RELIC

Inspiration

Complex projects rarely fail because they lack information. They fail because important decisions, evidence, experiments, assumptions, and unresolved questions become scattered across documents and project history.

RELIC was built to explore a different approach: turning messy project history into structured, traceable evidence without pretending uncertainty has been resolved when it has not.

What it does

RELIC converts project artifacts into an evidence-backed structure that preserves provenance.

Instead of producing only a summary, RELIC separates material into structured categories such as blueprints, builder designs, platform designs, runtime designs, and unknowns.

The goal is to make one thing clear: what is actually supported by evidence, where it came from, what remains uncertain, and what should happen next.

How I built it

I used Codex as the primary implementation environment and GPT-5.6 as a reasoning and review partner.

Codex was used to inspect the working project, implement and refine the prototype, run tests, generate evidence artifacts, and package the final submission.

GPT-5.6 was used to help reason about architecture, challenge assumptions, identify gaps, review outputs, and guide iterative development.

Rather than treating AI output as automatically correct, the workflow emphasized verification, provenance, explicit uncertainty, and reproducible evidence.

Challenges

The hardest challenge was deciding what the system could legitimately claim.

Project history is messy and ambiguous. A useful system must organize evidence without silently turning inference into fact. RELIC therefore preserves unknowns and keeps claims tied to their source material.

Another challenge was packaging a larger body of experimental work into a focused, reproducible demonstration.

What I learned

I learned that AI-assisted development becomes much more useful when generation and verification are treated as separate responsibilities.

A model can help build quickly, but traceability, evidence, testing, and explicit uncertainty are what make the resulting work inspectable and trustworthy.

What's next

RELIC could evolve into a broader developer tool for reconstructing complex project histories, identifying architectural decisions, tracking provenance across artifacts, and converting accumulated work into actionable system knowledge.

Built With

• codex
• css
• gpt-5.6
• html
• javascript
• openai
• python