Inspiration
Public records are often available, but they are not always usable for decisions.
I saw this while building FloridaLicenseIndex.com, a separate prior project focused on license verification. The hard part was not just finding public records. The hard part was preserving what the record said, when it was checked, what evidence supported it, what changed, and what still needed human review.
ProofPacket Studio was built from that problem.
What it does
ProofPacket Studio turns public-record style data into dated review packets.
The app preserves source chains, flags unresolved gaps, generates a review packet, and exports the result as Markdown. The demo uses fictional judge-safe fixtures so the workflow can be tested without live scraping, private data, or API keys.
The core principle is simple: sources stay attached, gaps stay visible, and humans retain judgment.
How we built it
ProofPacket Studio was built during OpenAI Build Week using Codex and GPT-5.6.
Codex helped build the app structure, product workflow, deterministic demo route, tests, documentation, interface design, and Markdown export flow.
GPT-5.6 is used as the interpretation layer: field interpretation, packet language, ambiguity summaries, and verification guidance. The submitted demo defaults to mock mode so judges can test the product reliably without API credits.
Deterministic code handles source-chain preservation, unresolved-gap counts, status comparison, and export behavior.
Challenges we ran into
The biggest challenge was avoiding overclaiming.
It would be easy to make an AI system sound more certain than the evidence allows. ProofPacket Studio is designed to do the opposite: preserve uncertainty, keep source evidence visible, and separate AI-assisted explanation from human review.
Another challenge was scope. Public-record workflows can include licensing, permits, insurance, vendor review, and compliance. For Build Week, I focused on the smallest complete workflow that demonstrates the larger idea.
Accomplishments that we're proud of
ProofPacket Studio is a working product demo with a deterministic recording route, fictional judge-safe fixtures, Markdown export, source-chain handling, unresolved-evidence handling, documentation, and passing contract tests.
I am proud that it is not another chatbot. It is a workflow product for turning messy records into reviewable packets.
What we learned
The main lesson is that AI is most useful when it operates inside a trustworthy workflow.
Codex can turn clear product direction into working software very quickly. GPT-5.6 can help interpret and explain messy information. But the system still needs deterministic structure, visible evidence, and human judgment.
I also learned that the design has to match the seriousness of the workflow. We went through several visual directions before landing on a restrained, product-led interface that better fit a public-record review tool.
What's next for ProofPacket Studio
The next step is to expand from fictional fixtures into real user-provided public-record datasets.
Potential use cases include license verification, vendor review, insurance checks, permits, compliance monitoring, and regulated local services.
The larger vision is simple: public records should not just be available. They should be reviewable.
Built With
- chatgpt
- codex
- css
- csv
- govtech
- gpt-5.6
- javascript
- json
- markdown
- openai
- react
- typescript
- vinext
- vite
Log in or sign up for Devpost to join the conversation.