Project Curiosity

A nonpartisan, receipts-backed platform that helps people understand how members of Congress perform in seconds.

Comment

Inspiration

I was inspired to build Project Curiosity because I wanted a way to understand politics without feeling pushed toward one side’s narrative.

I kept thinking: why isn’t there one place where people can get nonpartisan political information, see the facts, and understand what is actually happening?

It happened to be football season when I first had the idea, so ESPN was the format that immediately came to mind. Instead of player cards with stats, Project Curiosity would have profiles for members of Congress showing their actions, voting data, legislative activity, policy focus, and supporting sources.

The goal is to make Congress easier to understand without telling people what to think.

I did not come into this project with any traditional coding experience. Without ChatGPT, this idea probably would have come and gone. Before the release of GPT-5.6, I did not fully trust the quality of the product output or feel confident that I could identify every important risk.

I do not know what the next problem will be that I cannot solve on my own, but I know there will be one. Human creativity, paired with AI’s technical ability and information recall, has me feeling very optimistic about what can be built next.

What it does

Project Curiosity helps people quickly understand what is happening in Congress.

Users can see:

  • who their representatives are
  • whether they are voting
  • what they are voting on
  • what bills they sponsor or support
  • which policy areas they focus on
  • where the information came from

The platform presents congressional information in a fast, scannable format while connecting important claims to real sources.

How we built it

My role is to come up with the product ideas, decide what provides real value, and determine how the platform should work in the real world.

My development workflow is a collaboration between me, ChatGPT, and Codex.

I discuss product ideas, problems, and priorities with ChatGPT. Together, we turn those ideas into a clear implementation plan. ChatGPT then helps me create detailed prompts and instructions for Codex.

Codex receives a narrow task, such as auditing part of the repository, tracing a data issue, adding tests, or making a specific code change.

After Codex completes the task, I bring its output back to ChatGPT for review. We inspect the changes, identify risks, verify whether the solution follows Project Curiosity’s trust rules, and decide what needs to happen next.

From there, we adjust, test, review, and repeat.

This workflow allows me to remain responsible for the product direction and final decisions while using AI to move much faster through engineering work.

Challenges we ran into

I did not learn about OpenAI Build Week until July 17.

Luckily, when GPT-5.6 became available, my first thought was to use it to perform a full audit of the Project Curiosity repository as soon as possible. That audit began on July 13, which was the first day of the Build Week submission period.

The audit uncovered serious data-integrity problems that needed to be addressed before the platform was ready for beta testing. These included incomplete vote-position data, unclear participation handling, missing official receipts, and situations where unusual congressional votes could be interpreted incorrectly.

The Build Week work focused on strengthening the trust layer behind the platform.

Accomplishments that we're proud of

I did not learn about OpenAI Build Week until July 17.

Luckily, when GPT-5.6 became available, my first thought was to use it to perform a full audit of the Project Curiosity repository as soon as possible. That audit began on July 13, which was the first day of the Build Week submission period.

The audit uncovered serious data-integrity problems that needed to be addressed before the platform was ready for beta testing. These included incomplete vote-position data, unclear participation handling, missing official receipts, and situations where unusual congressional votes could be interpreted incorrectly.

The Build Week work focused on strengthening the trust layer behind the platform.

What we learned

One thing I learned is that keeping up with AI takes a lot of effort.

New models, tools, plugins, and workflows can make my work significantly easier, but they also appear so quickly that it can be difficult to understand where each one fits into the project.

At the same time, the newer capabilities have already improved my workflow considerably. Being able to connect ChatGPT more directly to the repository and use Codex for focused engineering tasks has made a major difference.

I am now getting roughly three times as much work done as I was only a few weeks ago.

The repository audit also taught me that Project Curiosity still had substantial work remaining before it was truly beta ready. That was difficult to see, but it was important. I would rather discover those problems now than publish misleading political information later.

The biggest lesson was that speed only matters if the process still protects accuracy and trust.

What's next for Project Curiosity

The immediate priority is to finish the remaining beta blockers discovered during the audit.

After that, the next steps are:

  • build a consistent project workflow and schedule
  • create an idea parking lot so new ideas do not interrupt active priorities
  • run structured beta testing
  • fix issues that affect trust, clarity, or usability
  • begin Phase 2 only after the beta foundation is stable
  • reevaluate the roadmap based on actual user feedback and development progress

The long-term direction remains the same: prioritize trust in how political data is presented and add the highest-value features at the appropriate time.

Project Curiosity will continue to follow a simple rule:

Trust is more important than feature count.

Built With

Share this project:

Updates