For most of human history, tyranny had a natural check. A tyrant needs people. Soldiers. Workers. An economy that runs on human labour. Push too far, and the people who make your power possible can simply stop. Bittensor's co-founder, Const, thinks that check is disappearing. War is becoming mechanised fast enough that a government could lose 90% of its workforce and barely notice the economic impact. When labor and intelligence both get automated, the leverage ordinary people have always held against concentrated power quietly evaporates. He calls it a risk of totalitarianism the world has never seen before. Not because anyone is plotting it. Because nothing is stopping it. That's the fear behind Bittensor's entire thesis. If AI ends up controlled by one company, one government, or a handful of insiders nobody can even name, everyone else just inherits whatever worldview gets built in. No vote. No audit. No way out. Here's the idea they built instead. Bitcoin proved something strange. A permissionless, competitive market of anonymous participants can out-produce any centralised company on Earth. By Const's own estimate, the Bitcoin network does more raw computation than every supercomputer on the planet combined. Built for a fraction of what one tech giant spends per year. Bittensor points that same mechanism at intelligence instead of hashing. Miners worldwide don't just burn electricity. They train, fine-tune, and improve real machine learning models, competing every block to be the best at a specific task. Instead of one company dropping a new model every six months behind closed doors, it's thousands of people improving models continuously, every block, forever. Const's line for it: "Imagine Bitcoin mining, but you're mining for intelligence directly." Here's the part that should stop your scroll. This isn't theoretical. One Bittensor subnet was already serving more open source AI tokens than any single commercial provider on the planet. People were already using it. They just didn't know Bittensor was underneath it. Ownership matters here too. Const has no separate equity from his TAO tokens. No secret cap table. No venture round only insiders got into. Everything runs on a public blockchain anyone can inspect, including who owns what. Compare that to companies where you genuinely don't know who's really in control behind the structure. If this resonates and you want to go further than agreeing with it: Start here: go to (bitensor. com) and look at what a subnet actually is. It's simpler than it sounds, a market that pays people for one measurable output. You don't need to be a machine learning engineer anymore. Point an AI agent like Claude or Cursor at a subnet's docs and have it help you figure out mining. Have spare GPU capacity? Some subnets pay you for it directly, permissionlessly, no KYC required. Want exposure without mining? Every subnet's token has to be paired with TAO to function, so ecosystem activity flows back through it. Go slow. Verify everything yourself. This is early, permissionless infrastructure. Real upside sits right next to real risk. Const points to a past subnet rugpull as a reminder that decentralisation removes middlemen, not human error. The bet isn't that decentralised AI wins because it's righteous. It's that a handful of unaccountable companies deciding what every mind on Earth gets fed is a risk worth building against now, while there's still time to. Video credit @BitcoinNews