Token Loans and the Gentle Art of Grinding

Somewhere Between the Dojo and the Data Center

Jiu-jitsu has a way of rewiring how you see the world. You start noticing leverage in a bank's lending criteria, or the quiet patience in a startup's pivot. Last week, I read about Chinese banks handing out 'token loans' to AI companies, and I couldn't shake the feeling that I'd seen this before—not in a boardroom, but on the mats.

The Bank of China has already pushed out 8 million yuan to three firms, with another 20 million waiting in the wings. The collateral isn't property or equipment; it's the flow of tokens—the raw activity of AI usage. That's clever, but it's also a gamble. In jiu-jitsu, you don't base your game on your opponent's energy; you base it on their balance and their mistakes. A token loan feels like betting on a blue belt who's great at escaping but hasn't learned to finish. Sometimes that works. Sometimes you end up on your back wondering what went wrong.

The 90-Hour Week: A Cautionary Tale

OpenAI and Anthropic are reportedly pushing employees into 90-hour sprints. One former OpenAI tech staffer told me they were logging over 70 hours a week, and even after moving to a startup, it only dropped to 50-60—with the constant dread of a weekend Slack ping. That kind of intensity, I get it. To be good at jiu-jitsu, you have to drill until the movements are instinctual. But there's a difference: on the mat, you learn to breathe, to pace yourself, to recognize when you're about to blow a fuse. In tech, that self-awareness often gets buried under the next deadline.

A Berkeley study found that AI tools speed up individual tasks but also inflate the overall workload, because managers can't resist piling on more. It's the classic treadmill problem. In jiu-jitsu, if you go full blast every roll, you'll be spent in five minutes. The smart players conserve energy, pick their spots, and let the frantic opponent burn out. The tech industry could use a few more of those smart players.

When the Guard Pays Off

Not every company is sprinting. Google DeepMind is reportedly pulling back from the race to build the biggest models, focusing instead on efficiency. They're also bracing for layoffs that could cut a third of their team. That's not a retreat; it's pulling guard. It's saying, 'I won't play your game; I'll play mine from a safer position.' In jiu-jitsu, pulling guard is often seen as a defensive move, but it's also a strategic one—it sets up sweeps, submissions, and a chance to reset. DeepMind is doing exactly that, and it's a rare sight in an industry that worships speed.

Leaving Before You're Tapped

There's something to be said for knowing when to walk away. Anthropic's ethics lead quit after less than a year, citing concerns about surveillance and autonomous weapons. That's a bold tap-out. In jiu-jitsu, tapping out is a sign of maturity, not defeat. It's saying, 'I know my limits, and I'm not going to break myself for this.' The tech world often glorifies the grind to the point of self-destruction, so seeing someone prioritize their values over their career is refreshing—and a little rare.

Small Wins and Big Egos

On the flip side, there's the IPO lottery. Unitree Robotics' share offering had a success rate of 0.018%, with over 9.78 million applicants. One winner admitted they didn't post about it on social media, fearing envy. That's a lesson in humility—something jiu-jitsu teaches through getting tapped out repeatedly. But it's also a reminder that success is often a matter of timing and luck, not just skill. In jiu-jitsu, you can do everything right and still lose to a wild card. The same happens in business, and it's easy to forget when you're caught up in the hype.

The Uncomfortable Parallels

Here's where the analogy gets messy. Jiu-jitsu is a sport, not a war. The stakes are higher in tech, and the consequences of a bad decision ripple further. But the core principles—leverage, timing, pacing, humility—still apply. I'm not saying every CEO should roll on a mat, but the next time you're facing a tough call, it might help to think like a grappler: breathe, find your leverage, and move with intention. Or, you know, just take a break and come back when you're less likely to make a stupid mistake.