RWD · Insights / Early work

An early read, and the toolkit behind it

A few pieces of work I'd like your eyes on. The first is an outside-in read of the real-world-data landscape and where the data is genuinely hard to copy. The second is the Claude Code toolkit I use to do that work, down to the agents and connectors underneath it. It's all in progress.

Seven guiding principles I work by, to set a high bar for the product work I do.

Listen before you propose

Understand what actually ships, and to whom, before pushing for anything.

Define "good" before you build

Write down what a right answer looks like, and how you'll check it, before building.

Build the workflow, not the feature

Own the whole job, not the one easy-to-copy step in the middle of it.

The data is the moat, not the model

Anyone can call the same model. The edge is the data and what you do with it.

Every bet clears the same bar

Accretive, defensible, cheap to run. One out of three is interesting, not a bet.

Nothing ships from one desk

It runs through Legal, compliance, clinical, and Commercial. Build for sign-off from day one.

Two speeds

Move fast where mistakes are cheap. Slow down to defend the numbers a payer signs.

The work

01 Competitive Research Where the RWD field sits, who actually overlaps with us, and the open questions I'd want to pressure-test before saying anything firm. Read → 02 Data Maturity Why foundational data has to be right before any AI sits on top of it — governance, linkage, delivery, and inspection, each signed off before the fun stuff. Read → 03 PM Toolkit The toolkit I built for RWD product work: competitive reads, data-feasibility checks, voice of customer, and a board-ready business case. Open → 04 Agents & MCPs The agents and connectors behind the toolkit, starting with a payer cohort builder that turns a plain-English ask into a HEDIS-style cohort. Open →