ChangePlane

The description states that ChangePlane is a repository-scoped GitHub App designed to enforce an independent control layer between AI agents and code repositories. It uses:

• GitHub Actions, Checks API, Apps, and Refs
• GPT-5.6 Luna as the AI model for proposing repairs
• A controller that validates and applies patches after review
• OpenAI Responses API, Playwright, React, Node.js, Vercel, and libsodium

It is described as a system that:

• Watches pull request updates
• Binds the exact commit, approved paths, and behavioral checks to a revision-bound contract
• Reproduces failures with deterministic CI evidence
• Gives GPT-5.6 only the allowed evidence and source from approved paths
• Validates the proposal in a clean job against current head, protected files, and evidence integrity
• Applies accepted patches through a separately credentialed controller
• Triggers fresh CI on new commits

The system is built to prevent AI agents from both proposing and certifying their own work, with GitHub remaining the final merge authority.

Claim: ChangePlane is a GitHub App that enforces separation of duties between AI agents and code changes.

Evidence: The project description explicitly outlines this architecture, including use of GitHub infrastructure, AI model integration, and control flow logic.

The author states that ChangePlane was built to address a specific problem:

“An AI agent can write the patch, edit the test, push the commit, and report green. If the same actor controls both the change and the verdict, green is not assurance. It is self-approval.”

This is the core positioning claim — that autonomous agents should not be allowed to validate their own work.

The evolution described shows a shift from a purely autonomous workflow to one with explicit trust boundaries:

• AI proposes fixes
• A deterministic harness validates the fix
• A separate controller applies it
• GitHub remains the final authority

Claim: ChangePlane introduces a new paradigm of “autonomy without authority”.

Evidence: The description explicitly states this as a design principle and contrasts it with self-certifying workflows.

The description does not name specific customers or use cases beyond RouteThai, which is described as a production user. It also mentions:

• “Developers stay in their coding agent and GitHub”
• “Install the App, choose one repository, bind one real check, and merge one protected setup pull request”

This implies that ChangePlane targets teams using GitHub with AI agents, particularly those who want to automate code changes but maintain control over final validation.

Claim: The target customer is teams using GitHub with AI coding agents.

Evidence: The description describes the user journey and infrastructure as GitHub-native, without naming specific customers or industries.

The project description does not contain any information about pricing, monetization, or business model. It only states:

• “ChangePlane is now a hosted, repository-scoped GitHub App used with RouteThai in production.”
• “Hosted product: changeplane.vercel.app”

Claim: No pricing or business model details are provided.

Evidence: The description does not mention revenue, subscriptions, or any monetization strategy.

The author describes ChangePlane as built on:

• GitHub App infrastructure
• GitHub Actions and Checks API
• OpenAI Responses API with GPT-5.6 Luna
• Node.js, React, Vercel, Playwright, libsodium
• A controller that validates and applies patches
• Repository-scoped secrets encryption using GitHub’s public key

It is designed to be non-intrusive — no new code host, proprietary agent workspace, or database.

Claim: The product is built on existing GitHub infrastructure with minimal additional dependencies.

Evidence: The description lists the technologies used and emphasizes that no new systems are required.

The author states:

• “ChangePlane is now a hosted, repository-scoped GitHub App used with RouteThai in production.”
• “The public experience recreates that workflow safely: judges can replay the failure, Luna proposal, clean validation, trusted apply, new-head evidence, and PASS without credentials or access to private data.”
• “The replay ends with zero human repair actions. That is the normal path I wanted.”

However, there is no mention of:

• Number of users
• Customer retention
• Revenue
• Adoption beyond RouteThai
• Product roadmap or iteration history

Claim: There is limited evidence of traction or maturity beyond a single production case.

Evidence: The description does not provide any data on adoption, usage volume, or product evolution.

The description does not mention competitors or direct market positioning. It focuses on the unique architecture of separating AI proposal from validation and merge authority.

Claim: No competitive context is provided.

Evidence: The description does not name any competing tools or platforms.

• No third-party verification: All claims are self-reported, with no independent evidence of performance or adoption.
• Single production case: Only RouteThai is mentioned as a user; no broader customer base or feedback.
• Unproven scalability: The architecture is described but not validated at scale.
• No pricing or monetization strategy: Unclear how the product will be commercialized.
• AI model dependency: Reliance on GPT-5.6 Luna, which may not be available or stable long-term.

Inference: Without evidence of traction, customers, or scalability, ChangePlane appears to be a prototype or early-stage product.

1. What is the exact scope of RouteThai’s use of ChangePlane? Is it fully integrated into their CI/CD pipeline?
2. How does ChangePlane handle edge cases such as model refusal, timeout, or malformed output?
3. Are there any known limitations in how the system interacts with different GitHub workflows or CI providers?
4. What is the current roadmap for product development and expansion beyond the current prototype?
5. How do you plan to monetize this product? Is there a pricing model or target customer segment defined?