A Blueprint for a Machine Learning Sports League
Smut for AGI enthusiasts
The year is 2050. The NFL is dead after concussions and CTE got worse. The new hotness? The Machine Learning Sports League (MLS). Kids no longer dream of throwing touchdown passes. They dream of submitting state-of-the-art models to the LMSys leaderboard and being featured on TBPN. Parents now brag about their child’s “token efficiency” instead of their 40-yard dash time.
MLS is tiered with teams competing for promotion and to avoid relegation. The Premier League is made up of juggernauts: OpenAI, Anthropic, Google DeepMind, and now xAI. xAI was recently promoted after their release of Grok’s ‘Bad Rudi’. Cursor, Cognition, and Reflection AI sit just below, pushing hard for promotion. Then there’s a massive long tail: Poolside, Sakana, Reka, and dozens of stealth startups grinding it out in Tier 3.
Player comp is through the roof. For years, the NBA paid 50% of league revenues to players as salaries. McKinsey’s AI market size was dead on: $4.4T in total spend, so the top engineers collect $2.2T in total salaries. Some are making $3-5B in annual compensation, making Meta’s $200M deal to recruit Ruoming Pang look like peanuts. MVP caliber engineers who win annual awards like NeurIPS best paper are eligible for salary accelerators. Labs package talent and compute in major trades that rock the industry.
The support infrastructure has professionalized too. Agents (AI agents) negotiate contracts, compute stipends, and intellectual property. Ari Gold from Entourage now represents high school ML engineers. Coaches help with model tuning and eval strategies and training facilities double as data collection centers. In a move that surprised everyone, Bill Belichick was bought out of UNC Football and took a machine learning coaching job in the Berkeley EECS program.
And once a year, the internet stops to watch the All-Star Game. Team Mark Zuckerberg vs. Team Sam Altman. Two elite squads of engineers go head-to-head in a weekend-long build-off. Models are trained, fine-tuned, and deployed live. Twitch streams top 1B concurrent viewers (half of them are AI agents). The last match ended in an upset, with Sam’s team solving a novel protein folding task while Mark’s team unfortunately got too drunk because he forced them to take shots every 5 minutes. David Fincher is directing a 30 for 30 about it on the back of this viral footage.
And of course, there’s Fantasy MLS. Kids draft rosters of real engineers based on leaderboard performance and win or lose points with each model release. Merch too… Seven year olds want to spend their money on ML Engineer trading cards. One kindergartner turned a $50 starter pack into a million-dollar card portfolio. His crown jewel? An original, 2015-mint Ilya Sutskever rookie card (in Ghibli style), graded PSA 10. It's currently listed on eBay for $1.08M, and he’s fighting Floyd Mayweather Jr. live on TBPN next week.