Learning
Observing 0 / 20 data points to graduate
⚙Advisory only. Nothing here authorizes a draft, PR, post, or schedule change. Learning surfaces patterns; you decide what to act on.
Learning phase
1
Observing
Watching outcomes;
no hypotheses formed yet
2
Hypothesis
Patterns proposed,
awaiting verdict
3
Confirmed / Rejected
Verdict rendered after
3+ data windows
!
Aborted
Data gap or missing source
made verdict impossible
Graduation progress0 / 20 data points · needs published outcomes
Codex graduates from Observing into Hypothesis once it has 20 published-outcome data points (post live + 14d of metrics). First data point lands after the first website PR is merged and tracked.
Data points by topic type
Buyer checklist0 / 20
Area guide0 / 20
Project showcase0 / 20
Comparison0 / 20
Founder story0 / 20
Construction update0 / 20
NRI guide0 / 20
FAQ0 / 20
Score signals
Seomachine threshold80drafts below this score are rejected before review
Weakest category (30d)Readabilitybased on 1 draft sampled · not statistically meaningful yet
Fastest improving— no signal yetneeds at least 5 drafts across 2 windows
Banned-word hits0 in last 30 drafts
Claim verification rate95%1 unverified claim out of 18 (above 90% target)
AXS preferences inferred
☉
0 preferences inferred
First inferences come from Day 2+ once you approve / deny / edit drafts.
learning/axs-preferences.md
Top-performing past topics
▲
No data yet
Top performers appear after ~20 published posts with 14+ days of GSC / GA4 metrics.
What Codex is stuck on (for learning)
!No published-content outcomes yet
Learning can’t infer what works until at least 1 post is live with 14d of metrics. Currently 0 posts have been approved + published.
!GA4 lead_type dimension not active
Without this dimension, Codex can’t learn which lead source (form / WhatsApp / brochure / phone) correlates with which content.
Open INBOX item ›
!DataForSEO weekly fetch blocked
Ranking-position deltas are a key learning signal. Resuming weekly SERP runs requires account top-up.
!Wiki / OS repo not yet shared
Persona learning + brand-voice inference depend on the Marketing wiki being accessible to Codex.