← Back to Blog
Stack

Choosing Your Stack in 2025: What Matters and What Doesn’t

Tony Le·Feb 28, 2025·7 min read

Every founder we talk to asks the same question: "What stack should I use?" And every time, we give the same answer: it depends. But not in the wishy-washy consultant way. It depends on specific, measurable factors that most teams never think about.

After shipping seven platforms on different stacks, here's our framework for making stack decisions that you won't regret six months later.

The Only Three Questions That Matter

Forget the framework comparison charts. Forget the benchmarks. When choosing a stack for a new project, we ask three questions:

  • How fast can we ship the first version? Time-to-production is the single most important metric for a new product. Every week of additional development time is a week of learning you're missing.
  • Can this scale to the next order of magnitude? You don't need to handle 10M users on day one. But you need to know the path from 1K to 10K to 100K users doesn't require a rewrite.
  • Can we hire for this in 6 months? If the product works and you need to grow the team, the talent pool matters. Exotic stacks have real costs beyond the technology.

Our Default Stack (and When We Deviate)

We have a default stack that optimizes for speed-to-production while maintaining production quality. Here's what it is and why:

Frontend: Next.js + React + TypeScript + Tailwind CSS. This is the stack with the largest ecosystem, best DX, and fastest path to production. Server components eliminate entire categories of loading-state bugs. TypeScript catches errors before they ship. We've yet to find a project where this doesn't work.
Backend: Supabase (Postgres + Auth + Real-time + Storage). This eliminates 2-3 weeks of backend scaffolding on every project. Built-in auth, row-level security, real-time subscriptions, and file storage. It's Postgres underneath, so you're never locked in.
AI Layer: Claude API or OpenAI, depending on the task. LangChain for agent orchestration when the workflow is complex. Direct API calls when it's simple. Never add abstraction layers you don't need.
Deployment: Vercel for Next.js-heavy projects, Render for anything else. Push-to-deploy with zero DevOps overhead.

When We Deviate From the Default

The default stack covers 80% of projects. For the other 20%, here's when and why we switch:

  • High-frequency real-time data (like GammaLens): We add Redis for caching and WebSocket layers for streaming. The core stack stays the same, but the data pipeline gets specialized infrastructure.
  • Heavy computation (like BetEdge's arb scanning): Background workers in Python or Go for the compute-intensive scanning, with the web layer still in Next.js. Use the right language for the right job.
  • Mobile-first products: React Native with Expo for cross-platform mobile. Shares the React mental model and a significant portion of business logic with the web codebase.
  • Enterprise integrations: When the project needs SAP, Salesforce, or legacy API integration, we go with Node.js/Express for the middleware layer. More control over authentication flows and data transformation.

What Doesn't Matter Anymore

The framework wars are genuinely over. Here's what we see teams waste weeks debating that simply doesn't matter for most projects:

  • React vs Vue vs Svelte: They all work. Pick the one your team knows. If you're starting fresh, React has the biggest ecosystem. Done.
  • SQL vs NoSQL: Use Postgres. It handles relational data, JSON documents, full-text search, and vector embeddings. You almost never need a separate NoSQL database.
  • Monolith vs microservices: Start with a monolith. Always. Break it apart when you have the data to know where the boundaries should be, not before.
  • CSS framework choice: Tailwind, CSS Modules, styled-components — they all produce the same CSS in the end. Pick one, commit, move on.

The Real Competitive Advantage

Here's the uncomfortable truth: your stack is not your competitive advantage. Your ability to ship is. The teams that win are the ones that pick a proven stack, commit to it, and pour all their energy into building the actual product instead of endlessly optimizing their toolchain.

We've shipped platforms in Next.js, React, Vue, and Python. The technology mattered less than the methodology every single time. A great process with a good stack will always beat a perfect stack with a slow process.

The best stack is the one that gets your product in front of users fastest. Everything else is a distraction.
StackArchitectureNext.jsSupabase

Building something?

Let's talk about what hyper speed looks like for your project.

Book a Chat

More from the blog

Build Log
Hyper Speed Agile: How We Ship Platforms in Days
The methodology, the AI tools, and the stack decisions that make it possible to go from idea to production in weeks, not months.
AI
AI Agents in Production: What Actually Works
Lessons from deploying AI agents across seven different platforms and use cases. What we learned about reliability, cost, and user experience.