Introducing the AI Programme Personalisation Engine

Release v1.0.89

---

Today we're shipping the feature that sits at the heart of what we set out to build: an AI engine that takes everything we know about a member — their training history, their personal records, how recovered they are today, and what they're actually trying to achieve — and turns it into a weekly programme written specifically for them.

The problem with generic programming

Most gym software delivers the same programme to every member on the same track. That works as a starting point, but it leaves a lot on the table. A member who slept four hours and logged a heavy session yesterday needs a different prescription than a member who is fully rested and trending toward a new squat PR. Coaches know this intuitively — but they can't manually tune a programme for every athlete every week.

The AI Programme Personalisation Engine closes that gap.

What the engine does

Every week, before members check in, the engine runs a pre-computation pipeline for each member. It reads four sources of truth:

• Session logs — What work has been done, across which tracks, and at what volume.
• PR history — Where the member's strength and fitness are trending.
• Readiness scores — How recovered and ready to train they are right now.
• Goals — What they're actually trying to get out of their training.

From these inputs, it produces a personalised weekly plan: a blend of training tracks weighted to the member's goals, with loads adjusted to reflect their current state.

Claude writes the coaching context

A spreadsheet of sets and reps is useful. Understanding why you're doing them is better.

For every personalised plan, we use Anthropic Claude to generate a coaching narrative — a short, plain-language explanation of why this week is structured the way it is, what the member should focus on, and how it connects to their longer-term trajectory. Coaches see it. Members see it. It replaces the silence between the athlete and the programme.

Built to be fast and reliable

Personalisation at this depth can't happen in real time on demand — there's too much to compute. We use Inngest to run weekly pre-computation jobs asynchronously, so every member's plan is ready and waiting before they open the app. Inngest also gives us reliable job retries and observability into the pipeline, so coaches and gym owners don't have to worry about members landing on an empty plan.

What this means for coaches and owners

Coaches don't lose control — they gain leverage. AI-generated plans surface in the coach dashboard for review before members see them. Coaches can edit, annotate, or override anything the engine produces. The AI is doing the first draft; the coach signs off.

Gym owners get a platform where every member is receiving individually tuned programming at scale — something that would otherwise require a personal trainer for every athlete.

What's next

This release establishes the foundation. Upcoming iterations will tighten the feedback loop between completed sessions and plan adjustments, introduce injury-risk signals into the readiness model, and give coaches more granular controls over blending weights and load ceilings.

The full technical details are in the feature documentation and the changelog.

---