Pip

I notice people in my everyday life just checking bank accounts and using that as permission to spend money.

I would literally watch them pull out their phone, check their bank app, see that they had money, and then spend it.

I've seen this multiple times, and I know it is a common habit across many people in the US.

These people would never touch a spreadsheet. They would never touch a budgeting app.

It's too complicated. It's too much work. They don't want to do it.

The only personal finance they do is check the number in their bank app.

Why budgeting apps don't work for these people

People don't like to use budgeting apps.

They're complicated. They're intimidating. It's a lot of work.

And then you have to pay money to use them.

There's a lot going against them.

Pip is actually fun to use. It's got a cute character. It's simple.

You can just log in to a bank app, connect it, and ask about your finances.

You can ask it what bills are coming up.

You can ask it if you can afford a big purchase.

And if you can't, he'll help you start saving for that purchase.

He'll tell you how much you can save for that month, and he'll tell you how long it's going to take before you can get there.

And if you overspend on the Spendable Cash Today, he will adjust the rest of the month because that's how it works.

All the calculations are done programmatically. There's a formula it uses to make it work.

It takes out the thinking in budgeting for everyone.

The core idea: Spendable Cash Today

The whole app is built around one number:

Spendable Cash Today

The idea is simple.

Instead of checking your bank balance and using that as permission to spend money, Pip gives you a number that already accounts for bills, spending goals, savings, and future spending.

So now anyone has access to a simple automatic budgeting system without needing to deal with spreadsheets, budgeting apps, or a bunch of numbers.

Why I stripped the UI out of the app

I stripped the whole app of UI.

It's just chat with recommended chips to navigate conversations.

The whole idea was to make it dead simple.

There's also a little Pip speech bubble that pops up when you first open the app, and he just gives you updates based off of what's the highest priority.

He might say things like:

• I see payments to a credit card that you haven't connected. Do you want to connect that so I can get a better view?

• Your recent transactions sent you over the Spendable Cash Today amount.

• Your Spendable Cash Today is higher than normal because you've been saving money.

It was really hard trying to fight not putting more UI into the app.

All the current AI tools constantly want to throw more UI at stuff, and I just had to keep giving it a hard no constantly.

Yet I had to find a way to add ways to do account management or access settings, and most things are done through tool calls with the agent.

Some of them are just brought up programmatically, like settings.

Technical challenges

Calculating Spendable Cash Today was one of the main technical challenges.

I had to go through several iterations of that to make it work well, and it probably will still be tweaked in the future as well.

But I think I landed on a place that I like now for the time being.

I also tried to use very low intelligence lightweight models for this agent, which made it fairly challenging.

You have to keep the tools pretty simple and clear for the agent.

And you can't have too many tools.

So I tried to keep the cards minimal and clearly defined.

It took a while to have it regularly find upcoming bills and recurring payments that it could pick up as bills.

You have to have enough transaction history to do so.

With the low intelligence models, it's not always the easiest, but we figured it out.

Why I made it open source

I decided to make it open source so that I can contribute as a developer and showcase my work.

But also it works very well in this situation to just build trust and show that there's no funny business going on.

You can look through the code and see everything we're doing.

It's just a simple agent structure that reads your transaction history, uses that to calculate your Spendable Cash Today number, and answers questions about your personal finances.

Who Pip is for

The people who seem to like it most are younger women, from what I've noticed.

Which isn't surprising because I did think of my own girlfriend when building the app, and she helped with the design direction as well.

It's not for somebody who wants to see a bunch of numbers and tweak their budget and see forecasts on where they're going.

It's a dead simple app for a reason.

It's supposed to make people enjoy doing personal finance without too much thinking.

Current status

Pip just released.

It's still very rough around the edges.

It still has issues as an agent.

Sometimes he'll ask the same questions over and over again.

Sometimes he won't really know what you're asking him.

But the core structure is there and works.

The next thing I need to work on is definitely honing in on making Pip feel more alive like an agent, like an assistant.

Someone that can just talk to you about finances.

At the moment, it's pretty basic in what he can do in that area.

What I learned building it

I learned a lot about how agents work when building this.

Like the design philosophy for Pip, I tried to keep it as simple as possible and try to do a lot of the work outside of the agentic structure.

I assumed I could just plug everything into an LLM and it'll spit back a bunch of good information.

Which I guess is technically possible, but it doesn't feel good.

You'd have to use a very expensive model, and you'd get a lot of random outputs, and it would be slow.

It wouldn't work for a consumer app that's supposed to be dead simple.

They don't have the patience for that, and it'd be expensive.

So it would not work.

Why I built it

I'm not really sure if this is going to take off or if I'm going to continue working on it.

I mostly did this as proof to myself and the world that I can ship products from end to end.

This one's ready to go. Set up. And it's actually very useful.

I use it in my day-to-day life still, and so does my girlfriend.

If I want to see recent transactions, my bank balance, or where I stand at the moment with my personal finances, I open Pip now.

I truly do.

I genuinely think it's useful.