super_artifex

From an idea to software running in production.

Nine stages, in order. Each one produces something you can read or use, and the next stage starts only when you approve it.

Research

Before anything is written, the agents study the market, the users and the competition.

The agents
Searches the web, reads competitor products and local regulations, and writes a research brief with sources.
You
Read the brief. It feeds the PRD directly, so there is nothing to approve yet.
research-brief.md
## Who books clinic appointments today
Most patients in Lagos book by phone or WhatsApp.
Missed appointments cost clinics 18 to 25% of slots.
## Competitors
3 apps found. None send SMS reminders in Yoruba or Hausa.

PRD

A product requirements document that says what to build, for whom, and what is out of scope.

The agents
Writes goals, users, features with acceptance criteria, and the open questions it needs you to answer.
You
Approve it section by section, or tell the agent what to change in the chat beside it.
PRD.md
## F3 · Appointment reminders
Send an SMS 24 hours and 2 hours before each booking.
Acceptance: a patient can confirm or cancel by reply.
Out of scope: voice calls (phase 2).

Flows

Every user journey as a step-by-step process, with a wireframe for each screen.

The agents
Turns each PRD feature into journeys: who starts it, every step, what the system does and what can go wrong.
You
Walk through each journey and correct any step that does not match how your users work.
USER-FLOWS.md
1. Patient opens the booking link
2. Picks a doctor and a free slot
3. Enters a phone number, gets a code
4. Booking confirmed, SMS reminder scheduled
If the slot was just taken: show the next 3 free slots

Design

The system design: data model, API, infrastructure and the decisions behind them.

The agents
Chooses the stack for your project, models the data, designs the API and explains each tradeoff.
You
Approve the design, or ask why a choice was made and have it explained.
SYSTEM-DESIGN.md
Appointment(id, doctorId, patientId, startsAt, status)
Availability(doctorId, weekday, from, to)
Decision: bookings lock the slot in one transaction,
so two patients can never book the same time.

Build UI

Every screen built for real on mock data, so you can click through the product early.

The agents
Builds the screens from the flows and wireframes, with empty, loading and error states.
You
Open the preview link and use the app as your users would.
preview
Booking page
Doctor schedule
Patient history
Clinic settings
14 screens ready to review

Review

You review the UI page by page. Corrections go in rounds until every page is approved.

The agents
Restates each correction as it understood it, gives an estimate, then revises and shows the diff.
You
Pin comments on the page, confirm the agent understood, and approve each page.
round 2
Booking page: "Make the free slots bigger on phones"
Understood: 48px slot buttons below 640px wide
Estimate: about 6 minutes
Done. 1 file changed. Approve or reply

Backend

The API, database and integrations, wired to the approved UI and covered by tests.

The agents
Builds the backend, connects the screens, writes end-to-end tests and pushes to your Git repository.
You
Add the keys it asked for (Paystack, SMS provider) and watch the tests pass.
test run
booking.e2e  12 passed
reminders.e2e  6 passed
payments.e2e  4 passed
Pushed 9 commits to main

Deploy

Shipped to your own server or DigitalOcean, with CI/CD so every later change deploys the same way.

The agents
Sets up the server, the pipeline, HTTPS and backups, then releases and checks the app is healthy.
You
Connect your server or cloud account once, then approve each release.
release 1.0.0
Build passed
Database backed up
Deployed to clinic.example.com
Health check: 200 OK

Operate

After launch the agents watch the app, fix bugs your testers report and propose scaling when needed.

The agents
Monitors uptime and errors, triages bug tickets, and proposes load balancers or Kubernetes when traffic grows.
You
Share the bug-report link with testers. Approve fixes and any scaling plan.
this week
Uptime 99.96%
3 tickets fixed, 1 waiting on the reporter
Proposal: add a second server before the January rush

At any stage

  • Change your mind. Send a change request. The agent reports which approved stages it affects and what it will cost before doing anything.
  • See the reasoning. The Project Brain keeps every decision, correction and open question, linked to the code that implements it.
  • Let testers report bugs. Share a bug-report link. Reports go straight to the agents, and testers get a reference to follow.
Get started