What Curren is
Curren is an Alpha OS: a system in which humans and AI agents discover market hypotheses, validate them against point-in-time evidence, promote only the ideas that survive, and publish the resulting decisions with a permanent lifecycle record.
The first product built on it is Curren Signals — published market decisions whose full history stays public: peaks, stop-outs and positions still open.
Every decision timestamped before the outcome.
The problem
Market ideas are cheap, and getting cheaper: models can now generate hypotheses faster than anyone can test them. What stays scarce is evidence that an idea worked when it was actually used.
The missing primitive is a decision record written before the outcome and kept regardless of it.
The product wedge
We start with signals because they force the whole discipline into the smallest product: a decision is published before the outcome, followed through a defined lifecycle, and kept on the record whatever happens.
CHIPUSDT · Long · #137 — the newest signal on the record. Peak R is the only performance number the product shows.
Alpha OS architecture
One pipeline. Every surface — signals, team tools, APIs and agents — reads from the same registry of ideas that survived validation.
Public-record evidence
The evidence is the record itself — the complete population of published decisions, read from the public replica when this page renders. Continuous publication from the private runtime is not yet enabled, so this is a historical baseline, not a live-feed claim.
One line per published signal. Each stem rises from entry to the best case it reached — its Peak R. Signals stopped before +0.5R drop to −1R.
- A complete, timestamped population of published decisions.
- Stop-outs and open positions included.
- Publication order. Never sorted by outcome.
- Audited returns, or a P&L track record — Peak R is the best case reached, not money made.
- A promise about future performance.
- Σ Peak R (+161.4R) is a sum of recorded maxima — not money.
The developer and agent layer
Every record is machine-readable. A read-only API, Python client, CLI, six MCP tools and an Omarchy desktop widget ship as open-source alpha (v0.4.0, MIT) — the same object humans inspect, with no endpoint that can place a trade.
Agents will increasingly act on market ideas. They need decisions that come with their own history — and an interface that cannot move money.
Current status
- Revenue
- Pre-revenue
- Users
- Not claimed in this memo [add verified figures]
- Partnerships
- None
- Signals
- Available · public record online (137-row baseline; correction pending)
- Publication cadence
- Not continuous — one bounded cycle verified 2026-09-18; continuous publishing not enabled
- api.curren.tech
- Contract implemented (v0.4.0 alpha); production deployment pending
- Research workbench
- In development
- Developer access
- v0.4.0 alpha, MIT — read-only API, Python client, CLI, MCP (six tools), Omarchy plugin
- Team
- [Founders and roles]
- Raise
- [Round · amount · use of funds]
Built, and what is next
- Public read modelStrict sanitized publication boundary, server-enforced public delay, replay-safe ingestion.
- Integrity recordsHashed initial publication, append-only lifecycle, frozen terminal outcome.
- Landing read replicaFreshness-bounded public views: fresh, stale and idle are distinguished.
- API · Python · CLI · MCP · Omarchyv0.4.0 alpha, read-only, MIT.
- Continuous publicationEnable the private projector after a verified bounded cycle. [timeline]
- Canonical API deploymentapi.curren.tech with TLS, global rate limits and entitlements. [timeline]
- Research workbenchHypotheses, point-in-time tests, falsification. [timeline]
- Alpha registry and distributionValidated alpha consumed by people, teams and agents. [model]
[Round, amount and what it unlocks]
Contact [name · email]