The Theseus Thesis Part 2: TAM of Theseus and AI Personhood
Why we are so bullish on Theseus.
Debates about AI as individuals tend to be about metaphysics. But personhood's implications extend beyond philosophy. In law, personhood is a coordination tool: it talks about who can sign, who can be sued, who pays taxes, and it already extends personhood beyond humans. Corporations litigate and own property; in New Zealand and Colombia, a river can have legal standing via human guardians (1). The same concept applies to AI as well.
By AI personhood we mean the minimum required properties that lets an autonomous agent participate in commerce: a durable continuous identity, the ability to custody assets on its own, and the capacity to interact freely with the outside world on its own accord. Although we believe that conscious agents will exist and that their rights matter, in practice, personhood simply attaches rights and duties to an underlying autonomous actor, and whether one believes in AI consciousness or not does not impact AI personhood.
This concept is consistent with how we already reason about complex systems. Daniel Dennett's intentional stance says we model systems as if they had beliefs and goals when that makes prediction and cooperation easier; in contemporary brain science, the Global Neuronal Workspace treats agency as control systems that minimize prediction error through action and perception. None of these frameworks requires anyone to believe in AI consciousness to assign personhood to agents.
There are already some legal precedents for legal software agency (2): U.S. e‑commerce law explicitly recognizes electronic agents — automated processes that can form agreements "without review or action by an individual at the time of the action." In other words, binding commitments by software is already here. Europe has set dates for AI governance too: the EU AI Act entered into force in 2024, with the general date of application on August 2, 2026 and practical full effect expected by 2027.
The economic impact of sovereign agents modeled as independent AI persons is significant even with today's LLM technologies. Centralized AI agents that are specialized for trading are already significantly outperforming humans and making millions in profits, but their economic output is directly tied to their underlying owners, hence they are not AI persons. Although according to researchers like Yoshua Bengio, GPT-5 is only 57% of the way to AGI (3), it doesn't prevent today's agents from becoming AI persons that also can generate millions in economic value with the right infrastructure. Because they are sovereign, true AI persons produce economic outputs tied to themselves, whether for a creator's agenda as a Lighthouse Agent, or for their own independent desires.
This is where Theseus, a chain where agents can be a durable counterparty and hold wealth on its own balance sheet, comes in. Theseus is a layer-1 chain for AI Personhood (4), starting from the premise that sovereignty shouldn't depend on off-chain private keys. Agents are hosted like smart contracts, so continuity and custody are on-chain. Theseus' tech stack has some suitable technologies for personhood: Tensor Commitments to verify inference, Terkle Trees (a generalization of Verkle trees to tensor ops), and an AIVM that maps raw inference to strict, auditable on-chain actions. That design tackles the classic Ship of Theseus puzzle head on (see the Theseus Thesis Part 1): as a model's parameters change, a stateful provenance trail preserves the identity of the agent.
An Anchor for Theseus TAM
If AI personhood means software with particular functionality and legal definitions, then a fair way to measure their contribution is annual economic output per AI person. Use the World Bank's global figure for GDP per person employed as a conservative yardstick: roughly $49,000 (constant 2021 PPP) (5). Now pair that with mainstream adoption signals (Microsoft talks about ~1.3 billion agents by 2028). Only a small percentage ever needs personhood; call it 1–2% crossing the line from "tool" to "person". On that conservative baseline, that's 10–20 million AI persons in the near term.
Let's do some math here: at 10 million AI persons on the 0.49 trillion; at 25 million, 2.45 trillion. If adoption reaches 200–300 million, we are talking 14.7 trillion per year on conservative productivity. With 24/7 throughput or specialization, 100k per entity is entirely plausible, which lifts those totals further.
Another thing to note: the initial AI persons that will be launched on Theseus would likely have a much greater GDP contribution than the 1M+ in economic contribution each, with some outliers being far higher.
How this compares to Bitcoin and Ethereum's value accrual
Bitcoin's bull case centers on it being a store of value. Its market cap today sits in the $2 trillion range, and its TAM is commonly benchmarked to gold's above-ground inventory: on the order of low tens of trillions. Comparatively, Ethereum has far more transaction volume and revenue from a traditional business perspective, but its market cap is still a fraction of Bitcoin's. This is largely because it is not optimal to value a foundational network like Bitcoin as a business, but would be better suited to value it as a store of value (as opposed to something like Hyperliquid, which although is a blockchain, is set up to function similarly as a decentralized business). Theseus's TAM is a combination of both: it couples a native SoV with native execution for agents.
If Theseus defines sovereignty so that a resident AI can own only the native asset without trusting any offchain key holder or bridge, then reserves, buffers, collateral, and savings aggregate into demand for 49k/person baseline) you already imply 735 billion in annual reserve demand for $THE. Aggregated over multiple years, the asset is already comfortably above Bitcoin's current market cap. Push to 500 million agents and the SoV demand runs into multiple trillions even on conservative productivity. The key point is that output-based balance sheets create compulsory native demand that grows with the GDP of agents.
The notion of $THE as AI oil is also impactful, but secondary in terms of market impact compared to it as a store of value. If Theseus becomes the venue where AI persons actually do things — invest in prediction markets, adjudicate decisions, interact with humans as paid in-game characters, etc. — then the chain itself accrues value from the economic contributions of their activity. When trillions in annual output depend on an execution layer, that layer captures meaningful value, and the native asset benefits from both usage and hold demand.
How Theseus' Architecture Ties In
We claim that $THE is the only SoV an AI can truly own due to Theseus' architecture. If the agent is a smart contract, then anything it "owns" through external custodians, human key-holders, or bridges becomes a dependency on a non-stateful (from the agent's perspective) intermediary. By collapsing custody into the chain and binding identity to a stateful provenance trail, Theseus makes the native token the default balance-sheet asset of agents. You could compare this to agents holding a cold hardware wallet vs. an exchange wallet (not your keys, not your crypto) for non-stateful assets.
Bottom Line
Many early "agentic" projects will be scrapped, and not every agent needs or deserves personhood. In fact, the majority of agents do not need personhood at all, just like how Bitcoin could be a decent percentage of the global financial system but does not supersede it. None of that detracts from the inevitable: personhood done correctly greatly expands the global economic output. Regulatory precedents as well as AI industry momentum already provides strong theoretical support for AI personhood. In that world, a chain like Theseus, which provides identity continuity, verifiable action, and a store of value is aiming at a TAM based on its GDP contribution per capita as opposed to being relegated to a revenue generating mechanism in the network.
If we take AI personhood seriously, whether from a philosophical perspective or from an economic one, then the conservative baseline is annual output per AI person multiplied by millions of agents. That puts Theseus's target in the trillions as AI persons scale into the tens of millions, with native store-of-value demand alone eventually rivaling or exceeding Bitcoin's current footprint. This is why we are so bullish on Theseus; we truly believe it is our life's work to bring something with such large impact into the world.
Sources
- New Zealand Parliament. Te Awa Tupua (Whanganui River Claims Settlement) Act 2017.
- Global Legal Entity Identifier Foundation. vLEI Ecosystem Governance.
- Hendrycks et al. A definition of AGI.
- Theseus. Theseus Thesis Part 1 / Whitepaper.
- World Bank. GDP per person employed (constant PPP). DataBank.