An AI engineer just did something most investors are too smart (or too cheap) to do: he spent $9,600 a year on 23 paid investment newsletters, tracked every single stock pick, and published the results.
Spoiler alert: most of them don't work. And the one that does costs $100.
The Experiment
The researcher subscribed to 23 paid Substack newsletters—including Michael Burry's—and built an AI system to extract high-conviction stock picks from 1,782 articles published over the past year. Then he tracked returns at multiple time horizons and compared them to sector benchmarks.
Total dataset: 3,519 stock calls. Total cost: $9,600. The goal? Figure out which newsletters actually make subscribers money.
The Winners
At the 30-day mark, Global Tech Research crushed it with a +14.9% average return on 50 calls. That's a $100/year newsletter outperforming almost everything else.
Second place went to Paulo Macro (+9.5%, 21 calls), followed by Collyer Bridge (+8.7%, 89 calls) and Doomberg (+7.8%, 79 calls).
But here's where it gets interesting: the rankings completely change at 60 days. SemiAnalysis jumped from +7.5% to +16.7%, and Fabricated Knowledge went from +5.8% to +14.2%.
Why? Because some newsletters are built for swing traders (quick in and out), while others are deep value plays that need time to work. If you're judging a 12-month thesis by 30-day returns, you're measuring the wrong thing.
The Expensive Losers
Price doesn't equal performance. James Bulltard costs $1,099/year. Lord Fed runs about $1,000. 10x Research is $948. None of them cracked the top five.
And then there's —yes, Michael Burry from . His newsletter costs $439/year. His 30-day return? . His 60-day return? .
