AI Money Compass

I quit six budgeting apps before I built one, and none of them died at the dashboard. They died the same week I stopped believing the number.

But that is not really why I built this one. I built it because every single one of them stopped at the chart.

They will all tell you that you spent RM842 on food this month. Not one of them will tell you whether you can afford the thing you are actually thinking about. That question — can I do this, given everything else I have going on — is the only question a normal person has about their money, and the entire category answers a different one.

So the advisor is the product. You ask it in plain words, and it answers from your own numbers.

Concretely, what it is looking at when you ask: your goals, with target, current balance, percent complete and what you are contributing monthly. Your commitments, with the monthly amount, what is still outstanding and which day it is due. Your per-category budgets. Your last 30 days of spending broken out by category and by merchant. That is a real picture of both what you are aiming at and what you actually do, and it is why the answer can say "yes, but it pushes your emergency fund back two months" instead of "consider reducing discretionary spending".

There are four of them and they disagree on purpose. A plain finance buddy. Buffett, who will ask which of your holdings are durable and which are speculative. Kiyosaki, who will classify everything you own as an asset or a liability and tell you your passive income ratio. Ramsey, who will tell you which baby step you are actually on and stop being polite about the credit card. Same ledger, four lenses. You pick the one whose framing you need that day.

The unglamorous half of this is the part I underestimated.

An advisor on top of a half-empty ledger gives confidently wrong answers, which is strictly worse than no advisor. And a manually-typed ledger is always half empty by week three — that is the failure I opened with. So before any of this could work I had to make logging nearly free: photograph the receipt, say it out loud, or drop in a bank statement as PDF, CSV or a screenshot, and it comes back with amount, date, merchant and category filled in for you to confirm.

The usual answer to that problem is bank-feed aggregation. I am in Malaysia, where that infrastructure barely exists — coverage is thin and the banks people actually use are not in it. Which turned out to be lucky. It forced me to solve the input properly instead of outsourcing it, and capture-plus-advisor is a different product from sync-plus-charts, not a worse version of it.

Where it stands, since numbers travel further here than features:

- Live on the App Store, version 1.3. Android in internal testing.

- 13 accounts that are not mine or test accounts. 12 of them signed up in the past 2 days, which is when the listing went live.

- Revenue: zero, deliberately. AI usage is metered per user in cents, and when your free allowance runs out a top-up pack buys exactly its face value in model usage. The AI bill passed through at cost, not a business. T

- Built solo — Flutter on iOS and Android, Next.js and Postgres behind it.

What I would like to hear:

If you have used an AI money assistant, did it ever tell you something you did not already know? Mine is only useful to the degree it can see your real commitments and goals, and I am not sure yet where the ceiling on that is.

And if you have quit a budgeting app — was it data entry, or was it that it never actually told you anything?