An autonomous intelligence system for markets.
Anor is building an automated hedge fund with a continuously upgrading research stack. An evaluation harness tests new models and promotes better intelligences into production as evidence, priors, and calibration improve.
We'll reach out to interested investors and send research-stack updates.
The loop
How the edge compounds.
A research system that widens coverage, sharpens triage, and keeps retesting each layer as better models arrive.
- 1
Ingest
Ingest messy filings, feeds, and market data into a point-in-time memory the system can trust.
- 2
Context
Link entities, events, supply chain relationships, contract paths, and cash-flow routes in point-in-time memory so every situation arrives with context and citations.
- 3
Prioritize
Use priors, LLM reasoning, and explicit research budget to decide which contract setups, supply chain links, and cash-flow targets deserve deeper work next.
- 4
Learn
Feed outcomes into an evaluation harness that retests models, prompts, and policies, then promotes the best intelligences into production.
Each pass makes the next one better. Coverage widens. Triage sharpens. More arbitrage situations surface.
Why it compounds
A continuously upgrading hedge fund research stack.
The aim is to automate the stack gradually: first junior analysts, then senior analysts, then capital allocation, with an evaluation harness that keeps upgrading the fund to the best intelligences available.
Reason over cited evidence
Anor uses LLM reasoning to interrogate evidence, compare hypotheses, and say what still needs to be falsified instead of just summarizing one document.
Point-in-time memory
A temporal knowledge graph preserves entities, relationships, and economic context so the system can trace supply chain links, infer likely contract winners, and map downstream cash-flow beneficiaries across time instead of starting cold every run.
Evaluation harness and lead ranking
An evolving energy-based scoring layer helps rank leads and arbitrage candidates, while a continuous harness benchmarks models, prompts, and policies and promotes the best intelligences into production.