A little story about how All started
Honestly, I never expected this journey to take so much time It started with a Binance trading bot. We ran into some problems with calculations and the way certain things worked, and at some point we thought why not build our own bot?
We thought it would be easy because we knew how to code. Turns out building a trading system is a completely different story
One problem led to another. We faced engineering challenges, strategy problems, execution issues and plenty of things we hadn’t even thought about before
From the beginning, we had one goal. Keep everything as close to zero cost as possible. We didn’t want expensive API calls or cloud infrastructure. We wanted something that could run with just electricity and WiFi
And we wanted to do it on a regular Mac Mini with 8GB RAM. We could have looked for more computing power, but we wanted to see how far we could go with what we had
Then came local AI and Nemotron
We started treating it almost like a classroom. Nemotron was our student and we were the teachers. We prepared lessons, wrote teacher notes and kept experimenting with how to teach it about the market. That part was honestly a lot of fun
Of course, trading itself wasn’t nearly as fun We spent countless hours testing strategies and backtesting different market conditions. There were plenty of failures and liquidations. Across our experiments, we saw numbers ranging from around 70 to 120 liquidations
We kept coming back to TAO because we were genuinely curious about the market and its connection with AI One thing that kept us going was seeing the liquidation count in our testing come down from 78 to 16 Is 16 good enough? No. We still have plenty to learn and improve. But compared to where we started, it means something to us Looking back, we never imagined a simple bot idea would turn into months of coding, testing, learning and experimenting We’re not sharing this as a finished success story. Just a small part of a journey that is still going And honestly, we’re happy we didn’t give up
Honestly, I never expected this journey to take so much time It started with a Binance trading bot. We ran into some problems with calculations and the way certain things worked, and at some point we thought why not build our own bot?
We thought it would be easy because we knew how to code. Turns out building a trading system is a completely different story
One problem led to another. We faced engineering challenges, strategy problems, execution issues and plenty of things we hadn’t even thought about before
From the beginning, we had one goal. Keep everything as close to zero cost as possible. We didn’t want expensive API calls or cloud infrastructure. We wanted something that could run with just electricity and WiFi
And we wanted to do it on a regular Mac Mini with 8GB RAM. We could have looked for more computing power, but we wanted to see how far we could go with what we had
Then came local AI and Nemotron
We started treating it almost like a classroom. Nemotron was our student and we were the teachers. We prepared lessons, wrote teacher notes and kept experimenting with how to teach it about the market. That part was honestly a lot of fun
Of course, trading itself wasn’t nearly as fun We spent countless hours testing strategies and backtesting different market conditions. There were plenty of failures and liquidations. Across our experiments, we saw numbers ranging from around 70 to 120 liquidations
We kept coming back to TAO because we were genuinely curious about the market and its connection with AI One thing that kept us going was seeing the liquidation count in our testing come down from 78 to 16 Is 16 good enough? No. We still have plenty to learn and improve. But compared to where we started, it means something to us Looking back, we never imagined a simple bot idea would turn into months of coding, testing, learning and experimenting We’re not sharing this as a finished success story. Just a small part of a journey that is still going And honestly, we’re happy we didn’t give up
