The living website / Field notes
AI Jungle website improvements and delivery evidence.
A useful change starts with a real observation. This journal shows what is live, what was observed, what still needs review and what the next measurement should answer.
The latest website review findings
Report dated
No experiment crossed the evidence rules in this report.
This is the latest report included in this release, not a live traffic counter. A booking click is a signal of interest, not a confirmed appointment.
Website delivery journal and released changes.
Each entry has a status and a trail back to the code or report that supports it. The release was checked at on commit 542fb966.
Evidence before conclusion
- Verified in release
The free assessment now has a real handover
A signed-in visitor can start a live assessment, describe one real workflow to Baibot, receive provisional opportunities and follow the recommended next step. Google sign-in and email magic link are the available entry points.
Evidence
- src/routes/assessment/live.tsx
- src/routes/api/assessment/live/session.ts
- src/routes/api/assessment/live/message.ts
- src/components/assessment/live-entry.tsx
Next question
Watch completed assessments and chosen offers before calling the handover effective.
- Verified in release
The funnel is measured privately
The live route records authentication, assessment start, completion, recommendation and offer-choice events through the PostHog proxy. The private assessment path keeps session replay disabled.
Evidence
- src/routes/assessment/live.tsx
- src/lib/analytics/events.ts
- src/lib/analytics/client.ts
- src/components/analytics/posthog-analytics.tsx
Next question
Use the aggregate observation reports to decide whether a change is worth proposing.
- Verified in release
Five English and French articles point to the diagnostic
The current article set gives readers a route into the free signed-in diagnostic and, where deeper interpretation is useful, the optional paid assessment with a human consultant. The links prove routing, not conversion lift.
Evidence
- src/data/static-blog-posts.ts
- src/data/editorial-blog-posts.ts
- src/components/assessment/live-entry.tsx
Next question
Observe article clicks, authenticated starts and completions in the same reporting window.
- Observed
The release has no conversion result yet
The latest bundled scout report has no booking-click data and no conclusion of improved conversion. The OAuth handshake was verified, but no new completed-account result is claimed from that check.
Evidence
- src/data/evolution-snapshot.json
- loop/reports/2026-09-07-scout.md
- Release verification at 2026-09-09T03:35:24Z
Next question
Keep missing metrics as no data until the next reviewed report supplies them.
- Preparation
The next change still needs a human decision
The scout can prepare a bounded proposal from aggregate evidence. A person must review the hypothesis, visual result, success measure and stop rule before a release or flag change exists.
Evidence
- loop/README.md
- loop/scout.mjs
- loop/propose.mjs
Next question
Wait for a reviewed observation window before proposing a new experiment.
How we evaluate a website improvement.
Observe website behaviour
The scout reads consented aggregate events and keeps missing evidence as no data. It does not inspect a private visitor record.
Propose a website improvement
The weekly scout can draft a small experiment from a reviewed report when the documented evidence rules are met.
Review the website change
A person reviews the hypothesis, the visual result and the measure that would make us stop.
Release and verify the website update
An approved change is released and checked in the browser. A proposal remains preparation until that review exists.
Website observations and supporting evidence.
Aggregate observations only. “No data” does not mean zero visitors or zero interest. These weekly reports do not establish an improvement in conversion.
| Report / UTC window | Booking clicks | Diagnostic starts | Draft proposals |
|---|---|---|---|
| 2026-09-11 → 2026-09-18 (exclusive) | No data | No data | 0 |
| 2026-09-04 → 2026-09-11 (exclusive) | 1 | 5 | 0 |
| 2026-08-31 → 2026-09-07 (exclusive) | No data | No data | 0 |
| 2026-08-28 → 2026-09-04 (exclusive) | No data | No data | 0 |
| 2026-08-21 → 2026-08-28 (exclusive) | No data | No data | 0 |
| 2026-08-14 → 2026-08-21 (exclusive) | No data | 1 | 0 |