# PIXIE: join, contribute and resume

Updated 18 October 2026. One workflow across the existing ecosystem; GitHub holds shared contribution records.

[Open PIXIE](https://pixie-project-companion.ibloud-ivxx.chatgpt.site) · [Ecosystem](https://ecosystem.loptrlab.com) · [Training](https://loptr-lab.github.io/training/)

## Why we are testing PIXIE

Loptr Lab’s ecosystem is the testing ground for an open-source multimedia training studio. PIXIE is its contributor companion, helping people join, learn, coordinate work and return with context intact. Contributor testing is the immediate goal; a finished commercial app is not required before testing begins. Pilot evidence should inform a concrete project pitch and a sustainable revenue model consistent with the existing mission, open participation and project licenses.

Read the [owner-confirmed studio direction](https://github.com/ibloud/ibloud.github.io/blob/main/docs/ECOSYSTEM_ARCHITECTURE.md#studio-viability-pilot--owner-confirmed-direction-18-october-2026). This workflow is a pilot of contributor coordination, not evidence that public plugin distribution, studio viability or commercial readiness is complete. Record concrete contributions, learning outcomes, handoffs, repeated explanations and defects. Build the pitch from those findings, with business-model assumptions labeled as hypotheses.

## Contributor steps

1. **Start.** Open the existing PIXIE companion, choose a project and read its repository’s contribution instructions. Training is the default starting point.
2. **Choose a task.** Use “How can I help?” for a bounded suggestion. Then choose “Join & coordinate” and describe the proposed task, progress, blocker and next action.
3. **Review and submit.** Review the checkpoint draft and its public visibility. “Review & submit on GitHub” opens a prefilled issue in that project. You submit it using GitHub; PIXIE does not post automatically or enroll you.
4. **Agree scope.** Record maintainer agreement, time, credit and exit arrangements in the issue. A suggestion or issue does not itself assign a task or grant access.
5. **Keep coordinated.** Add progress comments with evidence, blockers and the next action. Keep the issue URL. GitHub holds the shared history.
6. **Resume.** In a later session, choose the same project, select “Resume my checkpoint” and paste the issue URL. PIXIE reads the current public issue and latest comment page. Open the full GitHub discussion for older comments.

## Remember your place on this device

Optionally choose “Remember this checkpoint on this device” in PIXIE. It stores only the public issue URL and project in this browser. “Forget saved link” removes that pointer; the GitHub issue remains. This does not synchronize across devices or sites.

## What persists

Submitted GitHub issues and comments persist across sessions. Drafts are not saved until submitted. PIXIE does not read private chat history, automatically remember your issue URL, create formal membership, grant repository access or collect email addresses. Keep private material out of public issues. Current open contribution is voluntary and unpaid.

## Build and release protocol

1. Edit the existing companion source; preserve its Site and App identities. Keep runtime code and browser instructions aligned.
2. Run the companion tests, TypeScript checks and production build. Verify a checkpoint draft does not post, another session can read a submitted issue, and wrong-project URLs are rejected.
3. Commit/push exact source, save and deploy that source through the existing Site, then verify public HTTP/API behavior. Record source commit, saved version and deployment result. GitHub source review and Sites deployment are distinct operations.
4. Keep ecosystem and training links pointing to this same companion. A domain change is not required. Do not create another map or replacement service.
5. To distribute the ChatGPT plugin publicly, verify the supported submission path for the existing Sites App. Prepare the verified publisher identity, website/support/privacy/terms URLs, actual icons, five positive and three negative review cases, a real demo recording, country targeting, commerce declaration and release notes.
6. Prepare the public submission separately from the private original. Author uploads omit private App bindings and use verified MCP configuration; preserve canonical App ownership and Portal-generated bindings. Do not create a duplicate private plugin.
7. Configure and test the exact saved submission. The verified publisher completes legal attestations and required scans. Submit for review, publish the approved version, and verify installation and tools with a separate permitted account. A private update, public website or uploaded draft is not a public plugin release.

## Current boundary

The companion website is public. ChatGPT plugin distribution remains private until the public release process completes. GitHub/AT Protocol membership signup and private participant storage are not implemented by this checkpoint workflow. This is shared project coordination using GitHub, not a claim of an integrated AetherOS system. Reported issue text is not accepted story canon or verified completion.

## Hosting and domain

Keep the existing repositories and domain. ecosystem.loptrlab.com remains the hub; training remains the contributor manual; the existing companion serves the workflow. A subdomain under an owned domain normally requires DNS configuration rather than another registration, subject to the DNS provider’s plan. Custom-domain setup does not publish the ChatGPT plugin. Do not repoint the existing ecosystem hostname or migrate MCP/OAuth addresses without testing their supported configuration.

## Source and maintenance

The companion README carries implementation commands and release details. Before editing, reconcile the actual deployed source and the GitHub contribution source; do not assume automatic synchronization. Follow the [canonical ecosystem architecture](https://github.com/ibloud/ibloud.github.io/blob/main/docs/ECOSYSTEM_ARCHITECTURE.md) and [PIXIE interface track](PIXIE_ANNOUNCER_TRACK.md). Existing work is reused; failed retrieval is not proof of loss.