Ryan Mouhajer
<-Back to home

Bloom: AI Coaching That Turns Relationship Insight Into Changed Behaviour

Bloom — iOS & Web

2025–2026 (pre-launch)|Founder & Product Designer — strategy, research, behavioural design, AI system, brand, copy, and code

Web

ReactTypeScriptTailwindVite

iOS

Native Swift

Backend & AI

SupabaseClaude API

Designed & built with

Bolt.newClaude CodeClaude DesignFigma

Context

Bloom is a coaching app for attachment patterns, the wiring that shapes how people bond, trust, and handle closeness in relationships. Naming the pattern is the easy part. The harder problem, the one that actually matters, is changing it.

What I built

Bloom is an iOS coaching app: a quiz names an attachment style, anxious, avoidant, fearful, or secure, then daily practice and an AI coach work to change it, one system, not a report with a chat bolted on. Built on real therapy frameworks: EFT, IFS, Polyvagal theory, Gottman Method.

AI coach
conversation
reads · writes
The profile
triggers · defences · readiness
reads · writes
Daily practice
exercises & reflections

A coach for everyday attachment triggers

The coach is trained on the same frameworks as the exercises. Three decisions made it genuinely helpful, not just fluent.

  • •It opens with context, not a blank chat box: people tap prewritten suggestions shaped by the quiz, so the first message, where they usually quit, runs on recognition instead of recall.
  • •It follows a clinician's order, calm first, then reflect, then act: reflection doesn't work while someone's still overwhelmed, so the coach holds back until that eases.
  • •It learns from actions, not just chat: which exercises someone finishes or skips, and how they feel before and after each one, all feed one profile the coach uses, so it sharpens with use.

The reveal: turning results into a story

The result was a personal report, and a long scroll buried the insight in the first version. I rebuilt it as a stories format, the same tap-through pattern from Instagram, so people could lean on a mental model they already had instead of learning a new one. Each insight got its own chapter through progressive disclosure, rather than surfacing the whole report at once.

The first three chapters are free and shareable, working as a viral loop, and they end right where the Zeigarnik effect, the pull of an unfinished thing, is strongest.

The daily practice

That one action a day takes a few forms, each built on a real framework and doable in a few minutes.

The daily card

The method surfaced as lineage (Emotion-focused · EFT), a plain "what you'll do" chip, and the why behind a tap.

Feeling Namer

Naming a feeling helps the brain hold it. One idea per beat: a family of feelings, then a precise word, then how strong.

Voice-guided grounding

A breathing ribbon that follows the read, one instruction at a time. "Done early" is always there; practice is never a trap.

The Pivot

Bloom started as a web-only insight product, a quiz, a report, an AI coach. It's strong at the top of the funnel: it acquires and activates, naming a pattern in the first session. But behaviour change is a retention problem, not an activation one. A pattern reinforced for years only shifts with sustained engagement, repeated practice over weeks, and a web report has none of the mechanics that drive it: no daily loop, no push notifications, no native surface to bring someone back tomorrow.

That's why the real product is now iOS: daily modules matched to each attachment style, feeding a coach that reads what actually happens in them. Web is the free layer now, the lead generation machine: quiz and insights that hand people into the app.

<- All projectsNext projectFlightCalls ->