Dedalus

Philosophy

Filled-in fields do not establish experimental success; execution results must belong to a specific candidate.

Architecture & control flow

1. 01

   Parse hard requirements into required routes

2. 02

   Run module-level solver, limit, benchmark and subprocess checks

3. 03

   Bind priors, deltas and experiment reports through candidate_id

4. 04

   Aggregate required-route gates into a candidate-bound package

Architecture outline derived from this version’s control flow.

What this version changes

Replaces several execution-shaped fields with executable module checks.

Inputs & outputs

Inputs

Candidate genomes, required routes, priors and task tests.

Outputs

Candidate-bound reports, numeric benchmarks, execution records and route verdicts.

Implementation & evidence scope

The archive’s default run is still Archimedes, so it should not be described as an end-to-end Dedalus default. Its mathematical checks also center on a preset graph-diffusion task.

Code & bundled material

The introduction draws on bundled notes, changelogs and central code. Software tests, synthetic diagnostics and scientific effectiveness use different evidence standards.