🚀 Ship Your Product

MVP isn't a bad product. It's the smallest version that solves someone's real problem. The classic trap: polishing features nobody asked for while the core problem remains unsolved. Claude Code speeds up development, but you need to know what to build.

Ask yourself: "If I had only 1 screen and 1 action, what would it be?" That answer is the core of your MVP. Everything else is backlog. Tell Claude Code: "cria o fluxo minimo pra [problema X]" and resist the temptation to add “just one more field.”

Every feature you add has a cost: code to maintain, bugs to fix, complexity to explain. Ruthless prioritization says NO to 90% of ideas so the remaining 10% can be excellent. Use the impact vs. effort matrix: high impact + low effort = do it now. The rest can wait.

## Matriz de Priorizacao

| Feature         | Impacto | Esforco | Decisao      |
|-----------------|---------|---------|--------------|
| Auth basico     | Alto    | Baixo   | MVP - faz    |
| Dashboard admin | Medio   | Alto    | v2 - espera  |
| Dark mode       | Baixo   | Baixo   | Nice-to-have |
| Notificacoes    | Alto    | Medio   | v1.1 - logo  |
| Multi-idioma    | Baixo   | Alto    | Backlog      |

Before launching, go through this checklist. It's not about perfection. It's about not breaking things in front of the user. Claude Code helps: "audita esse projeto contra a launch checklist".

## Pre-Launch Checklist

- [ ] README claro (o que faz, como instalar, como usar)
- [ ] .env.example com todas as variaveis documentadas
- [ ] Testes passando (pytest + CI verde)
- [ ] /health endpoint respondendo
- [ ] HTTPS configurado
- [ ] Secrets fora do repositorio (.gitignore)
- [ ] Backup do banco de dados
- [ ] Monitoring + alertas basicos
- [ ] Pagina de erro customizada (404, 500)
- [ ] Um usuario real testou antes de voce

Every item has a reason. A confusing README = no one adopts it. Exposed secrets = disaster. A real user testing it = bugs you would never see otherwise. The checklist isn't bureaucracy: it's the minimum safety net.

Launched? Now listen. The feedback loop is the product’s engine for improvement. Quantitative metrics (who used it, where they dropped off) plus qualitative feedback (what they said, what they asked for). Without a feedback loop, you’re coding in the dark.

Minimum tools: basic analytics (Plausible, Umami), a direct channel with users (form, email, Discord/Telegram), and a place to record what you hear (Notion, GitHub Issues, even an .md file in the repo). The secret: respond quickly. A bug reported and fixed the same day is worth more than 10 new features.

The product isn't "ready" at launch. It's just getting started. Post-launch iteration follows a rhythm: fix critical bugs (day 1-3), adjust UX based on feedback (week 1-2), add the most requested feature (week 3-4). Frequent micro-releases > occasional mega-updates.

With Claude Code, the iteration cycle gets dramatically shorter. Report a bug -> "corrige esse bug: [descricao]" -> tests -> deploy. What used to take days now takes hours. Use this to your advantage: response speed is a real competitive advantage for small projects.

## Ciclo Pos-Launch

Dia 1-3:   Hotfixes (bugs criticos, crashes)
Semana 1:  Quick wins do feedback (UX, copy, fluxo)
Semana 2:  Refactor se necessario (divida tecnica)
Semana 3:  Feature #1 mais pedida
Semana 4:  Retrospectiva + planejar proximo ciclo

Regra: nunca fique mais de 2 semanas sem release.

You completed all 5 tracks. You know how to install and configure Claude Code (T1), use Plan and Loop modes (T2), connect MCPs (T3), orchestrate subagents and Git (T4), and build and deploy with Python (T5). This isn't the end. It's the beginning of a new way of working.

The next step is yours. It could be a SaaS, an internal tool, an automation that saves hours every week, or an open-source project that helps others. Claude Code is the copilot. You're the pilot. Ship it.