GJE - Guardian Judgment&Execution

I think Judgment and Execution is a governed control layer for Codex and might be good for ChatGPT as well. its purpose is to give users more trust to the agent claims.

Comment

Inspiration

coding agents and Ai in general can produce a answer that sound compelling even when its incorrect. Usually when operating in a large scale , the user cant give the attention for each line of code or each expiation, and if he doesn't have the background he would probably be lost and drown in false claims. i built GJE- Guardian Judgment&Execution , to separate execution from judgment. (i know the name is not original).

What it does

GJE is a governed control later for coding agents. i built it as two part mechanism:

  1. Code Guardian: it helps ensure that the repair remains compliant with the declared architecture and does not produce unsupported or false claims. 2.Judgment&Execution: it continues from the repaired space. Its job is both to support Code Guardian’s work and to classify any problem that still exists, thus acting like an upgraded debugger with an independent judgment layer.

How we built it

I first developed it through a paper called TTC which i wrote. Because I don't yet code at the level required to build a project like this alone, I used GPT and Codex to help translate the architecture into a working prototype. GJE represents only one practical edge of TTC, not the full framework.,

Challenges we ran into

The hardest challenge was learning how to operate Codex and verify the results through PowerShell. In addition, we encountered many problems while adapting the project to Windows.

Accomplishments that we're proud of

Finishing this project, even if it won't get much attention or any at all , it was fun and i'm proud of it.

What we learned

First, I learned how to work with code and the pain it brings. Second, I learned that when working with agents or code, even one small mistake in adaptation can kill the code. Lastly, I learned how to work with Codex.

What's next for GJE - Guardian Judgment&Execution

I think the next step is to strengthen the code, produce stronger evidence of its effectiveness, and bring more mechanisms from TTC into the system.

Built With

Share this project:

Updates