FREE IS HARDER
THAN IT SOUNDS

When I started building Pythos, the idea seemed simple.

Build a mathematics and physics tutor that students could actually use.

Make it useful.

Make it accessible.

And most importantly:

Make it free.

That last part turns out to be the hardest one.

The Idea Behind Pythos

I believe students should have access to genuinely useful help with mathematics and physics without having to pay for it.

Not a seven-day trial.

Not a limited version with the useful features locked behind a subscription.

Not another $20 or $30 monthly bill for a family that may already be struggling to pay for everything else.

Just a tool that a student can open when they are stuck and say:

"I don't understand this."

That's what I want Pythos to be.

And I'm still committed to that idea.

But building something with AI has introduced a problem I didn't fully appreciate when I started.

Free is difficult.

Nothing About AI Is Really Free

The website may be free to visit.

The software may be free to use.

But behind every question is infrastructure.

There are servers.

There are AI models.

There is storage.

There is bandwidth.

There is image processing.

There are databases.

And when a student takes a picture of their homework and asks Pythos to help them understand it, that picture has to go through a surprisingly complicated chain of technology before Pythos can respond.

Someone has to pay for all of that.

Right now, that someone is me.

And I don't want that to become the model.

I don't want the success of Pythos to eventually result in students getting a message saying:

"Upgrade to continue."

The Free AI Problem

There are AI companies that offer free access.

That's incredibly helpful for projects like Pythos.

But free access comes with limits.

Requests per minute.

Requests per day.

Token limits.

Vision limits.

And those limits can become surprisingly small when actual students start using the system.

We've already experienced this.

Pythos can work perfectly, students can be using it normally, and suddenly the underlying AI provider says:

You've reached your limit.

That's a difficult message to give a student who is simply trying to finish their homework.

So now I'm having to think about Pythos differently.

Instead of relying on one AI provider, I'm looking at multiple providers and ways of intelligently distributing the workload.

Instead of making the AI do every calculation, I'm building deterministic mathematical tools that can handle problems directly when possible.

And instead of blindly trusting the AI, Pythos checks its work.

All of that makes Pythos better.

But it also makes building Pythos considerably more complicated.

I Don't Want Pythos to Accidentally Cost Me a Fortune

There is another problem.

If Pythos becomes popular, I don't want to discover that I accidentally created a massive monthly bill.

So we've built something I'm particularly proud of:

Pythos now has a $0 AI spending guardrail.

By default, Pythos operates in a free-only mode.

If an AI provider isn't explicitly identified as free, Pythos won't use it.

If a free provider reaches its limit, Pythos can look for another eligible provider.

If there isn't one available, Pythos stops rather than quietly charging money.

The goal is simple:

Pythos should never surprise its creator with a giant AI bill.

That may sound like an unusual feature.

For a free educational project, I think it's essential.

So How Do We Keep It Free?

That's the question I'm working on now.

I'm investigating additional AI providers.

I'm looking into educational grants.

I'm exploring partnerships with educators and schools.

I'm looking at sponsorships.

And eventually, advertising may help support the project.

The important part is that I don't want to solve the funding problem by simply passing the cost to the students.

If Pythos eventually generates revenue, I want that money going toward Pythos and education.

Better infrastructure.

Better AI.

Better accessibility.

More educational resources.

School partnerships.

And, eventually, opportunities to support students and educators directly.

The objective isn't to figure out how much money we can make from students.

It's to figure out how many students we can help.

Building a Brain for Pythos

There's another direction I'm beginning to explore.

Rather than simply relying on increasingly large general-purpose AI models, I'm starting the groundwork for a purpose-built reasoning brain for Pythos.

Not a superintelligence.

Not an attempt to build the next ChatGPT.

Something much more focused.

A system designed specifically around tutoring mathematics and physics.

A Pythos brain should understand what a student is trying to accomplish.

It should know where the student is in a problem.

It should recognize when a student is proposing a step rather than asking a new question.

It should know when to calculate something itself.

It should know when to ask for clarification.

It should check its reasoning.

It should know when it doesn't know.

And most importantly, it should help students understand.

The underlying AI models can change.

The Pythos brain can remain focused on education.

Free Is Hard

I started this project thinking that making something free would mostly mean choosing not to charge people.

I've learned that it's much more complicated than that.

Keeping something free means finding ways to pay for it without charging the people you're trying to help.

It means being careful with infrastructure.

It means finding partners.

It means applying for grants.

It means building smarter systems.

It means figuring out how to stretch every available resource as far as possible.

And it means accepting that sometimes the easy solution is to put up a subscription wall.

I'm choosing not to do that.

At least, I'm going to fight like hell to avoid it.

Because I still believe the original idea is worth pursuing:

A student should be able to need help with math or physics, open Pythos, and get that help without worrying about whether their family can afford another subscription.

I'm still one developer building this thing.

I don't know exactly where Pythos will end up.

But I know where I want it to go.

Free.

Useful.

Accessible.

Built for students.

And if it takes grants, sponsors, advertising, partnerships, a purpose-built brain, or a ridiculous amount of engineering to make that happen...

then that's what we'll figure out.

We're still early.

But we're building.

— LANZAR
Build Tomorrow. Starting Today.