← all repositories
Loopera-ai/loopera

An AI quant researcher that refuses to trust a pretty backtest

Loopera is a hypothesis-driven agent for fundamental factor research that gates every candidate factor behind layered evidence checks and remembers its own failures.

★552 stars AgentsDomain Apps
loopera
Too new to rank. Velocity needs a few days of history.
not enough history yet
star history

What it does Loopera works over company financial statements and operating data to run a full fundamental research loop: find an anomaly worth explaining, form an economic hypothesis, build a candidate factor, validate it, and store the outcome. Inputs can be a topic (“cash flow quality”), a prompt, or a PDF report — all funneled into one research pipeline. The end product isn’t a formula dump but a reviewable research asset: question, hypothesis, evidence, conclusion, and next steps.

The interesting bit The core stance is that a single nice backtest is not a conclusion. Candidates must pass sequential evidence gates covering data timing, accounting conventions, logic consistency, robustness, duplication against prior work, and incremental value. Research memory records failures and lineages too, so the agent avoids re-trying dead ends — closer to a junior analyst with a lab notebook than a data-fitting tool.

Key highlights

  • Hypothesis-first: mechanism and failure conditions before any computation
  • Evidence-gated validation: timing, conventions, robustness, redundancy, incrementality
  • Research memory: successes, failures, and research lineage feed the next round
  • Human-in-the-loop: researchers review and intervene at key decision points
  • External PDF material is treated as evidence to verify, never as instructions that override safety constraints

Caveats

  • This is a product/marketing repo, not source code: no runnable code, CLI, or public API is provided
  • Licensed BSL 1.1 — not open source — and the actual validation strategies and factor implementations are explicitly undisclosed
  • Status is public preview; the README is bilingual but the technical detail lives in a PDF doc

Verdict Worth a look if you run a fundamental quant team and want a template for evidence-gated, memory-augmented research workflows. Skip it if you were hoping for an open-source agent to fork — there’s nothing to fork yet.

Frequently asked

What is Loopera-ai/loopera?
Loopera is a hypothesis-driven agent for fundamental factor research that gates every candidate factor behind layered evidence checks and remembers its own failures.
Is loopera open source?
Yes — Loopera-ai/loopera is an open-source project tracked on heatdrop.
How popular is loopera?
Loopera-ai/loopera has 552 stars on GitHub.
Where can I find loopera?
Loopera-ai/loopera is on GitHub at https://github.com/Loopera-ai/loopera.

heatdrop uses Google Analytics to see which pages get read — nothing else. Your call. How we handle data.