EulerForge

Philosophy

Verification should inspect the candidate itself, with compilation, execution, benchmarks and evidence linked by the same identity.

Architecture & control flow

1. 01

   Resolve bilingual hard requirements and a domain contract

2. 02

   Construct and compile equation, algorithm and architecture genomes

3. 03

   Run solvers, property and metamorphic tests, benchmarks, ablations and evidence-relation checks

4. 04

   Aggregate Aurora verdicts by required route

5. 05

   Build and audit the package, then recompute artifact hashes

Architecture outline derived from this version’s control flow.

What this version changes

Connects requirements to candidate compilation, execution reports and evidence binding.

Inputs & outputs

Inputs

A research direction, mode, required routes and compiler/task contracts.

Outputs

Candidate-bound experiment reports, strongest-prior deltas, ablations and an auditable research package.

Packaging and integration variants within one generation

EulerForge

Turns candidates into compilable objects: equations use registered operator semantics, algorithms face subprocess tests and fixed benchmarks, and architectures produce executable forward paths, all bound to evidence and prior deltas.

Default lite is a deterministic demonstration containing a sorted() wrapper and preset graph diffusion; “replay” mainly checks JSON artifact hashes here. Passing gates does not establish independently proven originality.

EulerForge · Claude Code adaptation

Keeps EulerForge’s core algorithms while adding Claude Code and retrieval wiring: request-ID-matched agent responses, an AnySearch network transport and candidate-context enrichment.

Python does not launch Claude Code Task agents itself; pending requests require host dispatch. Additional retrieval hits are not evidence of scientific effectiveness.

EulerForge · Full-stack package

Packages EulerForge as a complete project with coverage manifests, operator scheduling, search trees and host-agent review templates, making the toolchain easier to inspect as a whole.

Importing a module or writing a review template does not establish that it improves discovery; app-level agents still require host execution.

Implementation & evidence scope

Default lite is a deterministic demonstration containing a sorted() wrapper and preset graph diffusion; “replay” mainly checks JSON artifact hashes here. Passing gates does not establish independently proven originality.

Code & bundled material

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