Having an AI assistant while reading has been really helpful. It’s like having a person who knows a lot about a lot of things at your beck and call. Considering the fact that LLMs like ChatGPT and Claude are pretty much trained on all the written knowledge of humanity available on the internet, they make for excellent teachers. So I built this simple eBook reader with an AI integration.

You can upload an eBook, and if you are having trouble understanding anything, you can invoke an LLM of your choice and then ask questions. All you have to do is set an API key using [Openrouter](https://openrouter.ai) or your LLM of choice.

[Download it here](https://github.com/bebhuvan/kruthi/).

What’s wild is that I pointed out a bunch of things to OpenAI’s Codex, and it just did them—suggested a bunch of other interesting and helpful features, figured out a bunch of bugs in the editor, and it just fixed it on its own.

And I have absolutely no clue how this entire thing works because I’m not a coder. But even thinking—I mean, I’ve said this before on the same blog—the fact that I could even think about doing this, let alone actually do this, itself is very, very wild.

And so I was just reading some of the book, and I was actually using this to keep my own cooking, and I kind of started figuring a bunch of things. And I kind of started noticing these small little improvements in the reading experience, the chat, and Codex is like, “No worries, we’ll just do it.” And it just kept on doing it.

I mean, this is not a serious piece of software. This is something I just built to solve my own issue and to improve my reading experience—improve my own… to help me read more.

And it’s kind of crazy that this is even possible. Wild, wild times.