Extends idea generation into mathematics, algorithms and architectures, organizing candidates through evidence graphs, strongest-prior checks and route-specific gates.
Philosophy
A research idea should identify its problem, evidence, difference from prior work and a way to be tested.
Architecture & control flow
- 01
Resolve a domain contract and route requirements to math, algorithms or architectures
- 02
Construct typed candidates; plan queries and build evidence graphs
- 03
Compare strongest priors and equivalence; demonstrate recurrent refinement and population meta-rules
- 04
Run route-specific Aurora gates, audit, then persist one package
Architecture outline derived from this version’s control flow.
What this version changes
Establishes the v6 contract, evidence-graph and candidate-experiment scaffolding.
Inputs & outputs
- Inputs
A research direction, mode budget and replaceable retrieval client.
- Outputs
Candidate genomes, evidence graphs, prior comparisons, gate reports and a discovery package.
Implementation & evidence scope
The default path uses offline evidence; numerical-success and architecture-smoke fields include synthetic values, so it does not establish empirically validated new equations or architectures.
Code & bundled material
The introduction draws on bundled notes, changelogs and central code. Software tests, synthetic diagnostics and scientific effectiveness use different evidence standards.
Source references
harness/discovery_v60/__init__.py· 1–29harness/run_agent.py· 1777–1792harness/discovery_v60/pipeline_v60.py· 1–20harness/discovery_v60/pipeline_v60.py· 109–156harness/discovery_v60/pipeline_v60.py· 189–248harness/discovery_v60/pipeline_v60.py· 490–503v6_final_iteration_plan_disco_anysearch_openmythos.md· 1–5harness/tests_v60/test_pipeline.py· 10–73