The running list of what the engine still needs, in priority order. We keep appending to this as feedback comes in. The rule from Friday holds: nothing below Tier 1 matters until the output is worth paying for.
Tim's verdict is still “generic.” The engine reads the page and applies templates; it has to produce genuinely site-specific, insightful work. Better generators, richer use of onboarding signal, real drafting where it counts — until this lands, it is a recommender, not a department. Plan (7/31): engine researches the real site + competitors and generates grounded content it's a recommender, not a department.mdash; no-LLM first (within the no-API rule); revisit a metered content-LLM only if still needed after Tim reads it. Focused build next session.
Approving a fix only ships live on our clones today. A real customer needs a connection — CMS credentials, a snippet/plugin, or a publish integration — before “approve” can touch their site. Tim named this: it's the difference between a tool and staff.
Run it on the real site and track actual traffic/rankings over time — the evidence that the value is real, not theoretical.
Each branded domain gets its own marketing page + auth-aware root.
Real screenshots, a demo video, honest proof — replace the placeholder testimonial/video blocks on the splash.
Example customers, competitors, article ideas that measurably drive the work — not decoration.
Week/month/quarter movement, pending vs in-process vs done, and a clean way to feed the engine more info.
Email verification, CSRF tokens, login/reset rate-limiting. (Fine for you two now.)
Cancellations, refunds, per-user activity, Stripe reconciliation on top of the accounts/payments view.
Dunning, failed-payment handling, plan changes/upgrades, receipts.
Filtering, pagination, search — Tim flagged the list will get unwieldy.
Same accounts today, but login is per-domain.
Scheduled reports at the frequency chosen in Operations.
Lifelog is on B2 daily; add the NAS leg once SSH/rsync + a backup user are set up.