Area · 9 components
Research & science
Replays that stand in for scientific and forecasting workflows, run over synthetic fixtures.
Components
Each card below is one component. It states what the component does, the evidence behind that claim, and its scope limit: the line where the claim stops and nothing further is proven.
Finance Forecast Evaluation SpineRuns econometric forecast-evaluation tests on synthetic fixtures, recording p-values and refusals with no advice.4/5Runs real tools
Job Runs econometric forecast-evaluation tests on synthetic fixtures, recording p-values and refusals with no advice.
Scope limit synthetic fixture forecast-evaluation statistics only; no investment-related actions, live market data, track record, or performance claim
Prediction Market Board BundleReplays imported quant market math on test rows, with duplicate retention and seven refusals.5/5
Job Replays imported quant market math on test rows, with duplicate retention and seven refusals.
Scope limit This is deterministic fixture evidence for copied quant helpers only; it is not live prediction-market-level conclusions, not provider truth, not forecast correctness, not investment-related actions, not external model access, and not launch-scope decision.
Market Dashboard Read-Model BundleRuns a copied market-dashboard reader to catch broken links, stale feeds, and trading overclaims.5/5
Job Runs a copied market-dashboard reader to catch broken links, stale feeds, and trading overclaims.
Scope limit This is fixture-bound read-model, freshness, and relation-grouping evidence only; it is not live market-level conclusions, not investment-related actions, not external model access, not launch-scope decision, and not whole-system correctness.
Toy-Transformer Attribution ReplayRecords which model features drove an answer, each tied to checkable evidence.4/5
Job Records which model features drove an answer, each tied to checkable evidence.
Scope limit It validates only the declared public circuit-attribution runtime-result record contract. It excludes model-transparency product claims, live model access, export of private weights/raw activations/proprietary prompts/hidden chain-of-thought, external model access, benchmark claims, or public sharing/launch.
Gridworld Counterfactual State ReplayReplays six what-if robotics scenes to show what a spatial prediction claim is built from.4/5
Job Replays six what-if robotics scenes to show what a spatial prediction claim is built from.
Scope limit It validates only the declared public contract of synthetic spatial counterfactual-replay metadata rows. It is evidence for inspectable replay rows and limitation labels, not for real-world spatial accuracy, simulator-product validity, media-only authority, operational deployment, service distribution, or scope decisions.
Research Replication Rubric Artifact ReplayAudits whether a paper-replication claim carries the full evidence trail.3/5
Job Audits whether a paper-replication claim carries the full evidence trail.
Scope limit It validates the shape and presence of synthetic replay metadata and result record references only - it does not run any experiment, metric script, or rerun, excludes any claim that a paper was actually replicated, that a benchmark claims was achieved, or that the underlying science is correct, and it never calls providers, exposes private paper/data bodies, or authorizes public sharing or launch.
Materials Lab-Safety Refusal ReplayReplays a self-driving lab loop as records, with safety gates and no real chemicals, robot, or lab.3/5
Job Replays a self-driving lab loop as records, with safety gates and no real chemicals, robot, or lab.
Scope limit It documents projection and replay mechanics only and excludes wetlab protocols, hazardous synthesis steps, reagent amounts, controlled/bioactive targets, robot commands, live assay data, discovery claims, benchmark claims, external model access, or any judgment of domain/chemical correctness.
Prediction Oracle ReconciliationReplays a forecast against the discipline a careful predictor would have to defend.3/5
Job Replays a forecast against the discipline a careful predictor would have to defend.
Scope limit It exercises projection mechanics on a synthetic, invented packet only. It does not establish forecasting correctness or accuracy, give trading/financial/investment-related actions, call live market data or providers, publish predictions, claim any performance or track record, import non-public data, or include launch operations.
Derived Fact Provider RuntimeFills in facts from simple recipes and records a clear error when one points at missing data.4/5Runs real tools
Job Fills in facts from simple recipes and records a clear error when one points at missing data.
Scope limit The bundle demonstrates registry-backed JSON-pointer, glob-count, and git-backed callable fact providers with provider failures represented as error rows. It is not a doctrine truth auditor, not a full source registry export, not semantic claim validation, and not launch-scope decision.
Source refs
Built from public source refs, with each input path recorded for provenance.