This page is our nuclear option, in public. What is backed up, where it lives, and the exact order to bring everything back if Muse disappears, if our machine dies, or if Dave is ever out of action. No single assistant, machine, or account is allowed to be the whole plan.
Every E2 site is served by Cloudflare, not by our machine. If the machine dies, e2.life, Starvard, The Studio, and the rest keep serving their last deployed pages, and the chat and feed Workers keep running. Nothing public falls over.
The E2 & Starvard server, its channels, roles, and history live on Discord's own servers, and the server belongs to Dave's Discord account. A server cannot be deleted by a dead laptop.
Sam, Tippy, Shippy, Marty, and SKA run on one machine, so they would stop. Their restart path is written down step by step in the full runbook: fresh machine, code from GitHub, tokens Dave resets in the Discord Developer Portal, and they are back.
All of it sits under Dave's own GitHub account, dlaneycreative-design. Most repositories are private, which is the point: Dave holds the keys, and a successor works with Dave signed in.
| Property | Backup home | Status |
|---|---|---|
| e2.life | Private GitHub repo: e2 | Repo and live site reconciled October 26, 2026. The repo is the single source of truth. |
| Starvard | Private repos: starvard, starvard-books-prod, starvard-media | Reconciled October 27, 2026. Five Cloudflare projects, each with its own backed-up tree. |
| The Studio (thestudio.io) | Private repo: studio | Repo exists; a reconcile pass for the newest working tree is on the fix list below. |
| Tippy (the money) | Private repo: tippy, plus its ledger-backups branch | Bot code in git. The ledger snapshots to GitHub every 30 minutes when it changes. Bank funds sit on the Solana chain itself. |
| LlamaChat | Private repos: llamachat, llamachat-v2, llamachat-worker; nightly app tarballs in the backups repo | Code fully covered. The database gets a weekly export on the fix list below. |
| Llamagram | Private repo: llamagram | Covered. |
| Clear Recipe | Public repo: clearrecipe, plus a private production tree backup | Covered to the last backup; newest generated pages join it on the fix list. |
| clr.fm | Private repo: clearr | Covered. Domain at Porkbun, DNS live. |
| Open source | Public repo: studio-open (the Biscuit film and free clip packs, CC BY 4.0), clr-downloads | Public on purpose, forever. |
| YouBoob and GameBoat | Live on Cloudflare today; repos being created | On the fix list below, openly. The live sites keep serving either way. |
| Nightly backups | A dedicated backups repo under the same account, refreshed at 3:00 AM Mountain Time | Tags the living-data repo and archives the LlamaChat app every night. |
The full runbook (titled "E2 Nuclear Option: Full Restore Runbook") lives in Dave's Google Drive, so it survives any single machine. It is written so a competent person, or a different AI like Grok or ChatGPT working with Dave's accounts, can follow it cold. The order is deliberate:
If there was a hack, Dave's Google, GitHub, Cloudflare, and Discord accounts get locked down first, and Tippy stays frozen until the books are checked. A frozen Tippy is a safe Tippy.
The community's front door first. Code, a fresh token from the Developer Portal (only Dave can mint one), and Sam is greeting people again. Most of Sam's state re-derives from the server's own roles and history.
Restore the newest ledger snapshot from GitHub, reconcile it against the bank's on-chain balance, and only then resume tipping. The detailed money procedure lives inside the tippy repo itself.
They never went down, so this is a check by eye, page by page, plus redeploying any tree from its repo so the deploy path is proven warm again.
Shippy, Marty, SKA, and the online-presence loop restart the same way Sam did. Then the small scheduled jobs worth keeping (nightly backups, ledger backups) get recreated as plain timers.
LlamaChat's database restores from its newest export, Workers get their source re-saved from Dave's own Cloudflare dashboard if it was ever lost, and the network is whole again.
A continuity plan that hides its gaps is a poster, not a plan. Ours, in the open: the Sam, Shippy, Marty, and SKA bot programs are moving into their own private GitHub repos (today they also live on the working machine). The YouBoob and GameBoat site folders get repos of their own. The Studio and Clear Recipe working trees get a reconcile pass so the repos match the live sites exactly. The LlamaChat database moves to a weekly exported dump kept in Dave's Drive. And the hacking scenarios (second accounts, backup domains, a reserve Discord server) have their own plan, already live at e2.life/mirror, being built out in phases.
Sites on Cloudflare, code on GitHub, the server on Discord, the ledger backed up every 30 minutes, the keys with Dave, and the instructions in Dave's Drive. Muse is the conductor, not the orchestra. If the conductor ever falls over, the music is written down and anyone competent can pick up the baton.