Platform

One system, one loop, one memory.

The modules below are not features bolted together: each closes part of the loop and reads from the same Brand Brain. Phases are the documented build order, not dates.

01 · Phase 1

Brand Brain

The per-brand, evidence-linked memory: what the brand is, what it has tried and what actually worked. Everything else reads from it.

  • Eight record kinds, never one bucket, each with its own schema, lifecycle and approval rule.
  • Grade, confidence and estimand travel with every evidential record into the UI and into every model prompt.
  • Derived is never source: summaries are caches and can never be cited as evidence.
  • Isolated per brand: no shared retrieval index, cache, context pack or evaluation set spans workspaces.

02 · Phase 1 onward

AI Analyst

Answers “what is happening, why, how sure are we, and what should we test next?” across organic, paid, site and commerce — without inventing numbers or causes.

  • Pre-specified, deterministic analyses; the model narrates results it did not compute.
  • An analysis ledger of every look it has taken.
  • Claims graded by rule; causal wording only where the design earns it.
  • Closes the loop by turning observations into hypotheses and pre-registered experiments.

03 · Phase 1–2

Organic content engine

Turns what the Brain knows into channel-native content that ships on time with human approval, and turns every post into evidence that improves the next.

  • One pipeline from idea to learning; per-channel variants validated against a format registry.
  • Calendar with pre-flight checks, held states and alerts with time at risk.
  • Curves at equal age against the account’s own baseline, with uncertainty.
  • Tracked links connect founder content to the business brand’s site and store.

04 · Text from Phase 2; media in Phase 4

Creative Studio

Copy, hooks, captions, scripts, images, video and voice — grounded in the Brain and the brand’s own performance history, checked by Brand Guard, marked for provenance, and proposed for approval.

  • Clean variants that change one thing only, to test a hypothesis.
  • Repurposing of long video and audio into channel-specific clips.
  • Provenance and disclosure fields per platform.
  • Nothing leaves HypaBrand without approval.

05 · Facts from Phase 1; module in Phase 4

Creative intelligence

Describes every creative consistently — its features — and asks, with honest statistics, which characteristics go with better or worse performance for this brand, on this provider.

  • Associational estimates with intervals, sample sizes and a design label.
  • Decay curves and fatigue observations.
  • Concept-level lineage and a creative diversity index.
  • Outputs are observations, patterns and hypotheses — never rules.

06 · Clocks from Phase 1; analysis in Phase 3

Meta growth intelligence

Turns a brand’s Meta ad accounts into explained, evidence-linked, governed growth decisions; the template for Google Ads and TikTok later.

  • Any metric change outside its band gets a deterministic decomposition of drivers.
  • Account health, fatigue, change history and landing-page match.
  • Meta’s recommendations shown beside your own evidence; applied only if you approve.
  • Meta’s claimed results set beside store orders and site analytics.

07 · Phase 2–3

Measurement ledger

Keeps every lens separate, shows where they disagree and why, states which lens is authoritative for which question, and turns large disagreements into experiments.

  • Meta-attributed, GA4-attributed, store, survey and blended figures never merged.
  • Every revenue or conversion number states its lens.
  • Authority rules decide which lens answers which question.
  • Restatement windows and freshness are visible on every figure.

08 · Capture from Phase 1; analysis in Phase 5

Market intelligence

Learns patterns from competitors and turns them into original, on-brand concepts and testable hypotheses. Not a swipe file, not an ad-spy database.

  • Official APIs, the brand’s own polite crawler, newsletters it subscribes to and captures the user makes — nothing else.
  • Competitor performance is never observed; longevity is a proxy with known biases.
  • Market-derived entries are capped at grade G2; “it works for them” is always a hypothesis.
  • Anti-copying by construction: the Producer never sees competitor copy or media.

Open the portal.

Brand Brain, Analyst, Studio, Calendar, Performance and Experiments, as built so far.

Open portal Security model