These days, while suffering from the aftermath of Trump's troubles in the stock market, I learned about something called the TACO Index.
I heard that my American friends have already memeified this and mentioned it a lot.
So, I made one by slightly imitating https://www.pizzint.watch/...
This index compiles information such as S&P 500, Nasdaq index volatility + VIX + treasury bond yields + dollar index + oil prices + interest rate hike probability (referencing Polymarket).
I was already crawling US stock-related news and collecting brief summaries of the news content using AI (using a 3090 and gemma4 for inference). Since I'm collecting time series data, I judge whether it's TACO or not by running agy (antigravity) cli json output once on the corresponding date. If a comment appears, I save it back to the database as a time series.
I'm using a mix of MongoDB and Supabase (PostgreSQL) for the database, and for daily news aggregation, I'm aggregating MongoDB raw data into DuckDB, analyzing it with AI (qwen3.6 local), and then saving it back as a time series in the form of headlines and lead articles.
This way, I've been able to accurately capture how TACO came about, and so far, there haven't been any omissions.
I'm using AI to create all sorts of stock-related things. Now that I have a lot of functions, I'm studying how to use and analyze them.
It really feels like with enough motivation, you can do anything with AI these days.
