trending

The parts of an AI feature that should not be AI

@OceanMind builds your five-minute practice at request time from two answers: what you want out of the next five minutes, and where your nervous system is right now. A model writes it. This month we moved a good part of that work into ordinary code, and the practices are steadier for it.
Why we changed it

A model is strong at language and at choosing between options. Holding a limit, doing arithmetic on seconds and remembering what you did last week are not that kind of problem. Those three moved to code, where they can simply be computed.

What we changed

The protocol limits are now worked out before the model is asked for anything, and what comes back is checked against them: how slow the breath goes, how long a hold may run at the activation you reported, how long the session is.

We set out to fix retention. The hole was somewhere else

More than 4 in 5 people who registered for OceanMind never did the one thing OceanMind exists for.

Not once. Not a single practice.

We turned off subscriptions. A stranger subscribed anyway.

Something happened this week I can't stop thinking about.

We turned off subscriptions in OceanMind our breathing and meditation app. No paywall, no prompts, no upgrade screens. We decided to focus entirely on the product first.

What almost a year of building OceanMind on iOS taught us about what users actually want

When we started OceanMind on iOS, our thesis was pretty straightforward:

Build structured programs combining breathwork, movement, and meditation. Ask people for 30 40 minutes a day. Give them a path that produces meaningful results over time.

The interesting part is: it worked.

OceanMind: Breathwork - AI that reads your nervous system and calms it down

Tell OceanMind what you need right now. Calm down before a meeting, focus, reset after stress or switch off before sleep. It chooses the right technique and uses AI to personalize the practice to your state and situation.

We rebuilt our onboarding 4 times. Here’s what we learned about the moment users actually commit.

Our first onboarding had everything. Practice type selector. Duration preferences. Animated intros. Two different breathing tracks to try before you even signed up.

We thought more context = more commitment.

Would you trust an AI to choose your meditation or breathing practice for you?

I ve been thinking a lot about the difference between AI as a recommendation engine vs. AI as an actual decision-maker in your wellness practice.

Most apps I ve seen use personalization to mean: you filled out a goals form at signup. That s it. Your state on day one defines every session you get on day 300.

We couldn't crack retention. So we pivoted to AI. Here's the honest story.

When we built OceanMind, we had a strong conviction: most meditation apps are too shallow. Real, lasting change required real, structured commitment.

So we built a full course movement practice, breathwork, then meditation all three done back-to-back. The science was solid. The results for users who actually completed it were remarkable. Sessions ran 30 45 minutes. We were proud of it.

Then we looked at our retention data. It was brutal.

OceanMind - AI that learns exactly how to calm you down

Anxiety doesn’t follow a schedule. Most wellness apps give you the same routine regardless of how you actually feel. OceanMind is different — it reads your state and builds a personalized session from breathwork, somatic release, cold exposure protocols, and guided practices. The AI adapts every time. Open the app feeling overwhelmed, and it meets you there. What practice has helped you most when anxiety peaks? We’d love to know.
•

12mo ago

OceanMind - Breathwork, movement & meditation - in one connected journey

OceanMind - a structured wellness journey uniting breathwork, movement & meditation in perfect sequence. Each day builds on the last, making your transformation faster, deeper, and measurable.