MY STORY
I started getting into trading in 2020. Like a lot of beginners, at first I was doing anything and everything. I gambled, switched strategies every day, listened to the wrong people, and paid gurus who were not even profitable.
I lost five years thinking that what I was missing to succeed in trading was a good strategy. Except that is completely wrong. Trading is simply math. It is probabilities.
Yeah yeah, but past performance does not guarantee future performance. Sure, but if your strategy has never worked in the past, it’s not going to work in the future either…
Anyway, all of that to say that what I was missing was a way to easily test whether a strategy works or not.
That is where it started.
THEORYCRAFT
TheoryCraft is an AI research lab for traders and investors who want to test, create or improve their strategies with a data-driven approach.
Basically, you talk to an AI. You ask it to create a new strategy or validate an existing one, and the AI creates a Jupyter notebook in Python to validate all of that with real data. Since it’s a notebook, you can of course grab the code afterwards and run it on your own machine if you want to trade the strategy, for example.
But that is only the tip of the iceberg: the MVP. The goal is to go much further and offer a real scientific approach to trading.
Basically, a lot of people think creating a strategy is complicated, but it’s not. A strategy is just:
- an Edge — a pattern that repeats in the market and gives you a statistical advantage
- a Filter — your edge does not necessarily work in every market regime or every day, so sometimes you need to filter the signal
- Risk Management — position sizing, knowing when to stop trading for the day, etc.
If you have that, you have a strategy. Whether it’s trading or investing, it works the same way.
Now you still need to validate that strategy. Without going too deep into the details, because that is not the point of this page, there are methods like Walk Forward and Monte Carlo that I want to implement “natively” in TheoryCraft.
So the goal is to have a complete ecosystem that combines AI, process and data to create and validate strategies.
THE BUILD: 291 DAYS
I started TheoryCraft on September 3, 2025 and launched the MVP on June 21, 2026.
Too long, way too long… Looking back, I think I could have built the MVP in one month instead of ten.
My biggest mistake was over-engineering. I lost so much time on things that were useless for a first release… At the end, I had to refactor the code and I deleted three months of work because it was never going to be useful.
It was awful, but I learned a lot with this project.
LESSONS
- Stop trying to make everything perfect. I am a bit of a perfectionist, and I have spent years rewriting projects because they were never clean enough.
- Ship the MVP quickly. If nobody wants your app, you need to know as early as possible so you do not waste weeks/months of work.
- Focus MVP. This goes back to what I said above, but for the first version you need to ship the minimum: THE key feature.
- Deadlines work on me. Meaning I need to set tight deadlines for myself, otherwise I procrastinate too much.
JULY 2, 2026 - THE PAUSE
TheoryCraft is a project I really care about. The problem, and unfortunately I only realized it a few days after shipping the MVP, is that right now I do not have the financial resources or the time to run the project the way I want to.
So for now, I decided to put the project on pause. The platform is still available and functional, but until my situation improves, I have decided not to do marketing or keep moving forward on this MVP.
Instead, I am going to focus on smaller projects that will let me learn quickly and, who knows, maybe one of those side projects will make me rich and let me reinvest in TC :)