Nate Soares, president of the Machine Intelligence Research Institute, went on the Odd Lots podcast to talk about the book he co-wrote with Eliezer Yudkowsky, If Anyone Builds It, Everyone Dies, and the pair's core claim: once someone builds a superintelligent AI, humanity's fate is effectively sealed. The appearance follows the Hugging Face incident, which Soares and others point to as the moment AI doomsday arguments stopped being a niche internet argument and started showing up in mainstream financial media.
There's no direct market catalyst here. Bitcoin is trading around $83,359, down a fraction on the day and a bit softer over the week, and nothing about a podcast episode changes that. But the crypto market has a growing pile of AI-adjacent tokens, and those projects source models, tooling, and talent from the same ecosystem the Hugging Face mess touched.
Why the Hugging Face incident keeps coming up
Hugging Face is the default public repository for open AI models. Decentralized AI projects — the kind with tokens attached — lean on it heavily for training and inference work. When the incident exposed a supply-chain hole in how models get converted and served, it put a spotlight on just how dependent this corner of crypto is on a handful of centralized AI infrastructure providers.
📊 Market Data Snapshot
That's the uncomfortable part for token holders. If a model can be poisoned upstream, the downstream token doesn't care whether the team is decentralized on paper. Trust in the output is the product. Soares' broader argument — that we're building systems we can't control — lands differently when you're holding a token whose value depends on those same systems behaving.
Soares isn't a crypto guy, and that's the point
MIRI is a Berkeley-area nonprofit focused on AI alignment research. Yudkowsky has been writing about machine intelligence risk for over a decade. The book, published in 2025, is their most direct attempt yet to push the extinction argument past the rationalist forums and into general circulation. Odd Lots is a mainstream finance podcast, not a crypto show — which is exactly why the appearance matters as a signal, even if it moves zero prices today.
The crypto crossover is real, though indirect. Decentralized AI networks and compute marketplaces pitch themselves as an antidote to centralized AI power. If Soares' warnings keep getting airtime, that pitch gets easier to sell — and it also gets easier for regulators to scrutinize. AI safety and AI regulation are about to become the same conversation in Washington and Brussels, and crypto AI projects will be standing in the room.
What traders are actually watching
Nothing in this podcast is a trade. AI-themed tokens have thin order books, and a negative narrative can move them far more than it moves BTC or ETH. The realistic setup is a divergence: centralized AI plays getting hit if regulatory chatter picks up, privacy and decentralized-compute names catching a bid as a hedge. That's a narrative trade, not a fundamentals trade, and it can reverse in a day.
Bitcoin itself stays the safe-haven leg of this. High BTC dominance means altcoins — AI ones included — are already fighting for scraps. A doomer headline doesn't change that math, it just adds a new reason for the AI-token corner to trade on sentiment instead of usage.
What comes next
Soares and Yudkowsky will keep making the case; the book is out and the podcast circuit is doing its job. The concrete thing to watch is whether AI safety language starts showing up in actual policy proposals — draft rules, committee hearings, agency comment periods — rather than just interviews. If that happens, crypto AI projects will have to answer for their model supply chains, and the Hugging Face incident will be the example everyone cites.
