CHAPTER 01October 11, 2023
An early creative precursor
Themed mantra and password experiments in Colab explored the creative territory. This was an earlier exploration; a direct code lineage to Magic8 has not been established.
The development story
Every failure changed
the next experiment.
A chronology reconstructed from the development record. Historical milestones describe what was observed at the time; they are not a claim of current release readiness.
CHAPTER 02August 7, 2025
The Magic8 concept emerges
An Android and language-model idea brought together themes, complexity controls, and the shake-the-ball interaction. An HTML / Netlify phone prototype generated random output, while theme and readability experiments struggled. Language-model integration stopped at an API error.
CHAPTER 03August 10–11, 2026
The security model changes
The design separated AI fortune text from Web Crypto random secrets. EFF word-list passphrases of six to eight words became part of the direction. The accepted behavior and reported pull-request merge marked a shift toward a more explicit security model.
CHAPTER 04August 12, 2026
Android baseline and a first vault
Capacitor and PWA work introduced credential saving, but delivery back to the browser failed. Audio, motion, sonic fingerprint, and TOTP ideas were explored as experiments; these did not establish a production MFA system.
CHAPTER 05August 13, 2026
A narrower, testable workflow
The project focused on a concrete path: generate → Magic8 Vault → Magic8 Autofill → matching form. Sharing and external-vault paths were removed from scope.
CHAPTER 06August 18–27, 2026
Limited Autofill wins
Early reports included Firefox Autofill success, work on new-password and confirmation fields, and a successful Warmind flow. Shopify still failed. Compatibility became a site-by-site problem requiring direct observation.
CHAPTER 07September 9–13, 2026
Regression discipline takes hold
Stabilizing Test 22 became more useful than repeatedly changing build numbers. A regression harness and controlled checks helped reproduce failures. A reported Shopify success was corrected to a failure—a reminder that a green test and a filled field are different evidence.
CHAPTER 08September 28, 2026
Antstream save and reuse
A user-reported Firefox flow created an Antstream account, saved credentials, and used the saved vault credentials for a fresh login. This demonstrated a specific save-and-reuse journey; it did not prove every generated-password signup flow.
CHAPTER 09October 20–23, 2026
The real-device gap
Controlled Test 107 checks passed while the expected popup was absent on the user’s device. The direction shifted toward a personal beta, with repeatable physical-device results still required.
CHAPTER 10October 26–28, 2026
From local tests to physical evidence
Mac mini development, an isolated AuthCheck flow, and emulator journeys supplied more controlled evidence. Wireless ADB testing recorded a Samsung synthetic end-to-end pass. That flow used an explicit email; automatic email capture remained unavailable.
CHAPTER 11October 28, 2026
Hardening before release
The development record describes a critical dependency-advisory fix, verified-signer and Firefox-only beta policy work, and vault, backup, and release guards. Device checks continued. The current draft remained unmerged; a public test signer was not a production signer.
CHAPTER 12Next · Pending
Release readiness is its own milestone
Protected signing, device and artifact gates, independent review, and a synthetic private beta remain part of the path forward. Automatic email capture and public release are not complete.
Controlled success is not universal site support. The milestones above preserve that distinction throughout the story.
