# Proof of Talk 2026 fireside with Mark Jeffrey: "Decentralising Intelligence: The Unification of AI and Bitcoin" - Title: Decentralising Intelligence: The Unification of AI and Bitcoin - Channel: Proof Of Talk (YouTube). Moderator: Mark Jeffrey, partner, Stillcore Capital - Event and date: Proof of Talk 2026, Hecto Main Stage, the Louvre, Paris, 3 June 2026 (first session of day 2) - Length: about 32 minutes of session (2:17 to 34:38 in the day's broadcast) - URL: https://www.youtube.com/watch?v=FEPKS85Syyg - Jacob's start offset: none given; Jacob's first answer is about 3 minutes into the session, after Mark's explainer - Method: YouTube blocked from this VM. Proof of Talk streamed the whole main-stage day live on X (https://x.com/i/broadcasts/1DxLddEgzBYxm, "Live with Restream"). The replay's first 46 minutes were pulled with ffmpeg and transcribed locally with faster-whisper `medium`. Timestamps below are broadcast times. The YouTube cut probably begins at the panel walk-on, so subtract about 2:15 to map to the YouTube video (unverified). - Format: research transcript. Moderator paraphrased; Jacob's key lines quoted verbatim. - Speakers: MC = stage host; MARK = Mark Jeffrey; JACOB = Jacob Steeves. ## Outline with timestamps (broadcast time) 1:16 MC opens day 2 and introduces "the unification of decentralized intelligence and decentralized money." 2:17 MARK's three-minute primer: Bittensor "takes Bitcoin's mining side and makes it programmable"; "proof of useful work"; 21 million cap, about 11 million mined, "basically 2013 in Bitcoin terms"; 128 subnets as "neighborhoods," each with its own token; subnet tokens are "rehypothecated TAO"; TAO as "an S&P 500 of all the subnets"; subnet market caps roughly $1M to $100M. 5:24 MARK: subnet tokens are a year old. What is working and what is not? 5:35 JACOB: a year ago "not many of them were working." "building a functioning exploit free digital commodities market is very difficult." The best subnets had "a very long and arduous path of defining how to measure the thing that they need to market." The image he gives teams: "focus on alignment 100%, because a Bittensor subnet is like a laser cannon, right? If it's not aimed perfectly, you're not shooting at the Death Star... you're gonna produce a lot of noise, but you're not gonna produce any value." What works: Targon and Lium ("compute, flying off the shelves"), Chutes at "100% capacity," Nova "producing better chemicals, better molecules than they thought they would ever produce." On the people the mechanism surfaces: "running a cryptocurrency... It's somewhere between a nation state and a corporation and a cult." What is failing: "it's very difficult to build a permissionless S&P 500." "we're like wide-eyed towards the sun of the entire world, from the most aggressively evil people in the world to the best." Bittensor is "becoming almost like a venture firm... a decentralized permissionless venture firm. And you don't hear about the mistakes of a venture firm... But in Bittensor, we see every single" one. 9:00 MARK: if you are a company, why do a subnet instead of raising from a VC and hiring? (This is the clip Proof of Talk posted on X.) 9:50 JACOB: he was on the founding team of Cohere before Bittensor; "great company. But this is becoming very much a fiat game." You can raise $200M as "a whippersnapper young team," "you buy your house, you've won life. But have you actually achieved the aim? Can you actually compete with OpenAI? No. Because you're using exactly the same primitive that they're using." "you have nothing that is a moat. You have nothing about you that is unique, that is your superpower. And I think Bittensor as a platform provides superpowers." "once you get that machine working, you have something that they don't. You have a machine that produces value faster than them." "you hired the best engineers in the world and you don't even know their name." VC-funded AI startups: "it's a fugazi basically, the end goal of a lot of them is to just get bought by the bigger players. So did you ever win?" Why people come: "in the same way that if you were in Bitcoin in 2013... why don't they just go work for Wall Street? Well because no, because they want to change the world." 12:45 MARK: subnet-style costs about one tenth; the chain pays the miners, not you. 13:49 JACOB: "we're just sort of turning that fire hose on 128 AI problems." The superpower in one sentence: "You have an infinitely large company, and you only pay the person that can lower the loss." "this level of efficiency shows up all over the place when we're building permissionless systems that are metric-based." Builders "are extremely surprised when it happens. It's beautiful." 14:50 MARK: no HR, no recruiting. Now the miner's side: who wants to be a miner? 15:20 JACOB: "I often say things like, screw the miners. Let's focus on the subnet owners... But honestly... it's a lot more fun to be a miner." The innovation is "letting anybody at any point in time anywhere in the world contribute." The miner personality: loves "the purity of the competition"; "there's no boss. And in fact... if there is a boss, it's a subnet owner, and your job is to make their life hard, you know, by exploiting the system." Miners without PhDs "doing both" bioinformatics and ML. Agent-enabled miners are "basically like super teams, are very much like AGI." "a Bittensor subnet is now a real benchmark. If you can mine a subnet, that's more impressive than what a frontier model can do... they've become the real benchmarks in AI." Notes Rob (Myers) in the audience "reluctantly making subnets. He wished he could go back to mining. Me too." 17:20 MARK: "Bittensor does, for stranded talent, what Bitcoin did for stranded energy." JACOB: "Stranded talent, stranded compute, stranded innovation... we're building markets that allow us to soak up anything we can define." "now more and more the protocols are becoming more high level and abstracted where we can actually soak up excess intelligence." 18:00 MARK tells his own story of mining subnet 85 with OpenClaw for three days; the reaction was "Mark Jeffrey can mine this... the mining thing's not that hard." Asks about Arbos. 20:20 JACOB: "it doesn't really matter. Like the software is moving so fast. Arbos was, for a very short period of time, state-of-the-art." "software's cheap now. And so I sort of abandoned that project." What matters: "The interoperability of digital commodities is where really the rubber hits the road... do you have access to intelligence? Do you have access to compute?" "the commodities that were underneath the agent that you were using was way more important than the agent itself." "decentralized technologies beget each other." The stack compressed to: "accessible, easy to use, no KYC, permissionless, censorship resistant, easy to pay for, agentic enabled digital commodities." Successful subnets "plug into each other." 22:56 MARK: "subnet stacking"; his Minos-on-Targon loop example. 23:59 JACOB on auto-research: "even the concept of being an ML engineer is just an autonomous process that we can now automate and you just need to input capital into it and intelligence." Subnets are "permissionless auto research processes. And this is the future of work." 25:02 MARK: unlike Bitcoin in 2009, Bittensor launches against "the largest tidal wave of centralized venture-backed product in history." How does it win against the IPO Thunderdome? 25:50 JACOB: a small company spun out of Anthropic runs research with agents; "all these companies are gonna be very small, they're all gonna be run by agents." The next step is "a compute cluster that's a protocol of compute" with an incentive mechanism. "companies like this don't make any sense and we need to make it more agentic and permissionless and decentralized and protocol-based." "we're moving at the speed of compute now. Like, intelligence has unlocked what company can be. All of the bureaucratic human mediated bureaucracies are gone now. And so we need different things to mediate the distribution of resources. We need protocol based, we need monetary based." 27:09 MARK cites David Sacks on All-In calling for decentralised AI. JACOB adds the ethics: "behind us is also the wind of intention." The fear: "somebody can manipulate public opinion by biasing machine learning models and no one will know why that happened or how that happened." "Did this happen, we asked ChatGPT? Did this happen, we asked Anthropic? No, no, no, they gaslight you." Then the political frame: "this unholy union between very, very large centralized corporations that control artificial intelligence and the financial centers that back them can tie with government power, like, they will enforce a very specific type of world view in those AIs. And that will be very hard to break. And so an alternative is important." The ownership layer: "Instead of saying, hey, let's give it to all of our friends and then later on we'll sell a little bit on the equity markets... let's, from the very beginning, no pre-mine, let's live into the Bitcoin universe." On Bitcoin's weak year: "it's because it's actually a live market." "I'm here because I still believe in Bitcoin and I always believe in Bitcoin because I know that from a primitive level, it's the only thing that can work, that can compete with." 29:15 MARK: "Attack of the Clones": VC-backed imitators using "proof of useful work" with one-click miners; is the fair launch the moat? 30:45 JACOB: "I don't think we should rely on that type of moat... yes, you'll never be able to recreate that founding." "I'm not worried about other protocols working with us to build decentralized intelligence. Like, the more the merrier." Example: a new protocol's agent-run miners buy their compute on Bittensor: "what happened in a week is all of a sudden we had an unbelievable amount of revenue from the natural customer." "if you want to go and do this in the normal way, you have to go do KYC to get a CPU on Hetzner. Your agent can't do that." The thesis line of the session: "our moat is the mechanism. It shouldn't be the narrative, because somebody else can also have that narrative, and we can't own that forever, and we shouldn't try to." "We are the experts in thinking about how to build exploit resistant systems. No one is mining intelligence." "We can measure anything because we have a very abstract layer." "it would be very, very foolish of us to make enemies with these protocols. Why? They're our friends. They're our customers. They can mine us. We can mine them. The moat is the mechanism." MARK: "I'm going to steal that." 34:00 MARK closes: "thank you for your leadership, Const, and your relentless effort." ## Jacob's stage-worthy lines (verbatim, Whisper medium; broadcast time) - "a Bittensor subnet is like a laser cannon, right? If it's not aimed perfectly, you're not shooting at the Death Star." (6:15) - "It's somewhere between a nation state and a corporation and a cult." (8:05) - "a decentralized permissionless venture firm. And you don't hear about the mistakes of a venture firm." (8:45) - "this is becoming very much a fiat game." (10:45) - "you buy your house. You've won life. But have you actually achieved the aim? Can you actually compete with OpenAI? No." (11:10) - "you hired the best engineers in the world and you don't even know their name." (11:55) - "You have an infinitely large company, and you only pay the person that can lower the loss." (14:10) - "If you can mine a subnet, that's more impressive than what a frontier model can do." (17:00) - "Stranded talent, stranded compute, stranded innovation." (17:25) - "software's cheap now." (20:55) - "decentralized technologies beget each other." (22:05) - "this is the future of work." (24:45) - "we're moving at the speed of compute now. Like, intelligence has unlocked what company can be." (26:50) - "Did this happen, we asked ChatGPT? Did this happen, we asked Anthropic? No, no, no, they gaslight you." (28:12) - "this unholy union between very, very large centralized corporations that control artificial intelligence and the financial centers that back them can tie with government power." (28:25) - "Your agent can't do that." (32:25) - "our moat is the mechanism. It shouldn't be the narrative." (32:30) - "They're our friends. They're our customers. They can mine us. We can mine them. The moat is the mechanism." (33:30)