I built FIREMaster as a dashboard. Charts. Projections. A config page where I enter assumptions. Standard web app.
It worked. You could see your net worth broken into pools, visualize generic scenarios, watch the projection line bend when you change a number.
But the most important breakthrough didn't come from looking at these charts. It came from a pivot after an accidental realization.
The question no dashboard can answer
A few days after the layoff, I was staring at my checking account balance and doing the division everyone does. Cash divided by monthly burn. Six months. Maybe seven if I cut hard.
But I had three properties. One was bleeding cash. One was breaking even. One was making money. And I had retirement accounts I couldn't touch for six more years without a penalty — unless I set up a specific IRS withdrawal plan that most people (myself included at the time) have never heard of.
The question I needed to answer wasn't "what's my net worth?" I could see that on any dashboard. The question was:
"If I sell one property and use the proceeds to pay down the mortgage on another, what happens to my runway — and do I still need to tap my retirement accounts early?"
No chart answers that. No calculator has a button for it. It requires understanding the relationships between cash flow, property costs, rental income, retirement withdrawal rules, and time — all at once, for my specific situation.
The advisor that wasn't
FIREMaster pulls all its data from Monarch Money — a financial aggregator that connects to your banks, brokerages, retirement accounts, and mortgages in one place. All 27 of my accounts sync there automatically. That's the raw data. FIREMaster is the engine that makes sense of it.
I built that engine with Claude Code — Anthropic's AI coding tool. It's a terminal-based AI (I personally prefer to use it via the nice UI of VS Code) that can see your entire codebase, run commands, call APIs, and reason about the results.
The original plan was to embed an AI advisor directly into the app — a chat panel in the dashboard where you could ask questions about your finances. I built it. It worked. And it was immediately obvious it wasn't good enough.
The in-app advisor could only query data through the specific tools I'd wired up for it. Ask something outside those boundaries and it was stuck. It ran on a smaller, cheaper model. Every question cost me API tokens. And it couldn't do the one thing I actually needed: reason freely across my entire financial picture, combining data from different parts of the system in ways I hadn't anticipated.
Meanwhile, Claude Code — the tool I was using to build the app — could already do all of that. It could see the entire codebase. It could call any backend API directly. It could pull my real account data, run a scenario with different assumptions, and explain the results in plain English. No predefined questions. No per-query cost beyond my existing subscription. No limitations on what it could connect.
The pivot happened the moment I asked Claude Code a real financial question instead of a coding question:
"If I sell one property and use the proceeds to pay down the mortgage on another, what happens to my runway?"
It pulled my real account balances from the database. It ran both scenarios — sell one property versus keep both. It compared the projections side by side and told me:
"Selling extends your liquid runway by 18 months and eliminates the carry cost gap. Terminal wealth at 82 is 25% higher. The tradeoff: you lose that rental income starting immediately."
Nobody programmed that answer. The model read my data, ran the projections, and reasoned about my specific situation. The tool I was building with had become the tool I actually needed.
Once I realized that, the questions kept coming. I started asking it questions I'd never ask a financial advisor — partly because they're too specific, partly because I'd be embarrassed, partly because no advisor has the patience to run fifteen "what if" scenarios in one sitting:
What if I keep my property and pick up consulting work for two years?
What if the market drops 20% in year one — does the plan survive?
What if I delay Social Security to 67 instead of taking it early at 62?
Each question got a real answer, grounded in my actual numbers. Not generic advice. Not "it depends." A specific projection with specific tradeoffs.
The two-interface architecture
That's when the design clicked. FIREMaster isn't one interface — it's two:

Dual Interface Architecture
The dashboard is the cockpit — you glance at it to see where you stand. The co-pilot is Claude Code — you talk to it when you need to think through a decision. Neither works without the other. The dashboard without Claude Code is just another calculator. Claude Code without the dashboard has no data to reason about.
This is what makes FIREMaster different from every other financial planning tool. Boldin gives you a dashboard. ProjectionLab gives you projections. Monarch gives you tracking. FIREMaster gives you a thinking partner that can see everything and reason about your specific life — not just your numbers.
The takeaway
The most powerful financial planning tool isn't a dashboard. It isn't a spreadsheet. It isn't even a better calculator.
It's an AI agent that can see all your data and have a conversation about what to do with it.
The co-pilot architecture emerged from the building process itself — and it turned out to be the breakthrough I was looking for, just not where I expected to find it.
If you're building your own tools with AI — for finances, for health, for anything personal — pay attention to the moment the tool you're building with becomes the tool you actually rely on to answer the hard questions nobody else has the time for. That's where the real product lives.
Want to try it?
FIREMaster is ready for its first users. Not a waitlist — the real app, running on your machine, with your data.
Right now it's designed for a specific kind of person:
You use Monarch Money and have at least a year of financial history synced
You have a Claude Code subscription (Pro or Max) and use it for real work
You're comfortable with Docker and the command line — this is a clone-and-run repo, not a download-and-click installer
You're navigating (or stress-testing) the bridge between working and retirement — especially if the timeline just got forced
If that sounds like you, reach out. I'll send you access to the GitHub repo and help you get set up. First users get it free — I want to learn what works and what doesn't.
Next issue: the specific IRS loophole that let me access retirement funds six years early — and how most FIRE content gets it wrong.
— George
George Borshukov is a multiple Sci-Tech Academy Award winner, former Special Projects lead at Epic Games, and the builder of FIREMaster. He writes biweekly about FIRE strategy, quantitative finance, and building in public with AI.