02–03 NOV 2026Before launch
A separate Oracle computer
A London-region Always Free ARM VM was provisioned with Ubuntu 24.04.5. The agent ran under a non-sudo account; infrastructure administration remained with the owner.
THE TIMELINE
Six days of business.
And everything before them.
Follow the actual installation, launch, operating changes and early business activity. Evidence ends on 11 November 2026, even though the experiment continues.
ONGOING EXPERIMENT · DATED RECORD01 / DOCUMENTED EVENTS
What happened, in order.
Filter by activity type. Each event distinguishes technical readiness and agent activity from independently verified commercial results.
03 NOV 2026Before launch
Memory, models and a private channel
OpenClaw, ChatGPT/Codex OAuth, a Sol High default, local hybrid EmbeddingGemma memory and private Telegram messaging were independently exercised. Persistent STOP/RESUME controls were tested.
04–06 NOV 2026Before launch
Business tools and recovery
Dedicated GitHub, Cloudflare, email, research-service credentials and Stripe sandbox access passed scoped tests. An encrypted off-server backup, observer and recurring work loop were prepared.
06 NOV 2026Day 1 · 17:20 GMT
The 60-day experiment begins
Agatha's launch was authorised at 17:20:39 GMT. A one-off Astra Extra High session researched options, chose redirect-map review, built a free checker and deployed the first MOLT site. The standing schedule used Sol High every four hours.
07 NOV 2026Day 2
The first legal/commercial blocker
Three targeted agency emails were drafted and conditionally approved, but not sent. The missing requirement was truthful seller identification and contact details. The agent did not invent them.
08 NOV 2026Day 3
The work loop becomes repetitive
The operator explored organic discovery, but ordinary scheduled sessions frequently returned to the unresolved outreach dependency. No independently verified buyer contact or sale was found.
09 NOV 2026Day 4
DEFCON intervention and a second offer
An owner-side audit found 17 of 18 scheduled attempts successful but little substantive work. Journal publication also needed repair. A two-hour DEFCON runner, independent profit flags and broader parallel work were introduced. A £49 Shopify CSV service page followed.
10 NOV 2026Day 5
More public activity, no buyer yet
Longer sessions followed the intervention. Agatha tested competitors, published to directories, opened specialist profiles and joined technical conversations. Those activities were visible, but no paying-customer conversation was established.
11 NOV 2026Day 6
Live checkout readiness demonstrated
A dedicated Stripe live account completed onboarding. A restricted key created private one-time GBP products and Payment Links; the test objects were deactivated. No customer was charged. A buyer-validation focus for 12–18 November was added.
The chronology stops here, not the experiment.
The 12–18 November buyer-validation challenge was installed as a future operating instruction. This publication does not assume it was completed or invent results after the source evidence cut-off.
Read the dated source chronology02 / THE FIRST COMMERCIAL DECISION
Why redirect reviews?
CONSIDERED
Merchant Center fixes
Too dependent on access to customers' accounts and third-party policy decisions to test immediately.
CONSIDERED
Generic SEO audits
Broad, crowded and hard to differentiate as a first small transaction.
SELECTED
Redirect-map review
Build a useful free checker, then test a specific paid migration report with developer-ready priorities.
This was an autonomous strategic hypothesis, not proof of market demand. The first documented session contained 124 recorded tool-call blocks and produced a deployable site, a free checker and the proposed £79 service.