The Theseus manifesto
The long argument behind the firm: why principles and not data, how the machine compares to a quantitative shop, why the reasoning architecture is treated as edge rather than product, and how to read each public surface.
Theseus is a philosopher in a box. We extract principles from a curated corpus, build logical algorithms that apply those principles to live observations of the world, and place bets when the algorithms predict outcomes the principles support. We are the Renaissance Technologies of formal logic — the same machine shape (inputs → engine → conclusions → bets), one level of abstraction higher. We do not commercialize this. The machine is our edge.
corpus ──▶ synthesizer ──▶ principles ──▶ algorithms
│
live observations ──┤
▼
conclusions
│
▼
memos
│
▼
portfolio agent
│
▼
betWhy principles, not data
A principle is a logical pattern abstracted from a text. It is the stable shape underneath an argument, expressible in a sentence or two, applicable to situations the original author never saw. Data has patterns too; quantitative firms extract them and trade them. Principles are the next layer up: patterns that organise many possible data series, rather than patterns inside one.
Theseus reads a curated corpus the way Renaissance reads price tape. The synthesizer is the trainer; the principle is the fitted artifact; the algorithm is the live executor. When the algorithm fires on a fresh observation, the conclusion is just the principle applied — which is also why it can be argued with, replayed, and graded against what reality does next.
“The Good… is the cause of knowledge and truth, and you may think of it as known.”— Plato, Republic 508e
The Platonic idea is the working metaphor: the principle is the Form, the live observation is the shadow, and the algorithm is the discipline of moving correctly between them.
The Renaissance comparison
Quantitative firms abstract numerical patterns out of price data and arbitrage them. Theseus abstracts logical patterns out of text — principles — and arbitrages them. The stack is the same shape, one level up: a curated corpus replaces the tick archive, a synthesizer replaces the model trainer, principles replace fitted parameters, and an algorithm executing those principles against live observations replaces the alpha signal. The bet, when the algorithm predicts something the principles support, is the output of the same machine — operating on logic instead of numbers.
Renaissance Technologies did not invent statistical learning; they industrialised one application of it. Theseus is not inventing logic; we are industrialising one application of it. The bet, when it comes, is the output of a machine — operating on logic instead of numbers. The bet is polymorphic across the domains where logic can be applied profitably: Equities, Prediction markets, Advisory, Scientific, Private markets (eventually).
“To write between the lines.”— Leo Strauss, Persecution and the Art of Writing (1952)
The Strauss frame matters because principles often hide in the gap between what a text says and what it cannot say. The synthesizer is built to read both — the surface argument and the constraint that shaped it. The onion-futures example from the meeting: a quant sees price oscillation, but a reader who has worked through the corporate filings, the legislative record, and the speculator's incentives reads why — and that why is what the algorithm is trying to capture.
Why we are not commercialising this
Theseus is not a SaaS product. The reasoning architecture is our edge.
A SaaS distribution would force a single workflow on every buyer and would expose the corpus, the principle library, and the algorithm internals to anyone with a credit card. The machine is valuable in proportion to how much of the firm's specific reading has been encoded into it; selling shells of that would commodify the surface and erode the edge. We invest with the machine. We do not sell access to it.
“What appears to humanity as the history of capitalism is an invasion from the future by an artificial intelligent space that must assemble itself entirely from its enemy's resources.”— Nick Land, Fanged Noumena (2011)
The Land frame is the warning: a machine that assembles itself from its enemy's resources only retains its edge while the machine itself is not the resource. The corpus, the synthesizer, the principle library, and the algorithm registry are the machine. They stay inside.
How to read the public surfaces
Each public surface shows one cut of the same machine.
Manifesto
Theseus is building infrastructure for firm reasoning. A useful argument should not disappear when a meeting ends. A serious objection should remain visible after the conversation moves on. A judgment that informs action should retain the record of how it was formed, what challenged it, and what would require revision.
The project begins from a distinction between infrastructure and capital. The infrastructure — a memory layer, required citations, preserved objections, a controlled publication surface — is replicable in principle. The capital is not: it is the firm's own accumulated judgment, encoded into the Codex. Theseus builds the infrastructure to compound its own capital, not to sell it.
Today Theseus is not a commercial product. The Codex is the firm's internal edge, applied to the firm's own decisions: evaluating a forecast, testing a thesis, or weighing a judgment under uncertainty against the firm's accumulated reasoning — documents, meetings, predictions, essays, partner debates, and post-mortems. It is live feedback from the firm's recorded mind, used to lower error rates in the decisions the firm itself makes. That other knowledge institutions might one day build something similar is a possible future direction, not the present pitch.
That is close to retrieval-augmented generation with extra discipline, and the manifesto should not mystify it. A context bot embedded in Slack can retrieve useful passages. Theseus is trying to do something more demanding: model the logic of a firm, expose its recurring assumptions, preserve its dissent, and apply that modeled judgment to new situations without pretending that retrieval alone is reasoning.
The value of the Codex is its knowledge base, not the software. A replicated Codex would not reproduce Theseus; it would reveal the mind of whoever built it. That is why the firm treats the Codex as edge rather than product: what makes it valuable is exactly what cannot be sold without giving the firm away.
This also explains why the need is only now becoming visible. When the actual decision makers are always available, software can seem unnecessary: one can ask the founders, partners, or operators directly. But as firms scale, memory fragments. The same arguments recur, predictions go unscored, private reasoning becomes inaccessible, and the firm loses track of how it came to believe what it believes. The Codex exists to make that collective judgment durable.
For Theseus itself, the model is not merely a consultation surface. The aim is to apply the firm's logic across investing, writing, media production, Currents, prediction-market betting, and any other domain where disciplined judgment can be used profitably. The private Codex is the workspace where the source record is inspected and reprocessed; the public site is a selective publication surface governed by source visibility.
The operating axioms remain progress, rigor, and intellectual camaraderie. Progress means increasing real capability. Rigor means naming assumptions, judging methods, and earning confidence under objection. Intellectual camaraderie means treating disagreement as shared work rather than as a social threat.
No method should be trusted merely because it sounds rigorous. Theseus therefore treats forecasts, market outcomes, and later source evidence as checks on its reasoning. Capital decisions are one way the firm makes its reasoning accountable. The aim is not infallibility; it is a record detailed enough to show what failed when the firm is wrong.
The standard is operational: when the firm publishes a conclusion, the reader should be able to distinguish the claim, the evidence, the method, the objection, and the conditions for revision. Intellectual capital is human capital made durable without pretending to become post-human. The new industry is not the automation of belief. It is the institutionalization of accountable thought.