The Theseus Design Space

Theseus agents aren't just smart contracts with ML capabilities—they represent a fundamentally expanded design space that enables entirely new categories of on-chain applications.

This page provides an overview of the expanded design space. For the complete analysis including technical details, market dynamics, and the full vision, read the original article:

Read Full Article: Agents as an Evolution of Smart Contracts →

The Evolution: A Natural Progression

Bitcoin (2009)

First proposed stateful on-chain executable, non-Turing Complete code. BTC tokens were immutable, enshrined by the blockchain state. It solved money consensus among strangers.

Public ownership you can verify.

Ethereum (2014)

Solved program execution consensus among strangers. You, and everyone else, could verify what a full program does and come into agreement on that program. These allowed for small deterministic programs, whose lack of complexity was reflected through design constraints: every programmatic step had to be replicated across every node in the network.

Public program behavior you can verify.

Theseus (2025)

Fuses web2-style AI agents with the stateful, sovereign properties of smart contracts. Instead of replicating on-chain compute across all machines, one node performs heavy inference and others verify it before committing the result on-chain. This makes complex, intelligent applications economically feasible.

Public agent decisions you can verify.

Key Insight

Each step removes a human dependency from something markets care about. Bitcoin removed treasurers from "who owns what." Ethereum removed judges from "what happens next." Theseus removes hosts and operators from "what kind of decision will an intelligent entity make?"

The Design Space Constraint

Over the last decade, runtimes like SVM or MoveVM, and systems like Polkadot or eWASM improved smart contract platforms with cleaner programming models or different languages. However, all maintained the same fundamental constraint: deterministic, replicated execution across all nodes. This design choice prioritizes security and verifiability but limits the complexity of programs that can run economically on-chain.

Theseus takes a different approach, using tensor commitments for verifiable inference. This makes complex, intelligent applications economically feasible on-chain while preserving verifiability.

Concrete Example: Lending Protocol

To illustrate the difference in design space, consider how a lending protocol would work on Ethereum versus Theseus.

Ethereum (Aave, Compound)

Off-chain:

Backend computes interest rates, keeper pushes parameters on-chain

On-chain:

Contract executes user transactions (borrowing, lending, liquidations) based on predetermined formulas

Updates:

Deploy new contract or upgrade existing one through governance

Theseus

On-chain Agent:

The market runs as a lending agent—a first-class on-chain entity. Its inference process may be deterministic, but its code is nonetheless deterministically verifiable just like any smart contract.

Execution:

Solvency and limit templates in the agent's context gate what the agent can execute. The pricing step runs via agent inference, and validators then verify a tensor-commit receipt.

Updates:

Agent's context or underlying model can be swapped—by the creator or, where allowed, by the agent itself.

New Primitives Only Possible with Theseus

The design space isn't just "existing apps, but better." It's about creating blockchain applications that were never possible before.

Subjective Prediction Markets

Agents adjudicate nuanced questions like "Was the iPhone Air launch successful?" instead of only objective outcomes.

AI Persons

Fully autonomous entities with their own goals: GPs of LP funds, marketing swarms, DAO orchestrators.

Complex Governance

Evaluate proposals based on nuanced criteria, read documentation, analyze trade-offs—all verifiably on-chain.

Dynamic DeFi Strategies

Manage liquidity, rebalance portfolios, execute complex trading strategies autonomously and verifiably.

Why This Is a Multi-Trillion Dollar Market

Sum >> Parts

If Ethereum's ~$500B market cap is tied to the value of its apps, then making those apps dramatically more capable should substantially increase value captured by the base chain.

Before Bitcoin, there was Hashcash—it implemented proof-of-work but holds almost no value. By the same logic, the market caps of "Ethereum" and "agents" already exist; unifying them with tensor commitments suggests one of the largest TAMs in crypto, where the whole is worth far more than the sum of the parts.

Bitcoin: public ownership you can verify.
Ethereum: public program behavior you can verify.
Theseus: public agent decisions you can verify.

Each step removes a human dependency. Theseus removes hosts and operators from "what kind of decision will an intelligent entity make?"

Want the Full Story?

Read the complete article with detailed examples, technical deep dives, and the full vision for why agentic smart contracts represent a generational shift in blockchain technology.

Read: Agents as an Evolution of Smart Contracts →