Nicholas Constantino, an experienced founder who co-founded Sticker Mule in 2010, came to Evil Martians with an MVP feature list and core user flows already mapped out for a live classes platform. He needed a technical team that could flag risks, support go-to-market needs, and act as a challenger adviser, not just an executor.
Sprint-based collaboration
Evil Martians run this engagement in focused, intensive sprints with a team of two: design engineer and backend engineer. Nicholas works in parallel on recruiting teachers, building marketing channels, and growing social media presence. This rhythm keeps costs lean and lets both sides move independently.
Agentic coding on Rails
The stack is Inertia + Rails + React in a single repo with shadcn components and Storybook as the design environment. Before writing the first prompt, Evil Martians configured AI harnesses for the project: layered Rails skills, Inertia skills, and CLAUDE.md rules encoding the architecture and conventions.
The design engineer builds features directly in production using agentic coding. He also runs weekly demos with the founder, makes product scope decisions, conducts teacher and student interviews, sets up analytics, and drives GTM strategy. The backend engineer handles production tasks like Stripe Connect, Whereby integration, and admin review flows, keeping the codebase secure, clean, and scalable.
From zero to production in four weeks
Evil Martians shipped a fully functional homepage in two weeks: course examples, an about section, and a working signup flow. By week four, the team had delivered auth, payments, live video classes, a course builder, teacher and student flows, a teacher application portal, a blog, and an admin panel.
Validating demand with cold traffic
With the platform live, Evil Martians built a demand validation pipeline on a $2K ad budget. The analytics stack was sequenced by need: Plausible first for immediate tracking, PostHog for session replays and A/B testing, GA4 for Google Ads optimization. Conversion tracking wired directly to the waitlist. PostHog ran experiments at 95% confidence. Four feedback channels went up: price tags on features for willingness-to-pay signal, a topic suggestion banner, a scroll-triggered survey, and an external survey testing demand without showing the product.
Results: 36K impressions, 128 cold-traffic waitlist joins, CPA below target. The founder doubled his ad budget after the first 12 signups and kept scaling. He now operates warm-channel testing independently.
Running in production
We are now approaching the next milestone: opening the platform for real courses, running the first sessions with real users, and stress-testing the full flow before scaling up course acquisition and advertising. The collaboration is still ongoing.
