One Record

My career gets described in a lot of places: a site, a resume, LinkedIn, a links page, a set of interview decks. Each one is written at a different time, in a different mood, and they drift. The resume says one thing and the site says something close to it. My experience in designing systems drives me to treat content the same way I’d treat a product with five versions of the same button. It needs one source.

The record

The record is a folder of plain text files on my own computer:

• A profile: the roles, teams, projects and outcomes, one row per claim.
• Evidence notes: one file per employer, pointing each claim to the document it came from. Those are my own project files, readouts and research notes.
• A decisions log: each choice I’ve made about wording, with the date and the reason.

Every claim has two wordings: one for public pages, and one for interviews. It’s plain text on purpose. I can read it, an agent can read it, and a change shows up as a difference between two versions. All of this is built on top of a well organized series of folders, one for each stop along my career journey.

What it feeds

The site, the resume, my LinkedIn profile, the links page and the interview decks all take their wording from the record. When a fact changes, it changes there first and is carried to each surface from there. Nothing gets retyped from memory.

The audit

A record built by reading hundreds of files will misread some of them. So before this site launched, I ran an independent pass. A fresh session in Claude read the original documents first, without seeing the record, and only then compared the two.

It found three places where the record had misread a document:

• a survey result I’d described as approval, when the document measured whether people understood the change
• a result labeled as a controlled test, when the source didn’t say so
• research listed as the starting point for a strategy, when it had sharpened later versions

None of them were large. All of them would have been awkward in a reference check. They were corrected in the record, then on every surface, in a day.

The rule on numbers

One rule decides what’s public. Numbers about my work and how people responded to it are on the page: team sizes, research counts, engagement in rounded form. Numbers about a company’s business are not. Those I share only in an interview, where I can give the context around them.

Who does what

I work on this with Claude Code. I decide and approve. The agent reads, drafts, checks and carries a change across every surface.

What made the workload manageable was how the decisions reached me. The audit raised more than thirty questions. Instead of one long report to answer, the agent stepped me through them as question cards: one question at a time, with the options and its recommendation, and a check-in before each new stage of work. Each answer went straight into the decisions log, so nothing changed anywhere until I’d said yes to it.

If you want to try it

This is a design system applied to my career. There’s one source, the surfaces read from it, and a review step catches what drifted. I would recommend starting small: start organizing your sources of information, write down ten claims you make about your work, and next to each one, the document that proves it. The ones you can’t source are the ones to fix first.

Where I can go from here

Now that my source of truth is established, I’ve been thinking about other types of content I could create from my work. For example, a podcast talking about my work experiences, or an Instagram account that drips snazzy posts about my best work. The possibilities with this level of improved consistency and organization are super exciting.