Agentic Coding Agent

what the cockpit would look like

Inspiration

AI coding agents can move quickly, but it is often hard to understand what they were authorized to do, why they made a decision, or how to verify an outcome. We built Nova to make governed agentic coding visible and inspectable.

What it does

Nova is a constitutional agentic coding system. It models the path from intent through evidence, authority, execution, and verification. The Operator Cockpit presents that lifecycle alongside runtime state, receipts, continuity, replay readiness, and multi-agent coordination.

How we built it

The project combines a TypeScript and Node.js governance spine with a React/Vite Cockpit. During OpenAI Build Week, we used Codex and GPT-5.6 to extend the Cockpit and governance workflow, improve the evidence-oriented interface, and publish a browser-based demo.

Challenges we ran into

The central challenge was staying honest about system state. A governance interface must distinguish verified evidence from assumptions instead of presenting attractive but fabricated activity. We designed the Cockpit to surface that distinction directly.

Accomplishments that we're proud of

A public, browser-based Operator Cockpit for inspecting the governance model
A clear intent -> evidence -> authority -> execution -> verification lifecycle
A reproducible codebase and a public demo video for judges

What we learned

Agentic workflows become more trustworthy when authority, evidence, and verification are first-class parts of the developer experience.

What's next

We plan to connect the public Cockpit to more live governed execution sessions, expand receipt and replay tooling, and deepen the multi-agent coordination surface.

Built With

agentic
api
ciems
computing
constitutional
docker
evidence
fastapi
governance
gpt?4o
isl
kernel
kubernetes
langchain
layer
next.js
node.js
openai
postgresql
prisma
python
react
typescript
v2.0
vercel

Created by

I led the project's constitutional architecture, AI governance model, and overall product direction.
My primary contributions included:
Defining the constitutional computing architecture and governed execution model.
Designing the Intent → Evidence → Planning → Execution → Validation lifecycle.
Developing the constitutional governance concepts, evidence-first methodology, and replayable decision framework.
Shaping the product vision around accountable, explainable, and auditable AI systems.
Creating the technical architecture, specifications, and documentation that guided implementation.
Defining the constitutional contracts, governance boundaries, and conformance approach used throughout the project.
Leading the overall system design and ensuring the implementation aligned with the architectural principles.
Preparing the project narrative, demonstration strategy, and presentation for Build Week.

Darz' Morris
I designed and built the Agentic Coding Agent as a constitutional computing demonstration of governed autonomy. My contribution spans the full architecture, implementation, and conceptual framework:

Constitutional Architecture
Created the multi‑layer constitutional execution model

Defined the Intent → Evidence → Planning → Execution → Validation pipeline

Authored ISL v2.0 refinements, including intent/evidence/authority separation

Designed the Constitutional Inference Contract (CIC) and Constitutional Continuity Contract (CCC)

System Design & Engineering
Built the agentic reasoning loop and governed autonomy constraints

Implemented the Evidence Layer, validation contracts, and replayability mechanisms

Integrated constitutional safeguards into code generation and refinement

Developed lineage tracking, temporal replay, and the constitutional state engine

Technical Implementation
Engineered the backend, agent logic, and constitutional artifact lifecycle

Implemented structured intent parsing, evidence generation, and conformance checks

Created the governed execution environment and constitutional verification routines

Research & Innovation
Extended CIEMS sovereignty stack with constitutional computing principles

Formalized the Evidence Layer as a first‑class architectural component

Advanced the concept of constitutional software generation and governed intelligence

Project Narrative & Documentation
Authored all constitutional descriptions, architectural explanations, and system rationale

Wrote the Devpost project story, technical sections, and conceptual framing

Ensured the project communicates the principles of constitutional computing clearly and rigorously

Jon Halstead
Engineer of Bayesian phylogenetics, proto‑language systems, 4D rendering, and advanced governed intelligence architectures.

Updates

Jon Halstead started this project — Jul 21, 2026 09:02 PM EDT

Leave feedback in the comments!

Log in or sign up for Devpost to join the conversation.