Xowit

If you're not embarrassed, you're too late

I’m always full of ideas.

But they never go beyond ideas.

Sometimes I would buy a domain and promise myself that “this is the one”. And the list of domains grows.

Experiment#

But this year I’ve decided I’m simply going to experiment. I’ve lowered the level of what I want to achieve.

My first experiment is about reading. I read too little. But I no longer want to ship the Netflix of books. I just want to use the wonderful Project Gutenberg. And if it is helpful for others, great.

AI#

However. It is not only that I have lowered expectations. The level of what I can do has increased.

AI is a multiplier. This first experiment has made signicant use of Grok, Claude and Claude code. From refining my idea, to writing the “copy” (text), to styles and images, to code.

I’ve passed all the usual steps using AI. From “OMG, this is frigging incredible”, to “no, claude-code, you cannot fix something just by putting a delay”, to “how the heck can you have copies of the same file in two different places and call it refactoring!!”, to “set aside and let me fix this project. I want something that even I can understand”. And finally, back again to the vibe coding of “just change the styling of the project to make it visually more appealing” and wait for a commit that touches 20 files and makes an incredible job (yes my styling was much worse).

Not only that. There is a bit of prompt engineering to get book recommendations. A very cheap model. Just to experiment without breaking the bank.

And more important. From idea to up and running in 3 weekends.

AIs are amazing. And if you’re not playing with them (yes, you, both my non-technical AND my technical friend), you’re missing a lot.

And this is it#

This is the result:

Xowit

Xowit helps you get book recommendations from your Kindle into your Kindle. It scans your Kindle, suggest things that are similar to the ones you’ve read, and let’s you one-click copy those suggestions to your Kindle.

Yes, it is free. And considering the costs, it is likely to remain free.

Please, play with it and ping me. What do you like? What you don’t like?

A bit of embarrassment#

Xowit probably will have bugs. I was going to launch it last Sunday and had the good sense of having my wife and one of my daughters to test it. It was not good.

Also, the aesthetics are inconsistent and not great.

The recommendation engine is poor budget friendly.

Yes. We should read more. But why only on Kindle?

I hate the name. What kind of name is Xowit?

Not all the books in my Kindle are interesting books. Let me select the ones I like.

A bit more and you would have used Comic Sans. Wouldn't have been better to have a consistent style across the app instead of this "collage" of an app?

You know the drill about how you should be embarrassed by a first version.

I am proud of what I have, and embarrased of what it could have been and it is not.

I will paste the original author of the quote.

So here. To an early launch.

Some details#

For those more into tech, I might detail a bit more in a future post. But just to name a few technologies:

  • Javascript. Svelte (not NextJS/React).
  • Cloudflare. No, not AWS. Seriously, not AWS. Yes, I still like AWS. Also D1 database. R2 (not S3).
  • OpenAI gpt-4o-mini.

The idea was to be as nimble as possible. I don’t want to pay much for a experiment. The only fixed cost is R2 storage to store all books (Project Gutenberg doesn’t allow you to link directly to a file). If my calculations are correct, that should be around 2.5$/month. OpenAI costs could be more significant, but all my testing resulted in a grand total of … 2 cents.

The actual code running might not end up costing a penny.

If people start using this, costs will increase. But I doubt it will be such a success that I need to optimize.

Just one more thing#

I mentioned it at the beginning. This is only a first experiment.

Sure, I will keep working on Xowit.

But I’ve got more ideas. April is about to start!!!

© Gonzalo Fernandez-Victorio 2025