TD-Gammon: first superhuman AI backgammon
In the 1990s, Gerald Tesauro created TD-Gammon, the first neural network to play backgammon at a superhuman level, pioneering reinforcement learning AI.
Snapshot · Freakonomics Radio
677. Can Backgammon Save Us from Ourselves?
Explore episode
Jun 12, 2026
Where this was said
Neural Networks Change Everything
At 45:06 · chapter starts 42:30
Bob Wachtel had spent years accumulating a foot-high stack of position cards he couldn't solve even through extensive rollouts. When neural networks arrived, he finally got the answers — pure intellectual bliss. But the same tools that enlightened him revealed the truth to his opponents. A lawyer in Marin County who had lost to Wachtel for five years straight had convinced himself they were basically equals; the neural net analysis proved otherwise. Once weak players could see their error rates, many stopped playing for money. Mark Olsson contextualizes the technology: TD-Gammon, created by Gerald Tesauro at IBM in the 1990s, was the first neural network to beat the world's best human players at any board game — roughly contemporaneous with Deep Blue vs. Kasparov in chess. Backgammon was the ideal test environment because it was complex enough to be meaningful but small enough to simulate infinitely. Today, Backgammon Galaxy is built atop a successor to TD-Gammon, and the game at the elite level has shifted from inspired genius plays to something closer to the zero-error-rate standard of gymnastics.
How Neural Networks Killed the Money Game
Before neural network analysis tools, bad backgammon players could lose for years and still believe they were near-equals to the pros. When the AI arrived, it gave everyone an error rate — and the illusion shattered. The money game dried up almost overnight.
TD-Gammon: The Birth of Superhuman AI
Gerald Tesauro chose backgammon as the case study for his new reinforcement learning algorithm in the 1990s because the game was complex enough to be meaningful yet small enough to simulate infinitely. TD-Gammon became the first AI to beat the world's best human players — predating AlphaGo by decades.
Similar snapshots
MVP built in one week
Sam's initial MVP was coded in approximately one week using ChatGPT voice mode and copy-pasting code, with no prior technical experience.
Discord vs email for young users
Sam argues Discord is 10x better than email for building relationships with younger users who rarely check their inbox.
Monthly infra costs ~$780–$980
Sam's monthly operating costs include Cursor ($200), AI image generation ($100), AI video generation ($200), hosting ($100), email marketing ($80), and AI compute ($300–$500).
ChatGPT-powered pain point research
Sam recommends copying days of Discord chat history into ChatGPT and prompting it to list recurring pain points as a fast, free market research technique.
50 free tools built total
Bhanu and his team built approximately 50 free tools to attract search traffic, each linked back to SiteGPT.
New tool built in under 5 minutes
With AI coding tools like Cursor, Bhanu can now create a new free marketing tool in less than 5 minutes by referencing existing tools.
KD filter < 10 for rankable keywords
Bhanu filters Ahrefs keyword results to show only those with a keyword difficulty below 10, making them realistic ranking targets for any decent website.
Minimum 1,000 monthly search volume
Bhanu sets a minimum search volume of 1,000 monthly searches when selecting keywords to target with free tools.
20 downloads/day at launch
PropGPT averaged 20 downloads per day right after launching on the App Store through influencer marketing.
