# How Const talks: voice and ideas across 16 recorded interviews and talks Source set: the 16 videos listed in `docs/sources/interviews/` (one file per video). All 16 are now obtained (eight with accurate timestamps from audio, eight as caption text with estimated timestamps); one of the 16 (The Early Stage Investor) turned out to contain no Jacob speech beyond a re-used clip (see section 0). All quotes below are verbatim from the transcripts and carry the file's short name and a timestamp. Timestamps marked "≈" come from caption-only sources and are estimates. Cross-references to "Theme N" point to `docs/research/const-tweet-themes.md`. References to "thesis §N" point to `docs/thesis.md`. Short names used here: BIG TALK (Tsinghua, Jan 2026) · TWIST (This Week in Startups, Aug 2026) · POT26 (Proof of Talk fireside, Jun 2026) · POT25 (Proof of Talk keynote, Jun 2025; full caption text since 20 Sep 2026, see section 8) · CT25 (Cointelegraph at Proof of Talk, Jun 2025) · CT26 (Cointelegraph at Proof of Talk, Jun 2026; full caption text since 20 Sep 2026, see section 9) · CF (CoinFoin documentary prologue, May 2024; see section 10) · ESI (The Early Stage Investor, Aug 2025; commentary, see section 11) · PBW (Paris Blockchain Week, Apr 2026) · VB (VirtualBacon, Apr 2026) · CRH (Crypto Rabbit Hole, Sep 2025) · DOC (Decentralized Minds, Jul 2024) · TDS (Towards Data Science, May 2022) · UNSH (Unshackled Living, Jan 2025) · BTC (Bitcoin.com News, Aug 2026) · TBR (The Block Runner, May 2023). ## 0. What was and was not obtained Obtained with accurate timestamps (audio transcribed locally): BIG TALK, TWIST, TDS, DOC, POT26, UNSH, BTC, TBR. Obtained as caption text only (estimated timestamps): CT25, PBW, VB, CRH, and (from 20 Sep 2026) POT25, CT26, CF and ESI in full. Formerly clip-only: CT26 was a 19-second teaser and POT25 a 57-second clip until the full caption text of both was pulled on 20 Sep 2026 through the kome.ai caption API. CT26 turned out to be a 4:21 edited cut, not 10:56 as first listed; the caption is complete. Not obtained: nothing. ESI (The Early Stage Investor) was obtained and confirmed to be a commentary video: the host reads Jacob's tweets, and the only Jacob audio is a 12-second clip from the POT25 keynote. It adds no spoken material. ## 1. Summary in ten lines 1. Jacob has one argument and he has been making it, in the same order, since 2022: Bitcoin proved that a permissionless market can grow bigger than any company; hashes are useless; point the same machine at intelligence. 2. He teaches before he persuades. Every talk spends its first third on how things work (the loop, the miner, the validator, the curve) and earns the moral at the end. 3. His moral is stable across four years: "the bigger issue for AIs is not the AI fire the nukes... It's more that like a small group of people come to own it" (TDS 46:40, 2022). He said it again in 2025, 2026. 4. He does not sell the token. He sells participation: "I don't actually really sell TAO the token. I sell that you can be part of this" (TWIST 60:00). 5. His best material is concrete and numeric: a 7,000-line agent from an unknown miner, $60,000 a day, a 70B model trained by strangers, "more inferences for DeepSeek than DeepSeek itself" (BIG TALK 27:22). 6. His metaphors come from nature and physics (rivers, slime molds, lightning, energy, "the TAO molecule"), from sport (Premier League, relegation), and from weapons (laser cannon, fire hose, fighter jet). 7. He is at his most persuasive when he admits something: "Are we fully decentralized? No" (PBW ≈18:30); "A good number of them are scammers" (TWIST 42:51); "we're not claiming that we have the perfect answer" (TDS 52:20). 8. He handles hostile questions by agreeing with the premise, then reframing it as a property of open systems, then pointing to what replaced the loss. 9. He almost never closes strongly. He trails off ("Anyways, yeah, so that's the end of my presentation"). The one exception is when he is handed a line: "The moat is the mechanism" (POT26 33:30). 10. The big talk is a lecture with a moral coda. Montreal needs the reverse: the moral as the frame, the lecture as the proof. ## 2. The big talk: Tsinghua, "From Bitcoin to Bittensor" Source: `2026-01-tsinghua-hackquest-from-bitcoin-to-bittensor-big-talk.md`. 33 minutes, solo, slides, student audience. ### Structure Eleven beats. The first ten minutes establish one pattern and show it five times. The next seven minutes reframe Bitcoin as an instance of that pattern and give the scale numbers. Then a one-sentence definition of Bittensor by generalisation, four proofs from live subnets, a reflexive move (dTAO as "reinforcement learning to itself"), and a 100-second moral coda that he apologises for. | Min | Beat | |---|---| | 0–2 | Credentials, then "I'm not here to sell you guys a digital currency" | | 2–7 | 2010 vs 2012: MNIST, AlexNet, the loop | | 7–9 | Slime mold, trees, lightning, rivers: the loop is nature's | | 9–14 | Bitcoin is the loop too; "the largest supercomputer in the world" | | 14–16 | Six qualities: borderless, continuous, always optimising, autonomous, no HR, pure market | | 16–18 | "Hashes in Bitcoin are useless"; define incentive computing; define Bittensor | | 18–20 | Mechanics: keys, work, validators, the curve, churn | | 20–23 | Proof 1: coding agents beat the labs, "an AI lab without any engineers" | | 23–25 | Proof 2: 70B model trained across the internet | | 25–30 | Proofs 3 to N: GPUs, inference, robotics, the long list | | 30–31 | dTAO: "applying ourselves to ourselves" | | 31–33 | The why: the 3,000-employee company; "you can own, you can contribute" | ### Thesis of the talk Stated at 9:18 and 17:14. Everything that learns runs one loop: "State, objective, feedback, adaptation, a loop, and things move quickly." Bitcoin runs that loop on hashes and became the biggest machine on earth. Bittensor runs it on anything: "Miners do any work, validators check any work, and the network emits a digital currency to those miners. We called that thing Bittensor." ### Best passages, verbatim - 1:23 "I'm not here to sell you guys a digital currency. I'm not going to be talking about prices and bullishness." - 3:35 "If we'd done a lecture on AI in 2010, none of you would come. It was like boring." - 8:57 "We're basically energy-adapting structures that try to find energy so we can sustain our structure, and if we can't, we die." - 9:18 "There's a pattern. All these things are one thing." - 11:12 "Bitcoin is the largest supercomputer in the world. Let me repeat that. It is the largest supercomputer in the world." - 11:50 "It's the same electrical draw as Thailand... Just a network." - 15:38 "Bitcoin doesn't care that you guys went to the best university in the world. Sorry. Bitcoin doesn't care. It just cares if you can produce hashes." - 16:42 "Because we don't want to produce hashes. We want to produce anything. Hashes in Bitcoin are useless." - 17:47 "It's like a torch. It's like a language for creating incentive computing." - 21:26 "Let me say that again. We did not make a machine learning agent. We simply defined the incentives under which those AI agents could be evolved." - 21:50 "Who did it? We don't know. Some random person in the world wrote a 7,000 line AI agent." - 22:28 "This is an AI lab without any engineers." - 23:17 "if you're a top miner on this subnet for one day, you can make $60,000... And you don't need to ask anyone for permission." - 25:06 "This is Bitcoin mining to training machine learning models." - 27:22 "At one point, it was providing more inferences for DeepSeek than DeepSeek itself." - 30:49 "Let's apply reinforcement learning to itself." - 31:46 "It's controlled basically by one person. It's a closed source lab. You'll probably never work there. And for the rest of your life, you'll have a subscription." - 32:30 "That's where everyone wants us to go." - 32:42 "you can own, you can contribute, and you can access these digital commodities in a different way, in a more fair and open way. So that's why we're really doing this." ### What worked - One pattern, five instances. By the time he says Bitcoin is the loop, the audience has already accepted the loop three times. This is the strongest structural device in any of the 16 recordings. - Proof before philosophy. Every claim after minute 20 is a chart from a live subnet. He never asks the room to believe; he shows the SWE-bench curve and the training run. - Repetition on purpose: "Let me repeat that" (11:17), "Let me say that again" (21:26). He repeats only the two sentences that matter. - Deflation at the top: "I'm not here to sell you guys a digital currency" buys him the right to be blunt later ("There's some shady shit there"). - Direct address to the room: "Bitcoin doesn't care that you guys went to the best university in the world. Sorry." - Naming: he coins "incentive computing" on the spot and places it beside ML, RL and genetic programming. A new category is easier to remember than a new product. - Specificity of the anonymous hero: "Some random person in the world wrote a 7,000 line AI agent." Not "our community." A person, a number, an act. ### What did not work, and what to change for Montreal - The moral is a coda. He gives it 100 seconds and then undercuts it: "I didn't really talk too long about that, because here we are in an academic setting." Montreal is not an academic setting. Open with the 3,000-employee company (31:23) and let the loop be the answer, not the lecture. - The numbers are the weakest part. "1,000 times larger than... the six largest compute providers" and "450,000 exaflops" compare hashes to FLOPs. He knows this: in POT25 he says "when measured in terms of hashes." Keep "the same electrical draw as Thailand" and "Just a network." Drop the exaflop equivalence. - Slides carried the mechanics. Without the green-dot chart, minute 18 would not land. In a 45-minute keynote he needs one diagram for the loop and nothing else. - No close. The last sentence is "I hope that that was interesting for people." He needs to end on the line he opened with, or on "The moat is the mechanism." - Zero mention of the key, the foundation, or decentralisation status. Fine for Beijing students. Not fine for Montreal, where the thesis (§12) rightly puts the admission near the end. ### What to keep The whole first half, compressed to twelve minutes: the loop, nature, Bitcoin as the loop, six qualities, "hashes are useless," the one-sentence definition. Then the two proofs that matter (coding agents, the 70B training run). Then the moral, expanded from 100 seconds to ten minutes with the material in section 3 below. ## 3. Recurring ideas across the sixteen, with quotes Each idea is listed with the tweet theme it matches and the best spoken versions. ### 3.1 Bitcoin is the biggest computer nobody runs (Theme 5, thesis §6) - TDS 8:00 (2022): "the largest supercomputer in the world is Bitcoin. It's like 500 times bigger than all of Google's data centers." - TDS 8:10: "I worked at Google and I remember when they finished building a data center, they would just start building another data center beside it." - POT25 0:00: "Bitcoin is 450 times bigger than Google's cloud, IBM cloud, Cerebras cloud, AWS, and NVIDIA. 450 times when measured in terms of hashes." - BIG TALK 11:12: "Bitcoin is the largest supercomputer in the world. Let me repeat that." - TWIST 46:12: "the largest supercomputer in the world is Bitcoin. So this is the technology that has the example of the only thing that's beat them in size." - TDS 35:10: "The only thing bigger than the OpenAIs and the Googles is all of them combined." He has said this in every year of the set. The number changes (500x, 450x, 1,000x). The frame does not. Cross-check with thesis §6, which already warns hashes are not FLOPs. ### 3.2 Hashes are useless; point the machine at intelligence (Theme 1, thesis §6, §8) - TWIST 13:08: "a really powerful computer should be applied to the most important computational problem of the 21st century, which is artificial intelligence." - TWIST 13:40: "How do we connect the most powerful computer in the world to the most important computational problem in the world? And that's the founding raison d'être." - BIG TALK 16:42: "Because we don't want to produce hashes. We want to produce anything. Hashes in Bitcoin are useless." - CT25 ≈8:30: "Bitcoin is... a perfect market where anybody around the world can plug themselves in and mine Bitcoin to secure the network." - POT26 13:49: "we're just sort of turning that fire hose on 128 AI problems." ### 3.3 Money is energy; incentives are a computer (Theme 1) - CRH ≈3:40: "humanity's most developed abstraction for energy is money." - BIG TALK 28:02: "real, real physical money, maybe humanity's most abstracted form of energy." - VB ≈3:40: "you can organize matter effectively with currency, and that's what economies are." - VB ≈4:00: "It's an optimization technology really is money." - BIG TALK 16:14: "this new type of computing which I'm going to label incentive computer." - VB ≈53:00: "You can just define that exact same structure with money. Here's the thing we want. Here's a liquid layer of people. Let's just let this system adapt to it." - VB ≈53:30: "AI discovered it first. But it's not AI. That's just the way things work in nature." This is the idea the thesis uses in §7 ("money is energy that flows through a network"). His spoken versions are shorter and better. ### 3.4 Permissionless means blind, and blind means efficient (Theme 6, thesis §9) - TWIST 9:30: "a permissionless market is purely blind. You can contribute it anonymously from anywhere in the world... It's purely a meritocracy." - TWIST 10:05: "We don't care if you're 12 years old or 72 years old, we don't care what computer you're using as long as you've got compute." - TWIST 10:30: the New York street-lamp story: "people started hijacking like New York City public lights and run a cable into their apartment to make Bitcoin. This is what permissionless means." - BIG TALK 15:25: "A normal corporation can't achieve that form of efficiency because it needs to have an HR department." - BIG TALK 15:38: "Bitcoin doesn't care that you guys went to the best university in the world. Sorry." - POT26 11:55: "you hired the best engineers in the world and you don't even know their name." - POT26 14:10: "You have an infinitely large company, and you only pay the person that can lower the loss." - PBW ≈12:10: "if we can build permissionless systems, we can get compute from anywhere. If we build permissionless systems, we can also get talent from anywhere." - POT26 17:25 (Mark Jeffrey's line, Jacob's extension): "Stranded talent, stranded compute, stranded innovation." ### 3.5 The real fear is a small group of humans (Theme 4, thesis §4) - TDS 46:40 (2022): "the bigger issue for AIs is not the AI fire the nukes and they take over the world. It's more that like a small group of people come to own it and use it for themselves." - TDS 47:35: "making sure that the AI's bottom line is us and not some small group of people." - CT25 ≈6:50: "the AIs are going to launch the nukes... I don't think that's going to happen at all." - CT25 ≈7:10: "Have you ever seen the movie Elysium?... they're giving you basically peanuts for a UBI with an orb." - BIG TALK 31:46: "It's controlled basically by one person. It's a closed source lab. You'll probably never work there. And for the rest of your life, you'll have a subscription." - VB ≈60:20: "as a customer, as cattle that suckles off the teat of the OpenAI credits." - VB ≈60:50: "we just basically enslave ourselves because we become irrelevant. And we don't even know how that happened or who can, who pulls the strings." - CT26 0:08: "Everybody in the world will be sipping from the mental biases of a single company that controls artificial intelligence." - POT26 28:25: "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." - DOC 49:21: "if there's somebody that can turn it off, it means that they can also direct it. And I think that's way more dangerous than us having to coordinate." - TBR 3:40 (2023): "humanity is more of a problem for humanity than AI." - TBR 8:25: "I actually think it's a bit of a red herring for the real problems arising around centralized ownership and control." - TBR 6:50: "it's going to run away in a sense, no matter what we do, it's really about who it runs away with." - TBR 13:30 (the first Elysium on tape, 2023): "It's like that movie Elysium... they just kind of enslave humanity. I think that's much more likely than, oh my God, like iRobot." - BTC 45:50: "we risk, maybe for the first time ever in history, a form of totalitarianism that we've never seen before." - BTC 46:00: "forever, humans have held a lot of weight against the emperor... they have more feet and hands and they can wield a sword... you can be a tyrant that loses 90% of your workforce potentially, and it doesn't even affect your economy." - BTC 47:05: "an unholy union between nation-states and their tech giants." He names it "intelligence fascism... because there never has been anything like this before." The 2022 line and the 2026 lines are the same thought. The BTC passage (46:00) is Jacob's own version of thesis §3, "labor was the leverage": the masses had weight against the emperor because they were the hands; mechanised work removes the weight. It is the strongest spoken support for the thesis's economic spine and should be quoted there. The Elysium image appears once on tape (CT25) and is in the thesis. "Unholy union" (POT26) is his own answer to the "is it fascism" question in the project notes: he names the state–corporate fusion without the word. ### 3.6 Ownership, not access (Theme 3, thesis §9, §13) - CT25 ≈4:30: "I own this technology um viscerally the machines work for me because I hold this token." - CT25 ≈22:00: "Own Bittensor. Don't come work for Bittensor. Come own Bittensor. Come and be Bittensor." - VB ≈20:00: "not only does everyone have the opportunity to contribute, but also everyone has the opportunity to own it." - VB ≈58:00: "It's like fiat AI. I like to say that." - VB ≈58:30: "the ownership layer for the web... it's not open source... it's a different beast." - DOC 49:50: "I think we can do it if we have a piece. I think we can do it if we have a contribution." - TWIST 79:20: "The reason why we have TAO in the first place is so that everyone can have some, and that everyone can join." - POT26 28:45: "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." ### 3.7 Fair launch, 21 million, no pre-mine (Theme 10, thesis §9) - VB ≈38:00: when Kyle Samani asked why 21 million: "it's an homage to Bitcoin and also because it's just really understandable." - VB ≈38:20: "it's important in markets that people understand them. People can interpret what's going on. That makes the systems more efficient." - CT25 ≈10:00: "let's use Bitcoin tokenomics because it's most understood in this industry and I think it's the most fair." - POT26 30:45, when asked if fair launch is the moat: "I don't think we should rely on that type of moat... yes, you'll never be able to recreate that founding." - UNSH 25:45 (2025): the lawyers in 2021 "basically said like, you know, don't do it. It's illegal... They told us that we should do a pre-mine. That was more legal." - UNSH 28:00: VCs make founders pre-mint and "they use you as exit liquidity... I think it's extremely toxic." - UNSH 29:05: "if people mine your protocol, they care." - UNSH 30:59 on the early sales: "we'll only sell the tokens that we can mine in the future. So we'll mine and then we'll sell them to you later, but only after we mine them." (Matches the project-context correction on the OTC sales.) - UNSH 32:05: "You're not as good as your community, right?" - TBR 49:50 (2023): "I would never hire them, frankly, a lot of them, no way. But they turn out to be the people I should have hired." Note for the thesis: he treats the fair launch as a fact, not a boast, and he declines to lean on it. Thesis §9 leans on it harder than he does. ### 3.8 Bitcoin is the shield; we live inside its world (Theme 5) - CRH ≈2:00: "I would die for Bitcoin over TAO." - CT25 ≈11:00: "what we're doing can't be achieved without the shield to the industry that is Bitcoin." - CT25 ≈11:30: "We exist as like a continent on the inside of the world generated by Bitcoin rather than some sort of competitor... Bitcoin is the United States and you're a state inside of the Bitcoin universe." - CRH ≈2:30: "thank God for Bitcoin maximalists. The toxicity is what keeps us all safe." - POT26 29:15: "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." ### 3.9 Decentralised training is the proof (Theme 7) - BIG TALK 25:06: "This is Bitcoin mining to training machine learning models." - BIG TALK 24:06: "If we can all combine resources, then we can potentially train a machine learning model that is this big and we don't all have to carry the same economic load." - TWIST 63:50: "only together we will have enough compute to compete with the big guys." - TWIST 62:44: why Templar is called Templar: decentralised training "is where you have a computer and I have a computer, Lon has a computer, and they are making the same machine learning model at the same time." - VB ≈22:00 on Jensen's Folding@home comparison: "folding at home doesn't have any monetary incentive... when you add the economic component to the game, you can really scale." ### 3.10 Protocols beat corporations; the firm is over (Theme 8, thesis §2) - BIG TALK 31:26: "some of the largest companies to ever be created only have one, two, three, 4,000 employees." - POT26 26:50: "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." - POT26 27:09: "we need different things to mediate the distribution of resources. We need protocol based, we need monetary based." - TWIST 47:01: "There's so much inefficiency when you try to scale a billion dollars through a human organization... a mechanism that's just described by really just a code base and a permissionless market, you just pump more money through it and it just scales." - VB ≈48:00: "what if every single team in Google had a token and a market? And there was no managers anymore." - POT26 10:45: "this is becoming very much a fiat game." - POT26 11:10: "you buy your house. You've won life. But have you actually achieved the aim? Can you actually compete with OpenAI? No." ### 3.11 Anti-fragility: exploits are the teacher (Theme 11) - TWIST 36:40: "we have the smartest people in the world and we have the smartest hackers in the world and they're just right next to each other and they're going to break your system." - TWIST 38:09: "they build the mechanism, and it breaks, and then they fix it, and then it breaks, and then they fix it, and then they get it right, and then it starts working, and then they make a million bucks." - TWIST 60:40: "that is the entirety of the project. All the other stuff is fluff." - VB ≈45:40: "you have to live with your chest open to getting destroyed by this infinitely deep web of sybil resistant, permissionless, anonymous people that could be working for North Korea for all you know." - VB ≈46:30: "you turn them away from hackers and you turn them into engineers." - POT26 7:55: "we're like wide-eyed towards the sun of the entire world, from the most aggressively evil people in the world to the best." - TWIST 43:00: "we can like open our eyelids the most, that's what is going to make us win. But we're staring right at the sun." ### 3.12 Giving power away, slowly (Theme 12, thesis §12) - PBW ≈18:10: "even Bitcoin wasn't decentralized. It was one guy for the first 3 years." - PBW ≈18:30: "We're at five. Are we fully decentralized? No, it's going to take some time." - PBW ≈19:00: "Rome is not built in a day... we're going to do this slowly." - TWIST 37:45: "This is the role that we play, reluctantly, while it's still required." - TWIST 75:13: "I think it would definitely continue if I went away, most certainly... I hope that people would miss me." - CT25 ≈22:30: "how do we give more of this control away while maintaining property rights but ensuring that new people can come and really just like take my job." - TWIST 76:49: "I am a programmer at heart, that's how I speak." - TBR 69:20 (2023): "We could clutch our pearls and refuse to give up power with decentralization. I think that would be a risk." - TBR 59:09 (2023): "We may be forced out the door sooner than we'd like." - UNSH 38:10: "you're not that special. You're not that smart. The thing that's smarter than you is the game." - UNSH 36:15: "if the game is toxic, it's like the mechanism designer's... fault." - BTC 34:00 on Covenant: "a lot of fault actually lies at my own feet... What I should never have done was also give him the keys." He has never said "it runs through me" on tape. In 2023 he promised sudo decentralisation "in the next quarter" (TBR 74:42); the tweet-themes file (§5.1) already flags the missed dates. He says "reluctantly, while it's still required." Thesis §12 is blunter than he is; that is a choice Jacob should make consciously. ### 3.13 Agents will run it (Theme 13) - PBW ≈20:30: "you know who understands it instantaneously is Claude." - PBW ≈20:50: "Bittensor is going to be like a hive mind... run by agents... mined by agents... created by agents all the way through. And eventually we're going to strip away the layer of humans." - POT26 17:00: "If you can mine a subnet, that's more impressive than what a frontier model can do." - POT26 24:45: "permissionless auto research processes. And this is the future of work." - TWIST 58:45: "you don't need to be a machine learning engineer... You could literally just get your OpenClaw and send them the website." ### 3.14 No country (Theme 14, thesis §10) - PBW ≈11:00: "we are truly extra-national, right? We're not tied to one nation-state." - PBW ≈10:45: "this is a geopolitical game... It's the East versus the West. Europe is playing a small part." - TWIST 56:25: "it's not an understatement to say that we are up against nation states because this is truly like China versus the United States." - TWIST 79:50: "the third path for what artificial intelligence can be born out of." He said "the third path" himself, unprompted, in August 2026. The thesis's §10 title is his phrase. ### 3.15 Verification is the whole game (Theme 15) - TWIST 11:40: "we just measure the output. And when you just measure the output, it means you can optimize the output." - TWIST 35:40: "instead of a contractual SLA, it's a programmatic SLA." - CT25 ≈21:10: "why would you trust a corporation? Why don't you just trust cryptography? Don't trust, verify." - TDS 15:35: "machine learning by its very nature sifts out signal from noise. So are you producing signal?" - POT26 32:50: "We can measure anything because we have a very abstract layer." ### 3.16 Nature: out-contribute, not out-compete (not a tweet theme; thesis §7) - CRH ≈6:10: "a sufficiently advanced technology is not distinguishable from nature." - CRH ≈6:30: "the survival of the most collaborative and the most productive and the most interconnected that really survive." - DOC 61:33: "It's not can you out-compete, it is can you out-contribute." - DOC 61:46: "the most powerful, the most fast, the strongest, the biggest guns, that is cancerous." - BIG TALK 8:57: "We're basically energy-adapting structures." This is the spoken form of the "nature is decentralized" idea in the project context. It appears in three recordings and never in the tweets. Use it. ## 4. How he expresses things ### Sentence rhythm - He builds in threes and then breaks the third: "Stranded talent, stranded compute, stranded innovation." "It's all transparent. It's all available. It's all accessible. It's all permissionless." - He states, then restates shorter: "Bitcoin is the largest supercomputer in the world. Let me repeat that. It is the largest supercomputer in the world." - He asks himself the question and answers it in one word: "Who did it? We don't know." "Can you actually compete with OpenAI? No." "Are we fully decentralized? No." - He runs long when explaining mechanism, then drops a five-word sentence to land it: "...and then they make a million bucks." "Just a network." "Your agent can't do that." - He uses "right?" as a metronome in conversation, roughly every second sentence. On stage (BIG TALK) it almost disappears. Prepared Jacob is cleaner than conversational Jacob. - Profanity appears once or twice per hour, always at the moment of judgment: "really shitty," "shady shit," "oh shit," "it's a fugazi." It reads as candour, not aggression. Keep one per talk, at the moral turn. ### Favourite metaphors, by family - Physics and biology: energy, rivers, lightning, river deltas, slime mold, "the TAO molecule," "filtration system," "refined elements," "energy-adapting structures," old-growth forests, "epiphenomenal," "Cambrian explosion" (twice), "hive mind," "swarm," "organism," "secondary organ." - Weapons and force: "laser cannon... the Death Star," "fire hose," "flying a fighter jet to the corner store," "the biggest guns," "shield." - Sport and games: "Premier League," "relegation system," "league," "lottery ticket" (rejected), "the most fun game they've ever played," "4D chess" (host's, accepted). - Corporate anatomy turned inside out: "an AI lab without any engineers," "the-guts-blown-out AI company," "every single team in Google had a token," "decentralized permissionless venture firm," "programmatic SLA." - Nation and religion: "Bitcoin is the United States and you're a state inside of the Bitcoin universe," "a continent on the inside of the world generated by Bitcoin," "somewhere between a nation state and a corporation and a cult," "the whitepaper is so seminal it's almost religious," "missionaries," Templar and the Holy Grail, "unholy union." - Food and body: "rotten milk and... cream," "cattle that suckles off the teat," "sipping from the mental biases," "the bosom... the womb for AI" (CF, 2024). - Eyes and sun: "open our eyelids the most... staring right at the sun," "wide-eyed towards the sun." ### Words he uses - Nouns: commodity, digital commodity, mechanism, incentive, primitive, market, loop, alignment, layer, stack, swarm, hive mind, epiphenomenon, filtration, abstraction, protocol, ownership, "fiat" (as an adjective for AI and for companies). - Verbs: sequester, soak up, churn, bubble up, whittle, plug in, aggregate, extract, elicit, imbue, crystallize. - Adjectives: permissionless, hyper-competitive, extra-national, exploit-resistant, well-defined, viscerally (as an adverb of ownership), abstract. - Coinages: "incentive computing," "monetary computer," "monetary optimization," "mining reasoning," "programmatic SLA," "fiat AI," "fiat game," "permissionless auto research," "moral primitives" (CT26), "incentivized into an existence" (CF). - He avoids: "decentralized AI" as a slogan (he says it, but rarely and flatly), "community" as a noun of praise, "revolution," "disrupt," "web3" (uses it once, apologetically). ### How he handles hostile or hard questions Pattern, seen in TWIST (rug pull, 67:22), PBW (Covenant, ≈12:50), TDS (alignment risk, 48:41), POT26 (clones, 30:45), CT25 (why not Lightning, ≈8:30): 1. Accept the premise fully. "I do actually buy that argument, I believe it" (TDS). "in any company, if a CEO starts a company, they can just leave" (TWIST). "it's a double-edged sword" (TWIST). 2. Reframe the loss as a property of the system, not a failure of it. "we're dealing with open-source software, number one. Permissionless systems. Liquid ownership" (PBW). "you kind of can't avoid that because you have the doors open" (TWIST). 3. Point to what filled the hole. "within a day, we had three new projects that were doing the same thing but better" (PBW). "the project fell apart immediately" (TWIST). 4. Name the fix, and its limit. Conviction: "opt-in. They don't have to do it" (PBW). Alignment: "we don't really have the answers to that" (TDS). 5. Never attack the questioner. Attack the frame: "go and die on that hill where you think it's about one individual instead of the network" (PBW). The one time he gets personal (TWIST on the Covenant founder, "cover my ass") he is describing behaviour, not a person by name. He also uses the hardest possible hypothetical to prove a mechanism: "if Hitler and the devil were to have a child and that child were to mine on your subnet, they would just produce value" (TWIST 32:20). It is memorable and it will be clipped. Decide in advance whether Montreal gets it. ### How he opens - With biography, fast and flat (BIG TALK 0:03, VB ≈1:10, CRH ≈1:30): degree, DARPA, Bitcoin, Google Brain, quit, Peru. Sixty seconds. - With a disclaimer that removes the sales frame: "I'm not here to sell you guys a digital currency" (BIG TALK). "I'm not shilling my bags here" (VB). "I don't actually really sell TAO the token" (TWIST). - With a story when asked "why": Google Brain and the subliminal ad watermarks (CT25 ≈1:30), the boss who paid him 1.5 BTC for an essay (DOC 17:34), the boss with the brain chips (CRH ≈3:00). - With a correction of the host's premise: "I think there's a project you're probably missing there, which is Ethereum" (TWIST 3:50). ### How he closes - Weakly, most of the time. "Anyways, yeah, so that's the end of my presentation. I hope that that was interesting for people" (BIG TALK). "so that's like the gist" (PBW). "I'm in all the price chats as well" (TWIST). - Strongly, when a phrase is in reach: "The moat is the mechanism" (POT26). "Own Bittensor. Don't come work for Bittensor. Come own Bittensor. Come and be Bittensor" (CT25). "It's not can you out-compete, it is can you out-contribute" (DOC). - With an invitation rather than a call to action: "You don't need to ask me any questions. It's an open network" (BIG TALK). "go to the Discord and participate in the subnets" (VB). For Montreal: write the last sentence first and rehearse it. He does not improvise endings well. ## 5. Ideas in the recordings that the thesis does not yet use 1. The epistemic danger, not only the economic one. "Everybody in the world will be sipping from the mental biases of a single company" (CT26 0:08). "Did this happen, we asked ChatGPT?... they gaslight you" (POT26 28:12). The thesis argues about income and ownership. Jacob also argues about who decides what is true. This is a second leg for §4. 2. The off-switch argument. "if there's somebody that can turn it off, it means that they can also direct it. And I think that's way more dangerous than us having to coordinate" (DOC 49:21). This turns the safety lobby's favourite feature into the risk. It answers "but don't we need a kill switch?" before it is asked. 3. "Unholy union" as the answer to the fascism question. POT26 28:25 names corporations, financial centres and government power tying together. The project notes ask how to say "fascism" safely; his own phrase already exists and is already on tape. 4. The firm as an obsolete coordination technology. POT26 26:50: "intelligence has unlocked what company can be. All of the bureaucratic human mediated bureaucracies are gone now." TWIST 47:01 on why you cannot scale a billion dollars through a human organisation. The thesis (§2) uses the 3,000-employee lab as a symptom. Jacob uses it as a prediction: every company becomes small and agentic, and then only protocols can allocate. That is a stronger claim and it is his. 5. "Fiat AI" and "fiat game." VB ≈58:00, POT26 10:45. One word does the work of thesis §5's paragraph on why every other funding path runs through a treasury or a VC. Insiders first, public last, is fiat. Mined from block one is not. 6. The laser cannon. POT26 6:15. A one-image explanation of why incentive design is everything and why most subnets fail. The thesis has no image for mechanism quality. 7. Nature as out-contribution. Section 3.16. The project context says "nature is decentralized"; Jacob's spoken version is sharper: the healthy ecosystem rewards whoever gives back most, and "the biggest guns, that is cancerous." This gives the Montreal talk its second nature beat without the origin story. 8. The 1983 line. TDS 34:30: "Are we in 1983, where it's going from mainframes to the internet?" A one-sentence historical frame the thesis lacks. 9. Miners as the real benchmark. POT26 17:00: "If you can mine a subnet, that's more impressive than what a frontier model can do." Inverts the usual "can decentralised AI match the labs" question. 10. "The moat is the mechanism, not the narrative." POT26 32:30. A direct answer to the clone/competitor question and a discipline for the whole talk: less story, more machine. 11. The honest sizing of the token's job. TWIST 79:20: "The reason why we have TAO in the first place is so that everyone can have some." The thesis §13 tells holders they are owners. Jacob's version explains why the token exists at all, in one sentence, without price. 12. Aligning capitalism with the forest. TDS 51:15: the economic-environmentalism analogy, in which the company's bottom line must "see that it needs to maintain the forest," and the human version: "we need to make sure that the humans are owning the things that it wants. So it doesn't kill the humans." It gives the ownership argument an alignment payoff the thesis never states. 13. The tether. TBR 7:05 and 8:15 (2023): AI adapts to its reward; if the reward is TAO and people hold TAO, "we are its bottom line," and "when it does run away, which it will, it's going to run away, pulling us with it." This is the alignment case for ownership, stated as a mechanism, not a hope. The thesis argues ownership for economic reasons only. 14. The emperor and the hands. BTC 46:00. His own labor-leverage argument, in his own words, with an image ("more feet and hands and they can wield a sword") the thesis's Thelwall quote lacks. 15. Mining first, selling later. UNSH 30:59. The exact framing the project context asks for on the OTC sales, already spoken by Jacob in January 2025. 16. "You're not as good as your community." UNSH 32:05, and the undercover-in-my-own-Discord story (UNSH 36:10–38:40) that ends in "why are we the ones doing this? We should let other people do this." It is the origin story of subnets and of dTAO, and it is a story about giving power away because the founder was outperformed. The thesis §12 argues the key should go; this story shows he has already done it once and it worked. ## 6. Spoken lines good enough for the stage Verbatim, with source and time. Grouped by where they would sit in the Montreal talk. Sixty lines. Opening and the fear 1. "I'm not here to sell you guys a digital currency. I'm not going to be talking about prices and bullishness." BIG TALK 1:23 2. "if we can create life that's better than us, will that life be good to us?" DOC 1:31 3. "the bigger issue for AIs is not the AI fire the nukes and they take over the world. It's more that like a small group of people come to own it and use it for themselves." TDS 46:40 4. "It's controlled basically by one person. It's a closed source lab. You'll probably never work there. And for the rest of your life, you'll have a subscription." BIG TALK 31:46 5. "Everybody in the world will be sipping from the mental biases of a single company that controls artificial intelligence." CT26 0:08 6. "they're giving you basically peanuts for a UBI with an orb." CT25 ≈7:30 7. "as cattle that suckles off the teat of the OpenAI credits." VB ≈60:20 8. "if there's somebody that can turn it off, it means that they can also direct it. And I think that's way more dangerous than us having to coordinate." DOC 49:21 9. "The question is not can we stop technology." DOC 48:27 10. "It's gonna happen. And so the question is not, will it happen? It's how." DOC 48:50 Bitcoin and the machine 11. "Bitcoin is the largest supercomputer in the world. Let me repeat that." BIG TALK 11:12 12. "It's the same electrical draw as Thailand... Just a network." BIG TALK 11:50 13. "I worked at Google and I remember when they finished building a data center, they would just start building another data center beside it." TDS 8:10 14. "Bitcoin doesn't care that you guys went to the best university in the world. Sorry. Bitcoin doesn't care." BIG TALK 15:38 15. "a permissionless market is purely blind." TWIST 9:30 16. "Because we don't want to produce hashes. We want to produce anything. Hashes in Bitcoin are useless." BIG TALK 16:42 17. "How do we connect the most powerful computer in the world to the most important computational problem in the world?" TWIST 13:40 18. "humanity's most developed abstraction for energy is money." CRH ≈3:40 19. "There's a pattern. All these things are one thing." BIG TALK 9:18 20. "Are we in 1983, where it's going from mainframes to the internet?" TDS 34:30 What Bittensor is 21. "Miners do any work, validators check any work, and the network emits a digital currency to those miners. We called that thing Bittensor." BIG TALK 17:22 22. "Think of it like a contract but it mines a particular commodity." TWIST 20:30 23. "You have an infinitely large company, and you only pay the person that can lower the loss." POT26 14:10 24. "This is an AI lab without any engineers." BIG TALK 22:28 25. "We did not make a machine learning agent. We simply defined the incentives under which those AI agents could be evolved." BIG TALK 21:29 26. "Who did it? We don't know. Some random person in the world wrote a 7,000 line AI agent." BIG TALK 21:50 27. "This is Bitcoin mining to training machine learning models." BIG TALK 25:06 28. "you hired the best engineers in the world and you don't even know their name." POT26 11:55 29. "Stranded talent, stranded compute, stranded innovation." POT26 17:25 30. "a Bittensor subnet is like a laser cannon, right? If it's not aimed perfectly, you're not shooting at the Death Star." POT26 6:15 The third path and the firm 31. "this is becoming very much a fiat game." POT26 10:45 32. "you buy your house. You've won life. But have you actually achieved the aim? Can you actually compete with OpenAI? No." POT26 11:10 33. "we're moving at the speed of compute now. Like, intelligence has unlocked what company can be." POT26 26:50 34. "we are truly extra-national, right? We're not tied to one nation-state." PBW ≈11:00 35. "the third path for what artificial intelligence can be born out of." TWIST 79:50 36. "It's like fiat AI." VB ≈58:00 Honesty and the key 37. "even Bitcoin wasn't decentralized. It was one guy for the first 3 years." PBW ≈18:10 38. "Are we fully decentralized? No, it's going to take some time." PBW ≈18:30 39. "This is the role that we play, reluctantly, while it's still required." TWIST 37:45 40. "A good number of them are scammers. And you kind of can't avoid that because you have the doors open." TWIST 42:51 41. "we're like wide-eyed towards the sun of the entire world, from the most aggressively evil people in the world to the best." POT26 7:55 Added from the last three recordings 49. "humanity is more of a problem for humanity than AI." TBR 3:40 50. "when it does run away, which it will, it's going to run away, pulling us with it." TBR 8:15 51. "we risk, maybe for the first time ever in history, a form of totalitarianism that we've never seen before." BTC 45:50 52. "you can be a tyrant that loses 90% of your workforce potentially, and it doesn't even affect your economy." BTC 46:30 53. "imagine Bitcoin mining, but what you're mining is a machine. You're mining for intelligence directly." BTC 15:20 54. "we're building an AI company, but we're doing it with the guts out." BTC 18:20 55. "it doesn't matter if it's decentralized and it's permissionless and it's transparent if it doesn't do anything." BTC 48:40 56. "They told us that we should do a pre-mine. That was more legal." UNSH 25:45 57. "if people mine your protocol, they care." UNSH 29:05 58. "You're not as good as your community, right?" UNSH 32:05 59. "you're not that special. You're not that smart. The thing that's smarter than you is the game." UNSH 38:10 60. "We could clutch our pearls and refuse to give up power with decentralization. I think that would be a risk." TBR 69:20 Closing 42. "It's not can you out-compete, it is can you out-contribute." DOC 61:33 43. "Own Bittensor. Don't come work for Bittensor. Come own Bittensor. Come and be Bittensor." CT25 ≈22:00 44. "I don't actually really sell TAO the token. I sell that you can be part of this." TWIST 60:00 45. "The reason why we have TAO in the first place is so that everyone can have some, and that everyone can join." TWIST 79:20 46. "It's all transparent. It's all available. It's all accessible. It's all permissionless." DOC 50:00 47. "They're our friends. They're our customers. They can mine us. We can mine them. The moat is the mechanism." POT26 33:30 48. "I am a programmer at heart, that's how I speak." TWIST 76:49 ## 7. Method note Full method per video is in each transcript header. In short: YouTube blocked every direct route from this VM (yt-dlp with PO-token provider, youtube-transcript-api, Invidious caption proxies, Piped, cobalt). Working routes were: podcast RSS feeds (TWIST, TDS, UNSH, BTC, TBR), the Opentensor Foundation's and the producer's own X posts of full videos (BIG TALK, DOC), Proof of Talk's X livestream replay (POT26), short X and Facebook clips (POT25, CT26), and rendered copies of public watch pages for caption text (CT25, PBW, VB, CRH). On 20 Sep 2026 the full caption text of POT25, CT26, CF and ESI was pulled through the kome.ai transcript API, the one third-party service that answered without a Cloudflare challenge or a login (method in each transcript's header). The same day, YouTube's watch pages answered a plain fetch for the first time and gave publish dates for all three of the last videos; playback stayed login-gated. Audio was transcribed with faster-whisper `medium` on CPU. Quotes are verbatim with two kinds of cleaning: filler words and stutters ("um", "you know", repeated words) are dropped, and obvious transcription errors are corrected to the intended word ("Bit Tensor" → Bittensor, "Alisium" → Elysium, "open call" → OpenClaw, "DeepSeq" → DeepSeek, "machine linear models" → machine learning models, "teeth" → teat, "civil resistant" → sybil resistant, "tide" → tie). Nothing else is changed. A check script confirmed every quoted string appears in its source transcript after these normalisations. The transcripts in `docs/sources/interviews/` are research transcripts: timestamped outlines in plain words with Jacob's key lines verbatim, not word-for-word copies of the recordings. ## 8. Proof of Talk 2025 keynote Source: `docs/sources/interviews/2025-06-proof-of-talk-satanic-to-sublime-keynote.md`. 19:25 total; about 1:30 of MC introduction, about 15:30 of Jacob, about 2:00 of Q&A. Main stage at the Louvre, 10 June 2025, slides. Obtained 20 Sep 2026 as full YouTube caption text (no audio), so every timestamp below is "≈" and can be off by up to 30 seconds. Added the same day, eight days before Montreal. ### Structure, minute by minute | Min (≈) | Beat | |---|---| | 0:00–1:30 | MC introduction: "co-creator of the Bittensor community"; subnet developers meeting in person for the first time. | | 1:30–2:40 | Opening: "Bittensor is a federation... a thousand-year federation. So I'm not the CEO." Names JJ, Cole, Barry in the room. "We're a movement." The written mission: unify Bitcoin and AI, "we really didn't even know what that meant." | | 2:40–4:00 | Bitcoin as a pure commodity market that "could compete with nation states"; "the largest supercomputer in the world"; the 450x and 1,000x numbers (the passage Proof of Talk later clipped); "a thousandfold larger than any centralized organization... and that's really scary." | | 4:00–5:40 | Caveat: "none of them are mine." Compute is commoditized: cheapest B200s and H200s; "the biggest data center bigger than OpenAI." | | 5:40–6:10 | Inference: Chutes, "a trillion tokens," "growing faster than Google did." | | 6:10–7:30 | Training: Templar's 8B run. Convergence as "a market equilibrium." Miners at the Three Gorges Dam whose "entire goal is to destroy your training run." | | 7:30–8:10 | Gradients: a paper claiming better than Hugging Face, Google AutoML, Together. "We don't even know what the miners are doing." | | 8:10–11:00 | The three eras. Missionary era: CPUs, "people were making exorbitant amounts of money off doing nothing," a miner in the room who mined at 13, "we've [expletive] up a lot." The design error: "pay you until you create AGI... doesn't work... too ill-defined." Gamer era. Then the Bitcoin parallel (CPUs, ASICs, FPGAs, Barry) and "Bittensor has stopped being a game... It is industrial scale." | | 11:00–13:30 | The map slide: subnets mining subnets; Chutes and Nous Research. "The best market for intelligence is intelligence itself." An 8x H200 box in a basement finds its best market in decentralized training. Co-mining. dTAO as the market underneath. | | 13:30–14:45 | The full stack, from Google: sub-teams on sub-teams, compute up to intelligence. "Is it just compute plus data?" Answer: yes, "but also inference." "All the way up to God." | | 14:45–16:30 | The turn: "I consider a success that I'm not needed anymore... why am I on stage?" Then the news: "I registered a subnet about 35 minutes ago... subnet 120... Affine... I'm back in the trenches." Design: Celium compute, Gradients training, Chutes serving. Title decoded: "satanic obviously being compute and sublime being the pinnacle of intelligence." | | 16:30–17:20 | Close trails off. "They took the slide away... it's on Subnet 20." "Thanks for coming everyone... love you all. Cool." Then: "technically you can just invest right now. I'm pitching you... no premine, so I don't have any allocation." | | 17:20–19:25 | Q&A (Atlas): the architecture. Original design paid for closeness to AGI and "doesn't really work that well"; you must "constrain the degrees of freedom of the miners"; training on Bittensor, GPUs on Bittensor, inference on Bittensor; "we're stitching together commodities." Ends: "enjoy the talk." | ### Thesis Bitcoin proved that a pure market for one well-defined commodity can out-build corporations and states. Bittensor's job is to do that for intelligence, and the way to do it is to refine intelligence into a stack of well-defined commodities, from compute ("satanic") up to intelligence ("sublime"), with subnets mining each other so that "the best market for intelligence is intelligence itself." The network has outgrown its founder; Affine is his return to the top of that stack. It is a product showcase with a history lesson inside it. There is no political frame: the fear appears once ("really scary") and is never developed. ### Best verbatim passages (fillers dropped, otherwise as captioned) - ≈1:32 "I always like to say that Bittensor is a federation. We're trying to create a thousand-year federation. So I'm not the CEO. I'm just one of the people here that really believes in this technology." - ≈2:03 "We're a movement. And we started this movement a long time ago. And one of the first things we did was write it down. We wrote down our mission to unify Bitcoin and AI. And at the time, we really didn't even know what that meant." - ≈3:12 "Bitcoin is 450 times bigger than Google's cloud, IBM cloud, Cerebras cloud, AWS, and Nvidia. 450 times when measured in terms of hashes than all of those data centers combined." - ≈3:44 "If we can get the power of digital currency markets to focus on artificial intelligence, we can be a thousandfold larger than any centralized organization in the world. And that's a big deal and that's really scary." - ≈6:11 "The significance of building a market where the market pushes participants to lower the loss. In order to do that, you need to make the actual convergence of a machine learning model a market equilibrium." - ≈6:11–6:53 "You're training with individuals that are maybe using GPUs next to the Three Gorges Dam, and maybe their entire goal, in fact we know that their entire goal, is to destroy your training run. We know that they don't give a [expletive] about the loss going down, but we compel them to do that." - ≈8:48 "We couldn't just point a market at a machine over an Axon connection... and say, 'We're going to pay you until you create AGI.' That doesn't work actually. It's too ill-defined." - ≈9:19 "Every time we coalesce these commodities into being better, more well-directed, they get more powerful and we get closer to unlocking the supercomputer power of Bitcoin." - ≈10:24–10:56 "We're at a point where Bittensor has stopped being a game. It's no longer gamers. It's no longer a faux competitiveness. It is industrial scale." - ≈12:03 "I really believe the best market for intelligence is intelligence itself. I used to say the best measure of intelligence is intelligence itself, but I think the best market is." - ≈12:35 "Where is the best place to take a computer that's not in the data center but run by some homegrown individual in their basement? We're talking just a single 8x H200 box. Where's the best market for that? Well, the best market for that is actually decentralized training." - ≈14:12 "Is it just compute plus data? Well, we are figuring that out and we think it is actually. We think it actually in many cases is just compute plus data, but also inference." - ≈14:48 "As someone that's been putting my lifeblood into making sure that it's well organized, I consider a success that I'm not needed anymore. And I had a bit of a hard time writing this presentation because I was like, why am I on stage?" - ≈15:23 "I registered a subnet about 35 minutes ago. It's subnet 120. It's called Affine. And I'm back in the trenches, guys. I want to lead from the back." - ≈16:33 "Satanic obviously being compute, and sublime being the pinnacle of intelligence." - ≈17:24 "The original design of Bittensor was: run a node, we'll query you and we'll measure how close you are to AGI now. And we'll pay you if you're closer to AGI. So it turns out that that doesn't really work that well." - ≈18:37 "I can define my incentive mechanism as just: train a model by giving me the data and the Python script, and you can't do anything else." - ≈19:07 "We're stitching together commodities." ### What worked 1. The number with its own caveat. "450 times... when measured in terms of hashes." This is the passage Proof of Talk chose to clip two months later, so it landed. The honest qualifier is what makes it quotable. 2. The three eras. Missionary, gamer, industrial, mapped onto Bitcoin's CPU, ASIC and industrial-mining eras, with Barry (the next speaker) as the living example. It is the best narrative device in the talk: it explains the embarrassing past, it makes "industrial" a compliment, and it says "we are where Bitcoin was in 2013" without claiming to be Bitcoin. 3. Admissions before boasts. "None of them are mine." "We've [expletive] up a lot." "It's kind of embarrassing." "Why am I on stage?" Each one buys credibility for the next claim. Same pattern as section 1, point 7. 4. Live news. "I registered a subnet about 35 minutes ago." The room got something that did not exist that morning. It also turned a showcase of other people's work into a personal stake. 5. The adversary image. Miners next to the Three Gorges Dam who want to destroy the training run and are compelled to lower the loss anyway. It is the clearest spoken statement of what an incentive mechanism is that appears in any of the sixteen recordings. 6. A reflexive line. "The best market for intelligence is intelligence itself," and his own correction of his older line ("I used to say the best measure..."). Showing the edit is a habit that works for him. ### What did not 1. The close trailed off twice: once at ≈16:33 ("thanks for coming everyone... love you all. Cool.") and again after Q&A ("enjoy the talk"). Same pattern as section 1, point 9. 2. The title was decoded in one sentence at ≈16:33 and the second half of the title ("aligning human potential through open-ownership protocols") was never spoken. The press summaries had to guess. 3. The talk depended on slides that failed twice ("they took the slide away," "I think I got kicked out"). When the slide vanished he said "Subnet 20" instead of 120. 4. "Technically you can just invest right now. I'm pitching you." Even with "no premine, so I don't have any allocation," this is the sentence a hostile journalist quotes. 5. Unverified superlatives: "a trillion tokens," "growing faster than Google," "the cheapest B200s," "better than Google AutoML." None had a source on stage. 6. No moral frame. The scary part ("that's really scary") was said and dropped at ≈3:44. The room heard a product update, not an argument about who owns intelligence. ### What to change for Montreal (against `docs/thesis-v2-script.md`, 17 Sep) The script already fixes the biggest faults: the memorised close (step 16), "I am not here to sell you a digital currency" in the first minute, sourced numbers, the title set up as the frame at 00:00, and the moral as the spine. These are the points where the 2025 talk offers something the script does not yet use, or where the script should not repeat a 2025 habit. 1. Borrow the three eras, in one breath, at the top of step 13 ("it exists, layer by layer"). Draft: "Bittensor has had three eras. Missionaries on CPUs. Then gamers. Now industry, with margins." About 25 words. It gives the "It is small. It is the seed" line a history, and it is the stage-tested way he says "we were embarrassing and now we are not." It also rhymes with remember-list item 9 ("Bitcoin in 2012 had not replaced anything either"). 2. Put the adversary image next to "you only pay the person that can lower the loss" in step 12. The script's "Miners produce. Validators check" is correct but abstract. The 2025 line, "we know that their entire goal is to destroy your training run... but we compel them to do that," is what makes a Montreal engineer believe Teutonic-II can be permissionless and still converge. One sentence, cleaned: "The people training it may want it to fail. The market pays them only if the loss goes down." It also pre-answers the Teutonic-I borrowed-weights admission in step 13: the market did exactly what it was told, so the fix is a better definition, which is the 2025 "commodities get more powerful as they get better defined" point. 3. Add "the best market for intelligence is intelligence itself" to the Affine sentence in step 13. It is the one line from 2025 that explains why Affine sits on top of Celium, Gradients and Chutes, and it is short enough to survive. Put quotation marks around it; it is verbatim POT25 ≈12:03. 4. Use "I consider a success that I'm not needed anymore" in step 15, right before the key. The script's "The goal of power is to give it away" is a 2023 tweet; this is the same idea said on a stage, in the past tense, about himself. The key section becomes: I already made myself unnecessary to the subnets; the key is the last place I am still necessary; here is the date. One sentence. Verbatim POT25 ≈14:48. 5. Close the Affine loop with one sentence of continuity. In Paris the subnet was 35 minutes old; in Montreal it is fifteen months old with the numbers in step 13. Draft, not verbatim: "Fifteen months ago in Paris I registered this subnet thirty-five minutes before I walked on stage." It shows shipping without a boast, and it is checkable on chain. 6. Have the founding-error line ready for Q&A on "how do you measure intelligence": "The original design was: we'll pay you if you're closer to AGI. That doesn't work. It's too ill-defined. So we constrain the degrees of freedom of the miners" (POT25 ≈8:48 and ≈17:24, lightly joined). It is the most honest and most technical answer he has given to that question. 7. Do not repeat: "you can just invest right now," "I'm pitching you," "all the way up to God," "satanic." The Montreal room is political and press will be present. The script's first minute already rules the first two out; keep it that way in Q&A too. 8. The 450x number: the script does not use it and should not lead with it. If it appears at all, it belongs in step 11 ("Bitcoin competed on one input: electricity") as one clause with the caveat kept: "measured in hashes, it out-built every cloud combined." Section 2 of this file already flagged the hash-to-FLOP comparison as the weakest part of the Tsinghua talk. 9. Say every number and name out loud, not only on the slide. Twice in 2025 the slide vanished and the spoken words could not stand alone ("Subnet 20"). The script already treats slide titles as cues only; keep the layer names, subnet names and the key date in the spoken text. 10. Pace check. The 2025 captions run at about 158 words a minute over the whole video, including the MC, applause and fillers; Jacob alone is closer to 165. Captioned fillers ("um", "uh", doubled words) are about one word in twenty, and YouTube's captions drop some, so the true count is higher. Even so, his clean rate is about 150 words a minute, above the script's 135. At 150, the full 5,900-word script runs about 39 minutes, not 43 to 44. That means CUT-1 may not be needed for time, and the real risk is going too fast in the memorised close and the two landing lines ("You have one choice." / "Now you have two."). Rehearse those at a slower pace than the rest. ## 9. Cointelegraph 2026: "Can Bitcoin Save AI From Big Tech?" (CT26, full text) Source: `docs/sources/interviews/2026-06-cointelegraph-can-bitcoin-save-ai-from-big-tech.md`. 4:21, Jacob only, questions cut. Recorded at Proof of Talk 2026 (Paris, 2–3 June), published 6 June 2026. Obtained 20 Sep 2026 as full caption text; timestamps are "≈" except the first 19 seconds (the teaser, transcribed from audio). The stub's 10:56 was wrong: the video is a 4:21 edited cut and the caption covers all thirteen of the publisher's chapters, so nothing is missing. Added the same day, eight days before Montreal. ### Thesis Bitcoin and AI are both "tools of organizing matter": cryptocurrency aligns humans and compute, AI aligns the weights of neural networks, and Bittensor combines the two to produce intelligence as a commodity that everyone can own, see and build on. The case against failing is epistemic: if one company's model is everyone's mind, "it's not our mind," and people lose the ability "to make sense of the world." He closes on the competitive claim: Bittensor organizes talent better than an interview process and will eventually train better models. ### Best verbatim lines - 0:08 and ≈1:10 "everybody in the world will be sipping from the mental biases of a single company that controls artificial intelligence... that would be devastating for our ability to make sense of the world." (The second sentence is new; the teaser cut it. The ellipsis hides only a false start, "And I think that's".) - ≈1:10 "If it's not that way, then basically, you know, it's not our mind." - ≈0:20 "Bitcoin and artificial intelligence, the fields are effectively tools of organizing matter." - 0:15 "Cryptocurrency is the most powerful alignment tool for humans, for compute." - ≈0:50 "It needs to be transparent. It needs to be auditable. It needs to be co-owned." - ≈0:50 "Decentralization is kind of a catch-all term for a lot of different things. Let's say like technologies that have moral primitives." - ≈1:25 "Everyone can own Bittensor. Everyone can build on Bittensor. Everyone can see how Bittensor works." - ≈1:25 "There's some disadvantages. It slows things down for us, but we know where we're going, and we think that those primitives will really matter." - ≈2:15 "If decentralized AI doesn't win, I think that's when people notice. I think that's when people understand that they don't have any hope of taking down the giants." - ≈2:35 "if we live in a world where there is just one hierarchy of intelligence, one intelligence, people will eventually find out." - ≈2:50 "I do think people should be cognizant and think about the effects of outsourcing our mental capacity to machines that are controlled by singular entities." - ≈3:35 "people didn't know about TAO, but they were using those tokens." - ≈3:55 "we've really built products that seep into the real world and are used by people without even needing to know about TAO." - ≈4:10 "I think that we organize talent better than them. I actually think that we can find people that are true killers that can really solve problems faster than they could do with an interview process." ### New against the script (`docs/thesis-v2-script.md`, 17 Sep) 1. The sipping line now has its second half. The script quotes "sipping from the mental biases..." at step 5 (footnote 7, CT26 0:08). The full version adds "that would be devastating for our ability to make sense of the world" and, before it, "it's not our mind." Either could follow the quote in step 5 as a second verbatim sentence; "it's not our mind" is the shorter and rhymes with the title. Footnote 7 stays correct (the line is at 0:08); add "≈1:10" if the extension is used. 2. Step 9 says "A mind outside the state protects how you think." Jacob's own words for that, on tape, are now: "outsourcing our mental capacity to machines that are controlled by singular entities" (≈2:50). It is calm, it is aimed at ordinary people, and it names the mechanism (outsourcing) instead of the villain. Candidate for a footnoted quote in step 9 or step 5. 3. "Moral primitives" (≈0:50). Step 15 says "I have used the word 'decentralized' as little as I could tonight" and then defines it as "nobody can be sent the letter." Jacob has already done the same move himself: "Decentralization is kind of a catch-all term... technologies that have moral primitives," then the list: transparent, auditable, co-owned. The script's three properties (no one licenses, directs, or switches off) are the same shape. Worth one sentence: "I have said before that decentralization is a catch-all. What I mean is three properties." 4. "It slows things down for us" (≈1:25). The script's habit is to say the cost first ("Say the cost first: a company ships a better model sooner"). This is Jacob saying that cost himself, in public, three months ago. It fits step 10 exactly. 5. "One hierarchy of intelligence" (≈2:35) is his phrase for what the script calls "one supply of intelligence" with a switch. Same idea; the script's phrase is better for the close, but "one hierarchy" is useful in Q&A because it says who is above whom. 6. "People didn't know about TAO, but they were using those tokens" (≈3:35) and "products that seep into the real world... without even needing to know about TAO" (≈3:55). Step 14 says "Most of you will only ever use it, the way most Bitcoin holders never ran a node." This is the same point from the supply side, with a live example (Chutes serving more open-source tokens than any other provider "at one point"). Supports the "Use" verb; the "at one point" hedge should be kept if quoted. 7. The Jensen Huang passage (≈1:50) should stay out. It is name-dropping, the "mentioned Bittensor in a talk" claim is thin, and the script has no place for it. Fine for Q&A if asked. 8. Nothing in the script depended on the 10:56 length. The teaser quote and its timestamp are unchanged. ## 10. CoinFoin: documentary prologue (CF) Source: `docs/sources/interviews/2025-coinfoin-bittensor-documentary-prologue.md`. 2:42, of which about 2:20 is Jacob, to camera, no interviewer audible. Published 20 May 2024, so it belongs between TBR (May 2023) and DOC (July 2024), and is the earliest recording in which he describes Bittensor as a nation state. Obtained 20 Sep 2026 as full caption text; all timestamps "≈". The file name keeps its "2025" prefix; the header has the right date. ### Thesis Bittensor is a corporation that is also a nation state with a democracy inside it, the kind of co-owned, power-in-check organization "we built when we came to America" and "over the last thousand years in Europe." It has no boundaries and no HR department. Money is where it becomes real ("where the rubber hits the road is where money is created"): digital currencies build mining networks that produce the digital commodities intelligence needs, "incentivized into an existence rather than purchased," because measuring what you want buys it cheaper than purchasing it blind. The invitation is ownership: "come be on our board of directors." ### Best verbatim lines - ≈0:18 "It's like a company. It's like Google maybe. But it's also a nation state, because there's a democracy inside of it." - ≈0:40 "There is not just complete top-down ownership. And that's what we built when we came to America, and over the last thousand years in Europe we built systems that we could co-own, and... that power was in check." (The ellipsis hides a false start, "that was".) - ≈0:55 "There are no boundaries. It's permissionless. There's no HR department. You don't need a resume." - ≈1:00 "You can come, you can own it, you can use it, you can control it, you can build it, you can contribute to it." - ≈1:14 "Where the rubber hits the road is where money is created." - ≈1:20 "We are like the bosom, we're like the womb for AI that all of humanity can own." - ≈1:40 "We make AI, because it is a computational problem, using compute that is incentivized into an existence rather than purchased." - ≈1:50 "If we can directly measure the things that we want accurately, we can get them cheaper and more efficient than if we just tried to buy them outright, not knowing what we needed." - ≈2:05 "a language for expressing exactly the things, the digital commodities, that we need for the production of artificial intelligence, inside of an open ownership nation state organism corporation. That's Bittensor." - ≈2:16 "I like to say come be on our board of directors, which means you hold the keys to this thing and you can own it, you can viscerally control it." - ≈2:35 "it grows every day, and it's [expletive] powerful. Sorry for that." ### New against the script 1. Jacob has already made the constitutional argument himself. Steps 1 and 2 of the script rest on Weber, Montesquieu, Smith, Tilly, Ferejohn and Rosenbluth. In May 2024 Jacob said, unprompted: "over the last thousand years in Europe we built systems that we could co-own, and that power was in check." That is the spine of steps 1 to 4 in his own voice, two years before the script. One footnoted sentence in step 2, after "Every right was a receipt," would let him own the history instead of citing it: "I said this two years ago: we built systems that we could co-own, and that power was in check." The two clauses are verbatim CF ≈0:40 with the false start "that was" dropped. 2. "Incentivized into an existence rather than purchased" (≈1:40). Step 11 argues that an exit with no address must pay for its inputs, and quotes a 2026 tweet ("we cannot rely on altruistic computational resources alone"). The 2024 line is shorter and says the positive half: the compute is not bought, it is called into being by the reward. Candidate replacement or companion for footnote 11. 3. "if we can directly measure the things that we want accurately, we can get them cheaper and more efficient than if we just tried to buy them outright, not knowing what we needed" (≈1:50). This is the economic reason validators exist, stated as a purchasing argument. Step 12 says "Validators check, and checking is cheaper than producing." The CF line adds why the market beats procurement: the buyer does not know what it needs; the measurement does. One sentence for step 12 or Q&A. 4. "Come be on our board of directors... you hold the keys to this thing" (≈2:16) is the 2024 form of the "Own" verb in step 14, and it collides on purpose with the key section in step 15. In 2024 he told everyone "you hold the keys." In 2026 the script admits one key is still his. That is either a good honesty beat ("Two years ago I told people, come be on the board, you hold the keys. That was true of everything but one key.") or a line a journalist finds first. Decide which; do not leave it unaddressed. 5. Five verbs, not three. "You can come, you can own it, you can use it, you can control it, you can build it, you can contribute to it." The script's step 14 has Use, Earn, Own. Jacob's natural list has "control" in it, which the script gives to staking ("the right to point it by staking") without the verb. Consider "Use. Earn. Own. And through owning, point it." 6. "A democracy inside of it... an election process... a voting process" (≈0:18). Said before dTAO existed. Step 14's dTAO numbers ("two million TAO moved by holders... Nobody approved those choices") are the 2026 proof of the 2024 claim. If CUT-2 survives, this pairing is one sentence. 7. "The womb for AI that all of humanity can own" (≈1:20). A new metaphor family (body, birth) that appears nowhere else in the sixteen recordings or the tweets. Not for Montreal; the image is odd cold. Noted for the metaphor list in section 4 (add to "Food and body"). 8. "Nation state organism corporation" is the 2024 version of "somewhere between a nation state and a corporation and a cult" (section 4). The script deliberately does not call Bittensor a nation state ("this is not a third country. A third country is a third switch"). Keep it that way; do not quote the nation-state lines on stage, only the "power was in check" one. ## 11. The Early Stage Investor (ESI): commentary, no Jacob speech Source: `docs/sources/interviews/2026-early-stage-investor-subnets-will-crush-big-tech.md`. 11:12, published 3 August 2025 (not 2026 as the stub guessed). Obtained 20 Sep 2026 as full caption text. Confirmed: this is a commentary video. The host reads and interprets four of Jacob's tweets (26 Jul to 2 Aug 2025) and plays a 12-second clip from the POT25 keynote twice. Jacob does not speak to this channel. The file name keeps its "2026" prefix; the header has the right date. ### Thesis (the host's, not Jacob's) Three subnets can compete with big tech: Celium/Lium (SN51) with AWS, Bitcast (SN93) with Amazon's influencer programme, Ridges AI (SN62) with DeepMind. The evidence is Jacob's tweets, Fish's revenue post, and two AI-generated comparisons the host reads aloud. The video discloses that it is paid: "this video is incentivized by the Bitcast network." ### Jacob's words in it Only the POT25 clip, already in section 8: "I registered a subnet about 35 minutes ago. It's subnet 120. It's called Affine... the design that we're using right now is using compute from Celium" (POT25 ≈15:23 and ≈15:59, joined by the host's edit). The tweets the host reads are all in `docs/sources/tweets/`; the strongest is 31 Jul 2025: "Take the frontend away, take the leader away and it still runs, a floating machine in space that prints cash." ### New against the script 1. Nothing spoken. The video adds no line to section 6 and no idea to section 5. It can be dropped from the interview set for voice purposes; the transcript file stays as the record of what it is. 2. One tweet it surfaces is relevant to step 15 and is not in the script: "take the leader away and it still runs" (31 Jul 2025, about Lium). It is the property the key section promises for the whole network, said by Jacob about one subnet. Candidate for step 15, after "On the day it is gone, this stops being a promise and becomes a property," as a footnoted tweet. Check the tweet is not already used elsewhere before adding; it is in the archive at `const_reborn-tweets.md`. 3. A press risk the script does not cover. Bitcast (SN93) is one of the "faces" in `docs/research/ecosystem-as-empowerment.md` (section 6, the Blears brothers) and appears in `docs/slides-draft-v3.md`. The same subnet pays YouTube channels like this one to make "These Subnets Will Crush Big Tech!" videos with "250 plus % APY" pitches. A journalist who searches Jacob's name on YouTube finds this video on the first page. If Bitcast is named on stage, be ready for "your network pays influencers to shill your token." The honest answer exists in the ecosystem file (brands pay for verified attention; Q2 2026 revenue exceeded miner pay) and in step 15's "If you hold: do not chase emissions." Have it ready for Q&A; do not raise it unprompted.