How we built Juniper with Specode in 2 days instead of 8 weeks
An 8-week healthcare MVP is mostly handoffs
We scoped Juniper the way a cash-pay practice would brief us: patients book, pay a deposit, and do intake from home; providers run their own schedules; admins see all of it.
Our developers put that scope at 5 to 8 weeks. The most optimistic estimate was 2 weeks, and it assumed zero back-and-forth.
That back-and-forth is where the weeks go. Every change routes from product owner to designer to developer to QA and back. Compliance adds a layer that never makes it into the brief: audit logs, MFA, role-based access.
Everything before the visit, in one app
Specode already knew most of what Juniper needed
.webp)
.webp)
.webp)
You can build this in Lovable and it takes you longer. Same with the other builders, because you have to specify every pattern that Specode already knows.
You can hit the same speed in Claude Code if you have a large skill library, but you only have one if you sat down and built it, and you only build it if you already have years of know-how building in that space. You have to hand the agent that knowledge in some form, and you pay for it with your time.
A live app worth building on
The verdict was keep building. The review found strong UI and UX, and code cleaner than a typical AI-built project.
What was left for a developer: 1 to 2 days of refactoring, plus the release. The extra security work Juniper needed is now built into the HIPAA agent.
.webp)
.webp)
.webp)

.webp)
.webp)
.webp)