Don’t start from trends
There is no universal “best stack.” Ask three questions:
- Who maintains this in 2 years? (internal skills, local hiring pool)
- What constraints are non-negotiable? (performance, compliance, legacy integrations)
- What’s the cost of change? (migration, training, tooling)
Simple matrix
| Context | Common direction |
|---|---|
| Greenfield B2B SaaS | TypeScript + API (Node or Java) |
| Enterprise / banking | Java, .NET, sometimes COBOL adjacent |
| Data / ML | Python + orchestration |
| Marketing site | Next.js, Astro, headless CMS |
Classic mistakes
- Picking Rust or Kubernetes “to learn” on a 3-month delivery project.
- Forcing one stack org-wide without looking at existing teams.
- Ignoring observability and CI from day one.
→ Selling a niche stack? See find freelance dev work.