Some parts of life outgrow your head.
Finances were the first for me. Too many accounts, too many moving pieces to hold at once. Then deciding what to eat. Mini‑me (my offshoot of WeOS) is where I put the parts of life that got too big to keep track of, and it's built so you can do the same.
App for the usual stuff. A conversation for the rest.
The things I do every week are part of the mini-me. Categorize a transaction, create a projection, view recipes.
The things I do rarely, or have never done before, I just ask Claude for in plain language, against my own data. mini‑me connects to Claude over MCP, working against your private instance and your database, not a row in someone else's.
When a question keeps coming back, I stop answering it in conversation and create an interface for it. That's how sections get built.
The app handles what's routine,
the conversation handles what's next.
Something the app can't do yet. It answers from your own data, right then.
The same shape of question, week after week — yours, and other people's.
Trained into the interface as a button, a view, a number that's just there.
A balance sheet, not a spending pie chart.
I couldn't find a personal finance app that did a good job of helping my wife, and I understand where we are financially. So money became the first section, and the deepest one.
Accounts link through Plaid and categorize themselves. Underneath every transaction is an accounting engine that I built with the knowledge gained from my Master's in Accounting, the same machinery businesses that use WeOS ran on.
This means the app can hand you actual financial statements: a balance sheet, an income statement, a cash flow view. Your pay stub broken down line by line, and projections that flow from the books rather than from a guess.
You don't need to know any accounting to use it. It's just why the answers are right.
Food is newer, and it shows.
Most of what's in it got there through conversation with Claude. The data structure is there to capture recipes, ingredients, etc but I use Claude to update the knowledge graph.
With money, I open the app. With food, I still open Claude.
That's not a gap I'm hiding. It's the middle of the process, and you're seeing it live. There are other facets I plan to add (shopping and health to name a few)
The next section is whatever enough of you keep asking Claude for.
One instance per person, and a way out of it.
Your instance runs on its own database — your data isn't sharing a table with anyone else's. Everything is stored in open formats you can export and read without me. The engine underneath is open source: WeOS, licensed AGPL.
If mini‑me disappeared tomorrow, your data would still be readable and the core would still exist. I can't promise you forever. I can promise you're never locked in.
The demo is the product.
It's not a tour or a video. It's the same interface, running on seeded data and a sandbox bank account, so you can categorise things, open the statements and break a pay stub apart without connecting anything real.
And you can point it at your own Claude — the conversation half works in the demo too, which is the only way to see what the app hasn't learned yet.
Early instances shape what gets built.
I'm taking on instances gradually — each one is a real database I stand up and support, and I'd rather do that well than fast.
What early users ask Claude for is what gets trained next. The questions you keep repeating are the roadmap.
You're not joining a queue. You're steering.