caelum0x 8bae186f19 feat(product): rolling-origin backtest, portable workspaces, cookbook
Wave 10 — rigor & sharing:
CLI — lawsynth backtest (rolling-origin/walk-forward: N origins x horizon,
per-state RMSE/MAE/R2 + skill-vs-horizon decay + verdict, --html report);
lawsynth workspace export/import (portable .lsworkspace archive: world bytes +
provenance + runs, per-entry SHA-256 integrity, non-destructive import).

SDK — Study.backtest -> Backtest (skill-vs-horizon decay, deterministic);
lawsynth.Project workspace persisting in the CLI library.tsv format (verified
bidirectional CLI<->SDK interop) with export/import archives.

Docs — a real cookbook (10 task recipes, CLI + SDK each) + 3 end-to-end
tutorials, verified against the shipped source (two-intervention-semantics and
preset-vs-recipe differences called out).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-08-17 23:46:25 +03:00

1.8 KiB
Raw Permalink Blame History

Getting started

LawSynth turns time-series observations into executable mathematical worlds: interpretable law systems you can read, simulate, forecast, compare, and share. The whole product is one loop, and every step is deterministic and offline:

observe (CSV) → discover (laws) → understand (explain) → use (simulate /
forecast / intervene) → compare → share (report / .lsworld bundle) → organize

A discovery is a portable .lsworld bundle; everything downstream — CLI, Python SDK, Studio, and the HTTP services — operates on that same artifact.

  1. installation — build the lawsynth CLI and the Python SDK.
  2. quickstart — run the core loop end to end.
  3. concepts — datasets, worlds, laws, and bundles.
  4. your first world — discover, inspect, and reuse a bundle.
  5. Pick a surface: CLI, Python, or Studio.

Then go deeper with task-oriented, copy-pasteable material:

  1. Cookbook — short recipes for each task (discover, clean, forecast, validate, monitor, export, organize, pipeline, service, Jupyter), each with a CLI recipe and its Python-SDK equivalent.
  2. Tutorials — longer end-to-end walkthroughs that chain many features into one narrative (messy sensor → trusted model, predator–prey ecology, monitoring a running system).

The checked-in examples and the workspace tests are executable and are the source of truth for supported inputs. Discovery finds a sparse fit from the implemented feature library — it is not evidence that an inferred relation is causal or that extrapolation beyond the observed window is valid.