I wanted to see what Claude was capable of and I have a keen interest in football statistics. So I started a new side project called False 9, it was meant to take a few days to get some interesting stats. It escalated. I started off manually syncing from the data provider.

Then I thought wait a minute let's automate this, now the site updates itself and syncs football score every minute they are playing. Then I decided to see if I could predict scores, I got Claude to create various prediction models. These were kinda cool they all used different signals to try and predict football scores.

I decided to automate adjusting the signals based on previous results so I got Claude to do a cron task to manipulate the signals and see if I could improve predictions. It worked quite well.

I got incredible juice out of a per month $20 budget but unfortunately I would need to spend more which I'm not willing to do. You can't get positional data for example without spending considerably more. Anyway if you like football stats check it out

https://false9.uk/teams/40?season=2025&league=39&tab=conceded&compare=42

Your Shazzer profile page now evolves the more public vectors you have. It changes colour, increases connected circles and your level changes.

https://shazzer.co.uk/profile?id=6605740959c283fb174e7998

You may have noticed I've been a bit quiet on social media recently, this is why...I'm going to present at Black Hat!

Can't wait to present these techniques! Here is a link to the abstract in case the screenshot is hard to read:

https://blackhat.com/us-26/briefings/schedule/index.html#css-the-bomb-inside-your-inbox-51909

Added AI features to Shazzer using Chrome's local model. They aren't very useful yet because the local model is very slow and isn't very smart but should improve over time when the model is updated. I've added:

- AI write description
- AI generate vector
- AI generate variant