Final merge, full SaaS demo & week 2 victory
The finale. PL-7 adds multi-user support and polish, Claude reaches for the freemium pattern on its own, we give the product an honest critique — then declare victory on a real, end-to-end SaaS build.
One ticket to go. PL-7 adds real multi-user support and the final coat of polish — and turns our prototype into something that genuinely looks like a product you could sell. Then we demo the whole thing, give it an honest critique, and declare victory on week two.
01PL-7: multiple users & final polish
The kickoff is now second nature: implement Jira ticket PL-7, then test and submit a PR. It
asks its usual good questions, then delivers a huge change set — around 1,500 lines across 22 files
— adding sign-up/sign-in, saved documents, and polish across every screen. The big question, as always: will it
work?
02The freemium surprise
Start the server, open it, and something clever appears: you can use the product without signing in, and sign in only when you want to save. Nobody asked for that — Claude chose it.
Let people experience the value first, then create an account to keep their work. It is the classic freemium / product-led growth move, and Claude reached for it on its own. Sign up, generate a document, see it under My documents, sign out, sign back in — and there it is, persisted. User management and saved documents, working first time, with a polished, professional look and the right brand colours.
03An honest critique
It is genuinely impressive — and still not perfect, which is the honest lesson.
Two blemishes: the chat gives a slightly empty message right at the end, and — more importantly — the legal disclaimer we specified in PL-7 is missing. Even with its own quality review, it dropped a requirement. That is exactly why there is still a role for us: to come back, check against the ticket, and catch what the model missed. Trust the productivity, verify the details.
04Merge, and declare victory
Final merge: “merge the PR locally, push, switch to main,” then one last
CLAUDE.md update. Step back and look at what we did across two days:
We read work from Jira (MCP), ran it through FeatureDev's disciplined process, called an LLM through our own Cerebras skill with structured outputs, built a containerised FastAPI + Next.js + SQLite app with user management, and shipped every step as a reviewed GitHub PR. A real SaaS product — not far at all from something you could commercialise.
What would make it truly production-ready? A persistent database behind it (Supabase or similar rather than a resettable SQLite), and deployment to production — both very doable, and on the roadmap.
05Your assignment & what's next
Now it's your turn. Build your own SaaS product with the same workflow — the general idea (templates + an AI chat that fills them + saved documents behind a login) applies to countless businesses. Do it the way we did:
Write Jira (or GitHub) ticketsplan
Break your product into a few clear tickets in a stakeholder's voice.
Run FeatureDev, ticket by ticketbuild
Stay at the helm — answer the questions, review the architecture, test each result, merge.
Share it on LinkedIncelebrate
Post what you built as your week-two project (tag your instructor!). Amplifying each other's wins is how the whole community grows.
Take a moment: you are now 67% through the program and can drive Claude Code like a professional. If week two was about getting professional, week three goes big — specialised agents, hooks, bigger codebases, and eventually swarms and orchestrators. Massive congratulations. See you in week three.
✓ Key takeaways
- PL-7 adds real sign-up/sign-in, saved documents, and polish — ~1,500 lines, first-time working.
- Claude reached for the freemium / product-led-growth pattern unprompted — explore first, sign in to save.
- Still verify: it dropped the specified legal disclaimer even after its own review — the human's job is to check against the ticket.
- We built a full SaaS end to end: Jira → FeatureDev → Cerebras skill + structured output → FastAPI/Next.js/SQLite → GitHub PR.
- Next steps to production: a persistent database and deployment. Your assignment: build your own, share it — then on to week three's agents, hooks and swarms.