I’ve done genealogy on and off for years, and I’m still not an expert. But I’ve finally fixed the part I was worst at. It wasn’t the research. It was everything around it.
How it used to go
It usually started with a few spare minutes. I’d open FamilySearch, pick someone more or less at random, and see if anything jumped out. A few quick searches. Maybe a hit, maybe not.
Most of the time I ran out of time and moved on to something else. Whatever I’d looked at was gone. No note, no record of what I’d searched.
The other times were worse, in a way. One thing led to another. A census led to a neighbor, the neighbor led to a newspaper, the newspaper named a brother I didn’t know about. An hour later I’d have a great find and no real idea how I got there. Somewhere along the way I’d passed three or four side branches I meant to come back to. I never did.
The problem was simple. The fun part for me is the discovery. The log keeping is not fun, so I didn’t do it. The result was a handful of interesting finds, the same searches run over and over, and leads that never got chased down.
How it goes now
I’ve already written about using Claude Code for the research and about the admin section behind the site. This post is about what all of that did to my workflow, which turned out to be the bigger change.
Every session starts the same way. I open Claude Code and type /start-session. It works through the same checklist every time:
- Pulls in anything I sent from my phone or dropped in a folder on my desktop since last time, and files it.
- Asks whether I want to refresh the tree from FamilySearch (a full pull takes about 30 minutes, so it asks) and from WikiTree.
- Re-runs an audit of the 368 people in scope: who’s incomplete, which dates are impossible, what changed since last time.
- Checks whether the family turned up anywhere new that needs its sources mapped.
- Reports what changed, what to work on next, and a short list of things I can ask it to do.
That last part matters. It doesn’t go off and do everything. It tells me what’s ready and I pick.

When I get an idea
I get ideas in the middle of things. A new archive, a different way to search, a source I hadn’t thought of. Before, I’d chase it right then and lose whatever I was doing.
Now I tell Claude Code and it files the idea into a future phase of the project. When I’m ready, I say “plan phase 21” and we work out how to build it. WikiTree and the National Archives catalogue both came in that way, as ideas I had while working on something else.
I even added a rule that it won’t start new technical work while a session is still starting up. That one is mostly a guard against me. Building new tools is fun. Working the to-dos is what actually moves the research forward.
The five minutes
The admin section has a to-do list. Right now there are about 640 items on it. Every one says exactly what to do: which record to open in FamilySearch, which field to change and to what, which search to run with which filters, and what a match looks like.
About 96 of them are quick wins. A field to change, nothing to look up first. Those are my five-minute items.
If I’m out near a cemetery, a “Near me” filter shows anything I could do there. When I finish an item I mark it done, none, or stuck with a note, and it gets filed at the next session. A stuck item becomes an open question in that person’s research log instead of disappearing. If it stays stuck, the project can write up a brief for a professional genealogist. I’ve hired one before and it was money well spent.
Same five minutes as before. It just isn’t throwaway work anymore.
Research I still do on my own
I’ll still go off on my own. I know myself.
When I do, I drop whatever I find into the tool. A photo of a headstone, a clipping, a scan, a few typed notes. From my phone it goes through the admin’s upload page. At my desk there’s a folder on my desktop.
At the next session, Claude Code reads it, transcribes it, works out who it’s about, compares it against the tree, and writes it up with citations. Then it does the part I never did. It follows through. For every person named in the new document, it checks their open to-dos, their open questions, earlier records that disagree, and anything else left hanging. The project won’t even let a new record be committed without that step.
And every search gets logged, including the ones that found nothing. There are more than 12,000 of them now. That’s the log I never kept.
What that looks like
A recent example. Newspaper searches turned up four items, 1905 to 1908, about a store in Kellyton, Alabama called Webb and Walden. The firm was one of the incorporators of the Kellyton Gin Company alongside C. D. Webb and two of his sons. It put up a new brick building. It advertised the whole business for sale. And it was the last stop for a gang of burglars who hit the town’s stores, express office and post office one Sunday morning in 1908. They gave up trying to bore through its front door when the sun came up.
A fun find. Before, that’s where it would have ended. This time the write-up says plainly what the papers never do: who the partners were. Probably James Pleasant Webb and his father-in-law William Walker Walden, but other Webbs and Waldens in Kellyton could fit just as well. So it’s written down as an open question with the evidence beside it, and it’ll still be there the next time I have five minutes.
Where this leaves me
I’ve gone from a disorganized researcher to an organized one, without having to turn into a different person. I still grab five minutes when I have them. I still go down rabbit holes. The difference is that nothing gets thrown away, and the side branches get written down whether I’m paying attention or not.
The finds end up on webbwade.com, where the Recent discoveries page is up to 88 entries. Not bad for someone who used to forget what he searched yesterday.
Say something