I've been building an entire AI trading ecosystem from the ground up, solo, for the past 2 years. Now that I'm finally in the last few miles of this marathon, I think it's a good time to start documenting the journey — while I still remember it.

Sometimes I hesitate to call it a startup. It’s really just something I personally want and believe in — AI agents investing on behalf of humans is inevitable. If it gets funded someday, great, I can turn it into a SaaS serving like-minded people. If not, I'll still run the

Lesson learned #1: One bad technical decision can cost you weeks — sometimes months.

It's easy to write code, but hard to build reliable software. I've been struggling to maintain multiple packages all at once. I think Cursor and Claude Code are good, but I believe we need a new IDE where AI agents are the main coders and we're their copilot. Imagine that

I'll be very surprised if we can reach true AGI without a real mathematical foundation behind it. I'm still waiting for something like Intelligence Theory that is as fundamental as Information Theory to emerge in this era to make AGI happen. If we can't define Intelligence, we

I ran a half marathon untrained years ago and thought it was easy. I remember in the last 1 or 2 miles, my knee hurt so badly and I had to walk to the finish line. That's exactly how I feel right now in my startup journey.

An IDE in this AI era should be: - AI agents as the main coders, and we act as their copilot - a way to glue multiple programming languages together so AI can write freely across them - interactive, blurring the line between dashboard apps, notebook and full IDE Unfortunately,

The world is far from perfect, that is also true in tech. There are lots of problems I want to solve, and web development is definitely the top one that bothers me the most. Not that I am interested in it, but it just annoys me so much that I want to fix it once and for all

I think one of the hardest parts of software engineering is estimating how long something will take. I always end up underestimating by a factor of 2–3.

Lately I’ve been working on something simple in concept: a tool that generates a standard project website for any Python package, so developers can start writing docs, tutorials, or blog posts right out of the box. The idea was simple… but integrating Svelte with MyST Markdown

Most of my packages will reach v0.1.0 in the coming months. Here's the first one: github.com/AfterPythonOrg… It's actually my first v0.1.0 release. Feels good to finally give sth back to the community.