Renigma

Context & intent

Requirements, technical models, and computational processes can change form without making the important distinctions easy to see. Renigma investigates which relationships must survive those changes.

Make interpretation reviewable, compare alternatives, and assess whether a transformation preserves a declared result.

Project context and development approach

This is an independent project from my portfolio across FractalDesignZ and ZuluYokohama. My practice centers on directing work with substantial assistance from agentic AI and development harnesses. The mechanisms and evidence below describe this project; they do not attribute every component to manual authorship or a specialist review.

Read how I work with AI →

Implemented mechanisms

1. A bounded language interface turns supported requests into editable plans and source-linked memory operations.
2. Scenario branches retain predictions, observations, and corrections as distinct records.
3. Transform Lab studies operation-preserving rewrites, intermediate reuse, and calibrated CPU/CUDA selection with checked outputs.

Evidence & merit

Source review describes implementation; recorded tests describe the cases that were exercised; functional review considers observed behavior. Independent specialist validation is a separate activity and is claimed only when documented. An automated checker does not itself establish independent professional review.

Reviewed implementation and controlled studies support these bounded mechanisms. The public exhibit uses synthetic examples and explains the scope of the work.

Explicit preservation conditions and retained comparisons make a transformation something that can be inspected, rather than simply asserted.