Lens-First Active Generation

Philosophy

Different contributions require different evidence. The value of the question, the claimed contribution and the work that can actually be delivered should align. Evaluation feedback reshapes the idea rather than merely adding components or raising its self-score.

Architecture & control flow

1. 01

   Start from a research direction and optional literature; commit to a contribution lens and resolve the field and baseline classes.

2. 02

   Branch across kernels, anchors and literature gaps, distinguishing a mechanism from a combination of existing parts.

3. 03

   Apply lens-specific scoring and audits of risk and applicability; reconsider the framing when refinement reaches diminishing returns.

4. 04

   Deliver a recommended direction, alternatives, strongest-prior comparisons and staged tests for human review.

Architecture outline derived from this version’s control flow.

What this version changes

Compared with the V2 active-exploration design described in the archive, V3 selects the contribution lens before generation. Embedded V3.1–V3.4 patches address self-evaluation calibration, theorem conditions, universal claims, measurement evidence and future commitments.

Implementation & evidence scope

This archive is primarily a model-driven Markdown research protocol with kernel and anchor libraries, rather than a persistent Python search engine. The two older ZIPs are identical and form one archive containing V3.4 patches. Its internal scores do not establish a guarantee of scientific novelty.

Code & bundled material

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