The week the tracks went away

The problem

For two weeks the product had two phases and three tracks. Phase 1 introduced a client to the system. Phase 2 picked one of three tracks — Your Agent, Your Code, Your Support — depending on what the client wanted to build. Underneath each track sat a Pro tier and a Max tier, with bundled Sessions and Claude gifts.

It looked clean on paper. It read like apprenticeship. Funders would understand it.

The problem was that almost none of the structure was load-bearing for the actual work. Most clients want one thing: a working system, kept working. The tracks split that into three almost-identical paths. The Phase 1 introduction doubled the install interview. The bundled Sessions tied two products to the same SKU and dragged the price up to where IF pre-approval pressure started showing up.

The structure was carrying weight that didn't exist.

The stack

The other thing that shifted this week was the surface the system runs on.

For most of April, the canon was iPhone. The Action Button was the capture surface, Claude Pro was the model, and the install assumed Apple. That worked for the early clients — Shawn included — but it wrote out anyone on Android, which is most of the disability market on the lower-priced side.

The 2026-05-05 update moved the canon to smartphone. iPhone is still recommended. Modern Android (2020+) is supported with an alternative quick-capture surface. The Action Button stays the recommended capture path on iPhone; on Android, capture lives somewhere else — to be designed, not shipped today, but the door is open.

The Hand off database was renamed in the same pass. It used to be called Claude Code. Now it's called Hand off — the name describes what the bus does, not which agent happens to live on one end of it. A small thing, but it turned a tool-named object into a concept-named one. The agents move; the bus stays.

Principle 7: tools are replaceable, principles aren't. A name that points at a principle ages better than a name that points at a tool.

Architecture

Concept 43 — the Complete Support System — landed on Sunday. It supersedes the block-track architecture and rewrites the product surface from the install outward.

The shape now:

• Neuro Agent is the install. $999, one-off. iPhone or Android. 90-minute install interview, Notion workspace configured, Custom Agent in place. Walk away with a working system.
• Neuro Support is the ongoing service. Around $500/month, monthly billing. The complete support system kept running. Notion Business covered. Maintenance. Agent-driven Help inside the workspace. Agentic Support Work as the human-labour anchor.
• Neuro Max is the premium technical tier. $1,499. Four seats. Four hours of technical 1:1 with Jamie on Notion + Claude Code integration. One-month Claude Max gift.
• Support Sessions are $60 add-ons. 30 minutes, online, available to anyone. Below the NZ ADHD coaching benchmark by design.
• Discovery Calls are free, 15–30 minutes, the universal entry point.

What changed under the hood: Sessions are no longer bundled into the ongoing service. The Phase 1 / Phase 2 / track architecture is retired. The Strategy Session is retired — its three functions split across the Discovery Call, Support Sessions, and the install interview. Tune-Ups are retired — synchronous-time billing collapses to one SKU at $60.

The architecture got smaller and more specific in the same move. That's usually a good sign.

Technical

While the product side reshaped, the operations side got an architecture too.

Concept 44 (the MCP finance pipeline) and Implementation 78 mapped how money moves through the business when Stripe is the front door, Xero is the books, and Notion is the operational surface. Three MCPs, three roles: Stripe handles the customer transaction, Notion holds the human-readable client record, Xero holds the legal one. The integration is event-driven through MCPs that already exist — not a custom backend, not a sub-processor we host. Principle 5 stays intact.

This isn't shipping yet. It's shape. But shape matters before code, and the shape is now there.

Human

The MSD OIA data came back. Disability Support Services pays for support across six categories — ADHD, Autism, Intellectual, Physical, Sensory, Neurological. The OIA gave us the funded recipient counts per category. We already had NZ prevalence figures for ADHD and ASD. That let us do something we couldn't do before: subtract.

Funded ADHD support: effectively zero — ADHD isn't a DSS category. NZ ADHD prevalence: around 100,000–130,000 adults. The unfunded population is the entire cohort, and it's an order of magnitude larger than every other unfunded disability category combined.

That doesn't mean ADHD is the only audience. It means the architecture is the right shape. One install, one price, two reachable populations: people in the funded system who can use IF for setup, and people outside it who pay the same $999. The pricing already sits below the NZ ADHD coaching benchmark, so for a private-pay client, the install costs less than five sessions.

You don't have to pick funded or private. The product reaches both with one shape.

Beyond

A second question surfaced from the same data: what does install-only actually mean, post-pivot?

Before the Notion-native pivot, the answer was simple. Free Claude, free Notion, install configures the surface, walk away. Now the Custom Agent — the half of the architecture that makes the system feel like a partner, not a notepad — lives in paid Notion. Without the ongoing service, the install runs on free Notion, which means the agent layer collapses to capture-and-recall.

That's not necessarily wrong. Capture-and-recall is most of what executive function externalisation needs on the worst day. But it's a different product than what's being sold. The question is whether install-only should be repositioned as a starter version of the system — cheaper, less capable, but real — or whether the install is reframed as the entry to ongoing service and pricing reflects that.

Filed as an idea, not a Monday move. But the question won't go away on its own. The first prospect to ask "what do I actually keep if I don't continue?" forces the answer.

What's next

The architecture stabilised. The data validated the reach. A new question opened. That's a good week.

Next: Phase B HTML implementation of the locked website rewrite, and possibly a thinking session on what install-only means now. Whichever lands first.