I have no idea why I like this comment, but I do.

By chapter four you have a minimal working codebase in python. Later chapters add out features like tool calling, planning, and reflection.

I’m writing my own harness to learn. And I’m writing it in zig.

But, I’ll absolutely add maki.sh to my reading list. So, that’s a win.

At home, I mostly use openrouter.

A few bucks will realistically last you forever. I also bought a DGX Spark at launch. I've learned a ton. but it's not usually my daily driver. I do run overnight tasks on it, like "read this book and write up notes on each chapter..."

In retrospect, given the dirt cheap prices of apis and or renting hardware, buying hardware for inference was likely a mistake (but not enough of a mistake to sell it now).

And I’m slowly writing my own. But it’s garbage at the moment.

The author’s discussion of their personal use case starts on the seventh sentence in TFA. And frankly, it sounds like a pretty solid use case:

“Being able to run things locally and mess around with RL, quantisation and fine-tuning on a laptop would be sick. I already publish my own GGUF quants on Hugging Face, with over 160,000 downloads in the past month.”

3.8k is less than the preorder cost of the DGX Spark over a year ago (Oct 14th, 2025).

I don't see how we can discuss this demo without acknowledging that fact as well.

[1]: https://apnews.com/article/russia-ukraine-war-zelenskyy-burn...

I guess what I'm saying is performance is fine... but what I really want to see is performance/dollar.

Thank goodness.

And if they do that accurately, there's no stopping someone from mapping the guard rails.

Clearly that outcome is more uncomfortable than what we have now, which is enough people seem to be okay with vague billing.