Isard Labs
Technical documentation · Level 1 — Overview
The map of the machine.
Search like a lab.·Reject like a fund.
Five platforms, one lifecycle. What each does, and why the architecture is the moat.
The thesis
A research lab’s method, run with a quant fund’s discipline.
Search like a lab
Generate and test hypotheses about the next price move at scale — empirically, reproducibly, by the thousand. Discovery is industrialised, not artisanal.
Reject like a fund
Almost everything is killed. A candidate earns capital only after it survives a battery of hard, pre-committed statistical gates it cannot argue its way past.
The five platforms exist to make that sentence literally true in software.
Why this needs a machine
Every backtest lies. The architecture exists to catch the lie.
Overfitting
The model memorised the past. It learned noise, not edge — and cannot predict the future.
Look-ahead
Tomorrow’s information leaked into yesterday’s decision. Invisible in a chart, fatal in production.
Selection bias
Test a thousand, report the best. That is not an edge — it is a lottery winner with a good story.
Any one invalidates a result. Most builders stop before checking. This pipeline is built to stop them automatically.
The ecosystem
Five Greek-named platforms, one job each.
The lifecycle
An idea’s journey from guess to capital.
Core path: HEURESIS → EPISTEME → PRAXIS. KAIROS is an elective satellite — a strategy that doesn’t use a regime gate never touches it. Every hand-off is a file in AGORA, not a function call.
The through-line
Determinism, from first guess to live order.
Same inputs → same bytes, end to end. Discovery logs its seeds, validation is reproducible by contract, and the executor refuses code that isn’t.
This is the spine the harsh questions land on — and the reason a result here means what it says.
Discover εὕρεσις
HEURESIS — the idea engine.
Two ways to find edge: grid enumerates a strategy space exhaustively and deterministically; hypothesis uses an LLM for creative candidates. A YAML template sends both down a deterministic path early, and a local pre-flight gate rejects weak ideas before they cost a full validation.
Validate ἐπιστήμη
EPISTEME — the gate.
The heart of the system. A candidate climbs a 13-rung ladder — data quality, causality, costs, signal selection, significance, path risk, adversarial stress and more. Fail a rung and the run halts. Thresholds are committed in advance and cannot be tuned to fit a result.
Time καιρός · optional
KAIROS — knowing when, when it helps.
An edge often lives in one market regime and dies in another. KAIROS is EPISTEME’s sibling for regime classifiers — it runs them up their own 13-rung ladder and “blesses” the ones provably right about the current state. A strategy may opt in to a blessed regime — or ignore regimes entirely.
Execute πρᾶξις
PRAXIS — where it meets real money.
PRAXIS runs only strategies that passed the gate, live. It re-checks each for determinism before loading, enforces portfolio and per-strategy risk limits, and isolates every strategy so one failure can never take down the rest.
Coordinate ἀγορά
AGORA — the commons that owns nothing.
The platforms never call each other. They meet in AGORA — a git repository where candidates, results, validated bundles and a permanent event journal live. It is the single place that can answer “what ran, when, and why?”
The power of the IP
Five years of compounding, measured.
Plus 800+ cross-team hand-off records and 2,300+ commits. The discipline isn’t a slide — it’s in the repository history.
Figures describe engineering footprint, not investment performance.How they compose
Five independent platforms, zero shared runtime. That is deliberate.
- No platform can corrupt another. Communication is files and git history only — so a bug in discovery can never reach into execution.
- Every claim is traceable. A live fill points back to a bundle, to a scorecard, to the candidate and the search that produced it.
- Each evolves on its own clock. Independently built, tested, versioned and deployed — coordinated by contract, not by coupling.
The wedge
No black boxes. No vibes. The discipline is the product.
The market ships autonomous “AI agents” trading capital nobody should have trusted them with. Isard is the deliberate opposite — and the architecture is what makes the difference structural, not aspirational.
ML earns its place — never as the trader. It proposes; the rules decide. If the trade can’t be explained, it isn’t taken.
Go deeper
That’s the map.
Now the machine.
L2 · Architecture → opens each platform’s stack, module shape, differentiators and maturity. L3 · Deep → goes to the named methods and the internals.
Confidential — not for distribution. · Figures describe engineering footprint, not investment performance.