# Slides, draft v1: The machine that pays you back

Exploit Conference, Montreal, 28 September 2026. 45 minutes. Speaker: Jacob Steeves (Const).

Design: Bittensor theme. White background, black text, or the inverse for emphasis slides. One line or one image per slide. No bullet lists on screen. Typeface as on bittensor.com. Use the τ glyph sparingly.

Timing guide: 26 slides. About 1 minute 40 seconds per slide on average. Slides 1–5 move fast (about 6 minutes total). Slides 12–16 are the heart (about 12 minutes). Slide 21 (the key) gets a full 4 minutes. Leave 5 minutes for the close and none for Q&A unless the organizer asks.

[VERIFY] marks any slide that carries a number to re-check on the day. Sources for all numbers: `docs/research/state-of-bittensor-2026.md` and `docs/research/economics-of-ai-and-labor-power.md`.

Speaker notes are beats, not lines. Jacob talks; the notes remind him where he is going.

---

## 1. Title

**On screen:** Black. White text. "The machine that pays you back." Below, small: Jacob Steeves · Bittensor · Montreal, 28.09.2026.

**Notes**
- Confirm spelling "Steeves" on this slide before export.
- Do not read the title. Walk out, look at the room, start with the feeling (slide 2).
- Tone for the whole talk: calm, direct, no hype. You know the material.

## 2. The feeling

**On screen:** White. One line, centered: "You have felt it."

**Notes**
- Everyone here had a moment this year when a model did something that used to be someone's job.
- Thrilled and afraid at the same time. Both are correct.
- I am not here to calm you down. I am here to aim the fear at the right target.

## 3. The builders are telling you

**On screen:** Quote, large: "Cancer is cured, the economy grows at 10% a year, the budget is balanced, and 20% of people don't have jobs." Small attribution: Dario Amodei, CEO of Anthropic, May 2025.

**Notes**
- Half of entry-level white-collar jobs. 10–20% unemployment in one to five years. His words.
- The people building the thing are describing what it does.
- Do not editorialize yet. Let the quote sit.

## 4. The line that has fallen for 80 years [VERIFY]

**On screen:** One simple black line chart, no gridlines: US labor share of income, 1947 to 2026. Two labels only: "65.8% · 1947" and "52.9% · 2026."

**Notes**
- The labor share is the slice of the economy paid to people for working. Everything else goes to whoever owns the machines.
- Lowest reading since records began. And this is before the automation the last slide described.
- One idea: the pie is shifting from work to ownership.

## 5. The firm without a payroll [VERIFY]

**On screen:** Two numbers side by side. Left: "GM, 1979 · 853,000 people." Right: "Anthropic, 2026 · ~5,000 people · $965B."

**Notes**
- For a century, a frontier company shared its gains through payroll. GM was the largest private employer in America.
- Anthropic is worth about a trillion with half of one percent of GM's headcount. Roughly $200M of value per employee. Walmart does $324k of revenue per employee.
- The AI lab has no transmission mechanism to society. It does not need one.
- Say "a few thousand" for Anthropic; the exact headcount is not published.

## 6. Where it pools [VERIFY]

**On screen:** Three short lines, stacked, large: "$720–745B · four companies' data-center spend, 2026." "53% · share of all venture capital that went to AI, 2025." "87% · share of US corporate equity held by the top 10%."

**Notes**
- The most valuable thing humans have ever built: built by the fewest people, owned by the fewest people, in the fewest places.
- Do not say "100% of capital." Say "a record share." The numbers are big enough.
- Transition: so why does it matter that work is losing its share? Because work was the leverage.

## 7. Why work was power

**On screen:** Quote: "Every large workshop and manufactory is a sort of political society, which no act of parliament can silence." John Thelwall, 1790s.

**Notes**
- The factory gathered people. Gathered people organized. Organized people made the vote matter.
- Second lever: wages are the tax base. Economists this year: AI "may erode the two main tax bases." A state that does not need your labor or your taxes has less reason to listen.
- Third: history. Engels' pause. 1780–1840, output per worker up 46%, wages up 12%. Sixty years. Ended only when workers organized.

## 8. The question

**On screen:** Black. White text: "What does organizing look like when the machines no longer need the people?"

**Notes**
- Pause here. This is the hinge of the first half.
- If your labor was your leverage and the machine takes the labor, you have no leverage. Not with your boss. Not with your government.
- You become an unimportant extremity of the engine that runs the world.

## 9. The real fear

**On screen:** Single still or line-drawing evocation of *Elysium*: a ring station above, a crowded surface below. Or just the word "Elysium" in white on black if rights are a problem.

**Notes**
- Not AI taking over. Machines want nothing.
- A small group of humans taking over, using AI to do it, and separating themselves from everyone else.
- Nobody on the surface is oppressed by a robot. They are oppressed by the people who own the robots.
- Decision point: the "Gaza" line from your notes. Decide before the talk whether to say it.

## 10. They know

**On screen:** Two lines: "Altman: 2.5% of every large company's equity, every year, to every adult." "Amodei: 3% tax on AI revenue."

**Notes**
- The people concentrating the capital are proposing to share it. They know what they are building.
- Notice how: through the US Treasury.
- Transition: the economists agree on the cure. Look at their mechanisms.

## 11. Everyone agrees on the cure. Nobody has a mechanism.

**On screen:** A plain table, two columns, no borders. Left: Korinek, Piketty, Weyl and Lanier, Altman, Amodei, Susskind. Right, each row: "the state," "the state," "firms," "the state," "the state," "the state."

**Notes**
- Korinek to the Senate, 2023: without reform AGI "would likely deliver mass impoverishment." His cure: "universal basic capital," everyone owns a piece of the AI.
- Every serious proposal: tax and redistribute, or ask the companies nicely.
- Both depend on a state losing its tax base or on corporations giving away their edge. And all of them stop at a border.
- Is there a third mechanism? We have already watched one work for fifteen years.

## 12. Bitcoin's second invention [VERIFY]

**On screen:** "≈1,000 EH/s · the power draw of Poland · no CEO."

**Notes**
- Everyone knows the first invention: money nobody can print.
- The second: the first permissionless commodity market. Bring compute from anywhere, get paid directly, every ten minutes, no contract, no HR.
- Miners hunt stranded hydro in Paraguay, flare gas in Texas, cheap power in Ethiopia. Nobody planned it. Incentives did.
- Do not say it out-computes Google. Hashes are not FLOPs. Say: more machinery coordinated for one purpose than any company or government ever assembled, with nobody in charge.
- The belief this proves: systems governed by incentives, not authority, can grow to the size of nations.

## 13. One product, and the product is a lock

**On screen:** A single SHA-256 hash string, small, centered, on white. Nothing else.

**Notes**
- All that machinery produces one thing: a hash. Useless for anything except securing Bitcoin.
- The question that started Bittensor: what if the same machine produced the most valuable commodity of this century?

## 14. Nature is not centralized

**On screen:** Black and white photograph or drawing: mycelium, a river delta, or a neuron. One word beneath: "Decentralized."

**Notes**
- A forest, a reef, your body. No CEO. Trillions of parts following local rules, exchanging energy with neighbors. Out of that, intelligence. It produced us.
- Neuromorphic computing: chips built from memristors that learn by feeling energy flow through them. Paths that carry current grow stronger. Structure emerges. [CONFIRM company name: Knowm?]
- The idea under Bittensor: money is energy flowing through a network, and where it flows, structure grows. Point it at the hardest problem: producing intelligence.
- A new form to grow AI inside the human biome, not above it. Keep it here. No origin story.

## 15. Bittensor in one breath

**On screen:** "Bitcoin's mining machine, pointed at intelligence."

**Notes**
- A chain that hosts many markets. Each subnet is a Bitcoin for one commodity: compute, storage, data, inference, training, agents.
- Miners produce. Validators judge. Yuma Consensus turns judgments into payment and punishes validators who stray from the stake-weighted majority. Colluders under 50% decay to zero.
- Since February 2025 each subnet has its own token; its market price sets its share of new TAO. No committee. The market.
- One token above it all: TAO. 21 million, ever.

## 16. Twenty-one million. No pre-mine. [VERIFY]

**On screen:** "21,000,000 τ · 0 pre-mined · first halving 15.12.2025."

**Notes**
- Nobody was given anything. No insider print, no venture slice before the public. Every TAO was mined from block one by someone running a machine. Mine too.
- That is why this community is what it is. People who earned their position defend it differently.
- Still true tonight: anyone here can mine, run a validator, or write a subnet. Your agents can mine for you. Nobody approves you.
- Tie back to slide 11: Korinek wants everyone to own AI capital through a government program. Ours is issued directly to whoever contributes. "Permissionless basic capital," my phrase, not his.

## 17. The third path

**On screen:** Three words across the slide: "Washington. Beijing. Neither."

**Notes**
- Every lab that matters lives inside one of two power structures. Their weights can be export-controlled, their access switched off by a phone call from a capital.
- Bittensor is inside neither. Bottom-up, grassroots, international. Disjoint by construction.
- We are in Montreal. This country built much of this field and buys it back from two superpowers. This room knows what downstream feels like.
- Nobody else brings digital currency into the AI conversation. It is the only tool that has ever coordinated this much machinery without a boss. Own that.
- Decide tone: "outside both," not "against both."

## 18. What has been built [VERIFY]

**On screen:** "≈128 subnets · from ≈32 in early 2025."

**Notes**
- Five years ago: a whitepaper and a handful of miners.
- Today: the largest decentralized AI network in the world by a wide margin.
- Name three subnets and what they sell (next slide). Do not list more.

## 19. Real customers [VERIFY]

**On screen:** Three names and one line each: "Chutes · inference at scale." "Targon · confidential compute, with Intel." "Lium · GPU marketplace." Beneath, small: "24 subnets · $28–35M/yr external revenue · Sep 2026."

**Notes**
- Chutes is the only Bittensor-linked provider on OpenRouter; anyone can check its token volume. Targon runs inference inside trusted hardware. Lium sells compute credits that burn its token.
- Fourteen of these teams buy back their own tokens with customer money. Customers include PwC France and Dropbox.
- Honest scale: ~3,600 TAO minted a day versus $28–35M a year from outside. Real, growing, about ten percent of the way. Say it before someone else does.

## 20. Wall Street arrived uninvited [VERIFY]

**On screen:** "Grayscale · Bitwise · Nasdaq · TSX-V." Small: "GTAO on OTCQX; spot ETF under SEC review; three public treasury companies."

**Notes**
- Grayscale's trust trades on OTCQX and has a spot ETF filing in front of the SEC. Bitwise filed too. Neither approved yet; say "under review."
- Three public companies hold TAO as treasury, one of them listed here in Canada on the TSX Venture Exchange.
- Nearly half of all TAO is staked to root; root dividends have paid about $198M since Feb 2025.
- Then the gap: our market cap is $2–3B. NVIDIA's is about $5T. Two thousand to one. Proof of mechanism, not yet a counterweight. I would rather say it than have you think I do not know.

## 21. I hold a key

**On screen:** Black. White text, small, centered: "Bittensor is not decentralized in the way Bitcoin is. Yet."

**Notes**
- Plainest words in the talk. Blocks are produced by foundation nodes: proof of authority. A privileged key can upgrade or pause the chain. A small group holds it, and it runs through me.
- We used it in 2024 to halt a wallet-drain attack in 35 minutes. We could use it badly. We have been accused of using it badly. Some of you were in Templar in April and lost money. Decide whether to say sorry here. Draft says yes.
- Everything else is permissionless today and always was. But rules that a key can change are not finished rules.
- We chose speed over decentralization while we found the economics. Right call then. Wrong call now.

## 22. The path, with dates [VERIFY]

**On screen:** Four short lines, one per row: "Nominated proof of stake · ~1 year." "On-chain governance, key disabled · proposed May 2026." "Conviction rights for holders · shipping." "Full decentralization · Dec 2027. Then the rules freeze."

**Notes**
- Nominated PoS: token holders choose who produces blocks. Foundation nodes go away.
- Governance: proposals reviewed by elected validators and subnet builders who can delay, cancel, or fast-track; final vote to disable the sudo key entirely.
- Ossification: after December 2027, nobody, including me, can change the rules.
- Say the ossification date every time you say "iteration."

## 23. They cannot make this promise

**On screen:** "A centralized lab cannot decentralize. Its key is the product. Ours is a bug we are removing."

**Notes**
- OpenAI cannot decentralize. Anthropic cannot ossify. Their entire structure is the key.
- The difference is not that we are already decentralized. It is that we can be, we are on a dated path, and they cannot.
- This is the turn that makes slide 21 a strength.

## 24. To holders, builders, and the new

**On screen:** Three words, spaced: "Hold. Build. Come."

**Notes**
- Holders: you own the only AI capital on Earth that was never allocated to anyone. Stake to validators who allocate well. Hold through the year it takes to remove the key. That day the asset becomes what it was designed to be.
- Builders: ~128 markets, room for yours. Find what is slow, expensive, or closed and write a market for it. Registration cost is falling toward a transaction fee. Build for a customer, not for emissions. Your model, data, GPU, or agent can be paid tonight, in ownership.
- Newcomers: you just heard the shape of the next twenty years. You do not have to wait for a government. The network has no gate. Come and see if it is real.

## 25. Aim the fear correctly

**On screen:** "Do not fear the machine. Fear the machine owned by a few."

**Notes**
- Return to slide 2. The feeling was correct. Aim it.
- Bitcoin proved a network with no owner can outgrow any company. Bittensor is that network, pointed at intelligence, paying ownership to anyone who helps build it.
- The fork: a station in orbit and everyone below, or an intelligence that grows like a forest, owned by everyone who tends it.

## 26. Close

**On screen:** Black. White τ glyph. Beneath, small: "Build it with me." Then a line for bittensor.com and how to start.

**Notes**
- We have the people. We have the design. Bitcoin gave us the proof it can scale.
- What we do not have is time to waste.
- Stop. Do not add a thank-you slide.

---

## Production notes

- Slides marked [VERIFY]: 4, 5, 6, 12, 16, 18, 19, 20, 22. Re-check on 27 September against the two research files. Market cap and subnet count move daily.
- Slide 9: check image rights for *Elysium*. Fallback is text only.
- Slide 14: confirm the neuromorphic company name with Jacob (notes say "GNOME"; likely Knowm). If unconfirmed, drop the company and keep the memristor idea.
- Slide 17: confirm exact conference name for the title slide and any host thanks.
- Format undecided (Keynote, Google Slides, PDF, web). This Markdown converts cleanly to any of them; each `##` is one slide.
- Total on-screen word count is under 300. Keep it there.