The Zero-to-Launch Stack I Use in 2024

Same client request comes up multiple times a year. Here's the stack I use for fast, defensible MVPs.

Founders ask me regularly: "if I want to ship in 6 weeks, what stack should I use?"

Here's my 2024 answer.

The default

• Next.js 15 (App Router, Server Components)
• Postgres (Neon for managed)
• Drizzle ORM (or Prisma if your team prefers)
• Clerk for auth
• Stripe for payments (if relevant)
• Vercel for hosting
• Resend for transactional email
• Sentry for errors
• PostHog for analytics

Why these

• Each one is a managed service with a generous free tier. Total infra cost for an MVP: <$50/month.
• Every component has obvious docs and a large community. Onboarding new engineers is fast.
• Boring tech. None of these will be deprecated in 18 months.
• Vercel + Next means deployment is git push. Engineers focus on code.

Variants

If you need a mobile app: Add Expo + React Native. Reuse logic between web and mobile via a shared package.

If you need real-time: Add Pusher or Liveblocks. Don't build it from scratch.

If you need AI features: Anthropic API. Maybe Pinecone if you need RAG.

If you need search: Postgres full-text first, Algolia or Typesense if Postgres isn't enough.

If you're regulated: Use AWS, terraform everything, and add a few weeks for compliance scaffolding.

What I avoid in MVPs

• Custom auth (use Clerk)
• Custom queue infra (use Postgres SKIP LOCKED or QStash)
• Self-hosted anything (until you have a person to run it)
• Microservices (one app, modular structure)
• Complex deploy pipelines (git push to Vercel)

The 6-week breakdown

• Week 1: scaffold (auth, payments, marketing site, dashboard skeleton)
• Week 2-3: core feature 1
• Week 4-5: core feature 2 + polish
• Week 6: launch prep, monitoring, on-call setup

This timeline holds for ~80% of B2B SaaS MVPs. Consumer apps are harder; AI-native products are different (depends a lot on the model integration).

When to deviate

• If your product is fundamentally novel architecture (real-time multiplayer, on-device AI), the stack matters less than the bespoke parts. Get help.
• If you're enterprise-grade from day one, the stack scales. Use the same stack; just allocate more weeks.

The stack isn't magic. Discipline is. Pick boring tech, ship fast, iterate.

References

• Next.js
• Drizzle ORM
• Clerk
• Resend