-
-
Project Foreman identity graphic
Why I built Project Foreman
I started using ChatGPT while I was recovering from substance addiction. It gave me an outlet and something constructive to keep my mind occupied while I tried to get back on my feet.
At first, I was researching how to build a better connection with my spirituality. That led me to writing a study curriculum. Then I realized I could make a better curriculum if I had good source material behind it.
One thing kept leading to another. The curriculum became a research project. The research project became thousands of source files. Then I needed ways to organize those files, process them, trace where information came from, and preserve the decisions I had already made.
That is how I began building tools.
I bought an old refurbished Dell OptiPlex to play Anno 1404, not to build a tool forge. At the time, I knew almost nothing about computers, files, folders, Python, PowerShell, APIs, or GitHub. ChatGPT taught me by helping me solve one real problem after another.
Somewhere along the way, I fell in love with the process of weaving digital systems together into unique but stable tools for problems I encounter in archival work. Tool building has now become just as important to me as the original research project.
Project Foreman came from one of those problems.
My conversations with ChatGPT had grown long enough to contain real project history: decisions, rejected ideas, requirements, corrections, sources, unfinished tasks, and evidence of what had actually been completed. The work was still there, but finding it again—and knowing which parts could be trusted—was becoming increasingly difficult.
I needed a way to recover a project from the conversation that created it.
What it does
Project Foreman converts a long AI conversation into a traceable project workspace. It produces a Project Spine, Decision and Authority Ledger, Continuation Brief, Source Trace Index, and Evidence Boundary. It can then export the recovered project as a validated nine-file package with a manifest and SHA-256 checksums.
The public version runs locally using Python’s standard library. It does not require an API key, external Python packages, or an AI connection at runtime.
How ChatGPT, GPT-5.6, and Codex helped
Project Foreman grew from conversation-recovery research that existed before Build Week. During Build Week, GPT-5.6 working through Codex helped me decide what to enter, audit the existing project, define the release requirements, organize the work, implement and test the public application, reconcile inconsistent evidence, prepare the website, and produce the demonstration.
It also caught compliance problems I had missed, including requirements that needed to be added to the README and video before the submission was viable.
I remained responsible for the product idea, scope, decisions, public claims, merges, publication, and final acceptance. Codex could propose and implement work, but I decided what became part of the project.
What I learned
The hardest part was not simply generating a summary. I needed the application to distinguish my decisions from assistant suggestions, connect recovered claims to exact source turns, and admit when evidence was missing.
I also learned how much I enjoy building tools. ChatGPT assists me every day in being creative, being productive, and looking forward to a brighter future while doing something I genuinely enjoy.
Project Foreman is a useful product, but it is also part of that larger story, 5.6 Sol and this build challenge is what brought it all together.
Log in or sign up for Devpost to join the conversation.