# Paris Blockchain Week 2026 fireside with Bpifrance: "Decentralized AI: From Vision to Infrastructure"

- Title: Decentralized AI: From Vision to Infrastructure | Bpifrance & Bittensor | PBW 2026
- Channel: Signal Week (Paris Blockchain Week); host: Damien Avignon, Investment Director, Bpifrance
- Date: 16 April 2026 (Mona Lisa Stage, PBW26; published later)
- Length: 21:13
- URL: https://www.youtube.com/watch?v=lu49zFCnTR8
- Jacob's start offset: none given; Jacob's first answer is at about 1:50
- Method: YouTube blocked from this VM. Auto-caption text read from a rendered copy of the public watch page. No audio, so timestamps are estimated from position in the text and marked "≈".
- Format: research transcript. Host paraphrased; Jacob's key lines quoted verbatim (caption spelling cleaned: "Bitensor" → Bittensor, "SODA" → SOTA, "Shoots" → Chutes).
- Speakers: MC = stage host; HOST = Damien Avignon; JACOB = Jacob Steeves.

## Outline with timestamps

≈0:00 MC: last talk of the day (the one after it was cancelled). Introduces Damien Avignon and Jacob Steeves.

≈0:50 HOST introduces himself (emerging tech at Bpifrance, the French sovereign wealth fund) and Jacob ("known as Const," original architect, protocol valued near $3 billion).

≈1:30 HOST Q1: in a few minutes, why is a decentralised market for intelligence necessary, and what does Bittensor produce today that no one else does?

≈1:50 JACOB: Bittensor is "fundamentally a tooling which allows us to organize resources in a more effective way using the tooling of artificial intelligence." A merger of crypto and AI that creates "liquid markets that allow us to organize information, organize data, organize machine intelligence, organize individuals, organize capital." Key framing: "our goal is not necessarily purely one of ethics and philosophy. It's also a technique for solving problems better, for reaching SOTA, for lowering costs."

≈3:10 HOST Q2: one or two subnets that illustrate this.

≈3:20 JACOB: inference is subsidised today; "when this bubble pops or perhaps catches up with the real cost, people are going to feel the real expense of say inference." Chutes builds inference from computers around the world "next to stranded resources," as Bitcoin mining did. Other examples: fast training, cheap storage, data collection. Claim: "we repeatedly beat SOTA." Pattern: pick a domain (e.g., AI detection), build a well-defined market, aggregate global talent. "we actually solve problems and we make money by solving these problems better."

≈5:50 HOST Q3: explain how subnets are organised, deeper.

≈6:00 JACOB: "a Bitcoin mining style network for anything." To compete with OpenAI you need cheap inference, compute, storage, serving, data, training. Other crypto-AI projects "focus on one sliver"; Bittensor "built an abstract mechanism that allowed us to focus on every single layer." Each layer produces a commodity; an internal market organises them. Analogy: like Ethereum, where subnets are the smart contracts, each with its own token, competing to be useful. "It's like a DePIN L1."

≈8:30 HOST Q4: why should institutions look at Bittensor; what shifted in 12 months?

≈8:50 JACOB: allocate to projects that "actually push the field and do real things." "there's nobody in our class." The community "really really care about the fundamentals and the philosophy of digital currency."

≈10:20 HOST adds a sovereignty angle: an oligopoly where "all the cards are distributed between US and China."

≈10:40 JACOB: "this is a geopolitical game." "It's the East versus the West. Europe is playing a small part." Bittensor is "truly extra-national, right? We're not tied to one nation-state." Talent from France who "don't want to be in San Francisco." Closing logic: "if we can build permissionless systems, we can get compute from anywhere. If we build permissionless systems, we can also get talent from anywhere."

≈12:30 HOST: the elephant in the room, the Covenant incident. What did it teach you?

≈12:50 JACOB: it looks bigger to people from the "classic company" world. This is open-source software, permissionless, "liquid ownership." "the team is just 18% of a subnet." "All of those people just filled in the hole immediately." Within a day, three new projects did the same thing better. "go and die on that hill where you think it's about one individual instead of the network." "the real value is the narrative, it's the people, it's the commodities, and it's the miners, which are all in Bittensor."

≈14:40 HOST: what about the deprecated subnets? JACOB: compute subnet being built on 39; a 1B model training on 3; 81 never started.

≈15:10 HOST: explain conviction, the new governance change.

≈15:20 JACOB: Bitcoin invented proof-of-work; Ethereum proof-of-stake. Many things cannot be measured on a blockchain (reputation, IP addresses do not matter). One primitive that can be measured is time: "Block time is how we measure things on a blockchain." Conviction "is simply the introduction of this other dimension." Each new dimension brings "a Cambrian explosion of innovation." In a company, people are tied in by law; on-chain "we have to come up with new primitives." It is opt-in.

≈17:40 HOST: is it hard to watch your baby grow and give power away?

≈17:50 JACOB: many projects "fall off the cliff because they try to move too quickly into governmental systems." "even Bitcoin wasn't decentralized. It was one guy for the first 3 years." Bitcoin is "very simple... the equivalent of one subnet on Bittensor." "We're at five. Are we fully decentralized? No, it's going to take some time." Democracies took thousands of years of tooling. Plan: move to NPoS and a full governance system, "but we're going to do this slowly."

≈19:40 HOST: final question, 3 to 5 year vision.

≈19:50 JACOB: "Bittensor was always going in the perfect direction... we built the right primitives for an AI-first future." A software engineer with network access gets compute, inference, storage, markets for work, training, fine-tuning; "their biggest customer is fast-paced agents." "you can make a subnet with an agent in 20 minutes." "you know who understands it instantaneously is Claude." Agents optimising markets non-stop will drive exponential growth. Closing image: "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 and it's just going to be an abstract incentive market like it was intended to be... where we can all find harbor because we can have this machine work for us."

≈21:00 Applause; hand-off to Signal Week.

## Jacob's stage-worthy lines (verbatim, auto-caption source)

- "our goal is not necessarily purely one of ethics and philosophy. It's also a technique for solving problems better." (≈2:40)
- "when this bubble pops or perhaps catches up with the real cost, people are going to feel the real expense of say inference." (≈3:30)
- "there's nobody in our class." (≈9:00)
- "we are truly extra-national, right? We're not tied to one nation-state." (≈11:00)
- "if we can build permissionless systems, we can get compute from anywhere. If we build permissionless systems, we can also get talent from anywhere." (≈12:10)
- "even Bitcoin wasn't decentralized. It was one guy for the first 3 years." (≈18:10)
- "Are we fully decentralized? No, it's going to take some time." (≈18:30)
- "Block time is how we measure things on a blockchain." (≈16:00)