lawsynth/ROADMAP.md

3.3 KiB
Raw Permalink Blame History

Roadmap

Shipped (P0–P5 + product)

The local engine and the full product loop are implemented and verified: World IR, bundle integrity, numerical simulation, reproducible discovery (sparse + symbolic + lagged + Pareto + regimes + uncertainty + parameter refinement + causal hypotheses), and the language bindings — plus the product surfaces built on top: the CLI, the Python SDK (Study, recipes, ensemble, backtest, monitor, Project, Client), the Studio app, the discovery-as-a-service backend, the notebook dashboard and interactive widget, six export targets, and self-hosting/deployment scaffolding. See PRODUCT.md for the product view and ARCHITECTURE.md for the engine boundary.

Formerly-proposed areas (causal analysis, regimes, uncertainty, plugins, Studio, services, deployment) are now implemented with source, tests, and user-facing contracts — not proposals.

Next

The forward plan turns a complete single-user tool into a collaborative, extensible, governed open-source platform that users run locally or self-host. Each phase is bounded by a boundary specification with a conformance suite:

Phase Goal Spec
P6 Collaboration teams share, review, and version models specs/collaboration/
P7 Streaming discovery models that update on live data specs/streaming-discovery/
P8 Plugin marketplace safe community extensions specs/plugin-marketplace/
P9 Governance auditable, accountable models specs/model-governance/
P10 specification archive former managed-hosting boundary, retained as historical design input rather than an active service plan specs/hosted-platform/

The full plan — goals, user problems, and sequencing — is in docs/roadmap/next-phases.md.

Proposed quantitative-research platform

docs/roadmap/quant-research-platform.md defines the shared architecture and staged roadmap for LawSynth's open-source quant family: data and experiment contracts, stochastic models, calibration, derivatives pricing, portfolio/market/liquidity/credit risk, backtesting, microstructure research, and independent model validation.

docs/roadmap/quant-diffusion-stress-engine.md is the first proposed vertical inside that family. It defines a production stress-testing product built from classical SDE and jump-diffusion baselines, an optional PyTorch diffusion generator, portfolio valuation, risk backtests, and governed reports. Planning does not alter the current roadmap order. Implementation begins only after a design partner, evaluation data, specifications, and statistical acceptance thresholds exist.

How priorities are set

Priorities are decided through maintainership review and reproducible technical evidence rather than calendar promises. A phase ships only what its conformance suite can verify; capability gaps are documented, never faked. The local, offline, deterministic core never regresses as collaboration and self-hosting features are added. lawsynth.dev remains a static project website, not a managed LawSynth application.