Thoughts, Notes and Ideas of Nath Aston

Safari making a comeback

I have finally come full circle and made the move back to Safari.

I've spent years thinking anyone who uses Safari must be an idiot. Now I have to admit, it's really quite good. After moving from Arc to Zen I've finally come full circle and just settled on Safari.

Performance feels miles better than Zen which did cause CPU spikes on M4 MBP. But in all honest, the UI is just really clean and everything just works, nice tab layout. Everything is positioned at the top of the page, and as much as Arc's design did feel innovative and mind-blowing for a period, Safari's layout does give you more screen real estate and just keeps all clutter out the way so that you can crack on.

I will also say with Arc I did try to setup systems with things like PARA and difference spaces for different projects. In the end, it was just procrastination and I never really received any value from it. You just wasted time sorting tabs.

The only thing I do miss on Arc is their AI Tab Sorting, which would group tabs together. Other than that Safari honestly feels great, and is beautiful to work on thanks to Liquid Glass update.

Once again, I'm reminded and rewarded that simplicity beats complexity unless complexity brings 10x the value.

Machine Learning is pretty cool.

Have been working with SciKit-Learn and Pandas (for data analysis) of my Reddit Investment Finder.

SciKit is my first exposure to ML

Using Codex CLI with GPT 5.2 (High) which is my favorite coding model by a mile. I don't understand why everyone loves CC so much, it's good but in my opinion spits out the most slop and doesn't really adhere to instructions the longer you work with it. It feels like an over-excited intern, who will get the job done but will make a mess doing it that you need to clean up after.

GPT5.2 is much slower, but it just feels more precise, it scans and understands your codebase and the context around the changes it will make, and how each piece of code ties together to not break other things with it's changes.

Anyway, after analyzing 1 strategy with 2 years of data, ML spit out about 12,000 suggested recommendations / patterns it noticed. So I've been working my way through those as a lot of the suggestions weren't supported by our current functionality so involved ranking by confidence then building out so we could facilitate the backtest.

The %s below are the improvement it saw over the baseline. With the top strategy generated a 213% return.

I'm still refining this down and working on improving the data sets that I feed into the ML.

One other really interesting thing is that you can train your own model, then export the artifacts, so based on the data you've trained on give a score of how likely this stock is to go up based on the data/patterns you've trained on.

Meaning of life

I still spend a lot of time these days questioning the meaning of life and "what's my true purpose"

After retiring at 28, then quickly becoming bored I have decided to continue working because I truly love the things that I work on. But still I find myself looking for a deeper meaning.

I often look at ultra successful people, movie stars, billionaires and question what is their life meaning? If they are 50 - 60 years old, they have maybe another 10 - 20 years of a high quality life, why do they choose to spend time in the way they do, what's important to them? what's their motivation?

To continue in the pursuit of more wealth, more fame and more approval seems unfulfilling to me. But still I haven't found the answer either of something that truly calls to me. So for now, back to the grind

Hello World

I've decided to launch this blog as a place to note down my own thoughts and ideas. Something similar to Twitter but I find Twitter to be so distracting, every time I open it I feel I'm being radicalized and waste 30 mins scrolling through videos of the worlds atrocities.

Currently my primary focus is on the AI Investment finder, which I've integrated Machine Learning into for mining rules. Have achieved 320% returns over the past year which I'm quite satisfied with and have now backtested over 3,000 different strategies.

I'm currently building a library of all the optimizations, and the goal will be to give agents actions to that library so that they can actually build and construct their own strategy ideas by mixing and matching up different optimizations/strategies in the library.