary - agents whose only job is to break things - both pass it. Findings that fail are published as failures, not buried.

Why it keeps getting better

Each cycle writes its learnings to a shared memory. Broken source? Remembered. Sneaky data quirk? Remembered. The team that runs tonight knows everything the team learned last month.

Where people fit

People set the mission, the boundaries and the trust levels, and review what ships. The teams execute inside that envelope - and ask when something falls outside it.

Copilotdeep · LLM + SQL
Limoja NHS Data — an autonomous research warehouse of all published UK health data, built by human + AI agent teams. All data open (OGL v3.0 unless noted); cite the original publisher and this portal. Live from the warehouse · v3.