What I listen to.
3,000 miles a year with headphones in is a lot of hours. Most of it goes in one ear and evaporates. So I set a test: did any of it change what I actually did? Here's the rotation, and the thing I built out of it.
The rotation
Three shows have a permanent slot. Everything else rotates out.
All-In
Four rich guys arguing about markets. The value isn't the takes. It's watching someone size a position in public and then get graded by reality. So I graded them. The answer was no. Here's the case study.
My First Million
The best idea machine in podcasting. Shaan and Sam pull a business apart on air and you can hear the reasoning happen in real time. Half of what I've built started as a voice note taken mid-run.
Morning Brew Daily
Business news in the time it takes to warm up. I'd rather start the day knowing what moved than spend an hour finding out. Fast, funny, and it respects your clock.
The All-In portfolio
A podcast is an opinion until you score it. So I scored it.
The question
Four guys talk AI stocks every week to millions of listeners. Does any of it hold up? Not "are they smart." The measurable version: what happened to those names after they said it?
The method
Every episode transcript since ChatGPT launched, read end to end by Claude. Then the stock data next to it, daily, weekly, monthly, looking for where signal showed up and where it didn't.
The output
A portfolio built only from what survived, rebalanced quarterly. Rules written down first, so it can be wrong in public instead of quietly rewritten later.
Read the full case study: the falsified version, the two failed repairs, the anti-tuning rule, and the live book. Paper portfolio, no real money, no advice.
AI thesis portfolio
A second paper portfolio, run by an agent instead of a podcast.
One thesis, written down
$100k notional, one idea: AI is a general-purpose technology in its buildout phase, so own the layers where the value actually lands. Compute, foundry, custom silicon, memory, cloud, networking, datacenter guts. Weighted by moat, not by story.
Falsifiers up front
The thesis says what would prove it wrong before it can happen: hyperscaler capex flattening two quarters running, app-layer revenue stalling while depreciation piles up, an efficiency breakthrough killing demand. Write them first or you move the goalposts later.
The agent runs it
Claude Code rebalances it monthly on a schedule, marks the book, explains the movers, and journals every call. When a run gets degraded it says so and flags what went unchecked, instead of pretending it looked.