My Solo Founder Tech Stack in 2026 — What I Use to Run fromscratch.dev and Why

Every solo founder eventually faces the same question: what should I build on? The answer matters more than most people think — not because of performance benchmarks, but because the wrong stack will slow you down, drain your budget, or force you to learn ops instead of shipping features.

This isn't a "top 10 tools" listicle. It's the exact stack I use every day, with honest takes on what works, what's annoying, and what I'd change if I started over.

If you're still figuring out what to ship before you optimize your stack, start with the SaaS launch checklist.

The full stack at a glance

Total monthly cost: $35–55 depending on usage. Let me walk through each choice.

Frontend: Next.js 15 + React 19 + Tailwind CSS 4

This is the boring-but-correct choice for solo founders in 2026.

Next.js gives me server components, API routes, static generation, and incremental static regeneration in a single framework. I don't need a separate backend. I don't need to configure a build pipeline. I deploy one project and everything works.

Why not Remix or SvelteKit? Both are good frameworks. I chose Next.js because the ecosystem is the largest — which means better AI coding tool support (Cursor and Claude Code understand Next.js patterns deeply), more community answers to debug, and easier to find contract help if I ever need it.

Tailwind CSS 4 is the other non-negotiable. v4 dropped the config file requirement and added native CSS nesting. The DX improvement over v3 is significant. I pair it with tailwind-merge and clsx for conditional classes.

The real reason I'd never switch: every AI coding tool in 2026 writes excellent Next.js + Tailwind code. When your AI assistant can scaffold a complete page in 30 seconds, framework popularity becomes a multiplier on your productivity.

Database: Neon Postgres

This was the highest-stakes decision in the stack. Databases are hard to migrate away from.

Neon gives me serverless Postgres with autoscaling, branching (like git branches for your database), and a generous free tier. The connection uses their serverless driver (@neondatabase/serverless), which means I pay nothing when nobody's using the site and scale automatically during traffic spikes.

Why not Supabase? Supabase bundles auth, storage, realtime, and a database into one platform. That's great for prototyping but creates tight coupling. I wanted to pick the best tool for each job separately. Neon is a better pure database than Supabase's Postgres offering — the branching feature alone saves me hours when testing schema migrations.

Why not PlanetScale? They killed their free tier and pivoted away from individual developers. That was enough to rule them out.

Why Drizzle ORM over Prisma? Drizzle generates SQL that I can read and understand. Prisma's query engine adds a runtime layer that abstracts too much. When something breaks at 2 AM, I want to see the actual SQL, not debug a query engine. Drizzle also has better edge runtime support, which matters for serverless deployments.

Auth: Better Auth

Authentication is the most annoying part of any SaaS to implement. It's also the part where mistakes have the worst consequences.

I use Better Auth — a self-hosted TypeScript auth library that handles sessions, OAuth providers, and account management. My setup supports GitHub and Twitter login, with sessions stored in my Neon database via the Drizzle adapter.

Why self-hosted auth instead of Clerk or Auth0? Cost and control. Better Auth is free forever because it runs in my own infrastructure. Clerk charges per monthly active user, which means auth costs scale linearly with growth. For a solo founder pre-revenue, that's a scary cost curve.

The honest trade-off: Clerk's pre-built UI components are genuinely excellent. Better Auth means I built my own login pages and session management UI. That took a couple days of extra work. If I were optimizing purely for speed-to-launch, I'd start with Clerk and migrate later if costs became an issue.

Hosting: Netlify

I deploy to Netlify with the @netlify/plugin-nextjs adapter. Every git push triggers an automatic build and deploy.

Why Netlify over Vercel? Honestly, it was close. Both have great free tiers and excellent Next.js support. I went with Netlify because their pricing model is more predictable — Vercel's bandwidth and serverless function pricing can spike unexpectedly on viral traffic. Netlify's model is simpler to reason about.

The Vercel advantage I miss: Vercel's image optimization and ISR implementation are slightly more polished, which makes sense since they build Next.js. If Vercel's pricing becomes more predictable, I might switch.

What about Railway? Railway is excellent for backend services and Docker containers, but it's overkill for a Next.js app that Netlify handles natively. I'd use Railway if I had a separate API server or background job runner.

Analytics: Fathom + Google Analytics 4

I run a dual analytics setup, which sounds overkill but serves two different purposes.

Fathom Analytics (~$14/month) is my primary dashboard. It's privacy-first, doesn't require cookie banners, and gives me the numbers I actually check daily: page views, top pages, referrers. The UI is fast and simple.

Google Analytics 4 runs alongside for one reason: Search Console integration and SEO data. Fathom doesn't tell me which search queries bring traffic or how my pages rank. GA4 does. For a content-heavy site like fromscratch, that data is essential.

I also track custom events (clicks, content interactions, searches) in my own database. This gives me full ownership of behavioral data without depending on any third-party analytics provider.

AI coding: Cursor + Claude Code

This is the part of my stack that changed the most in the past year. AI coding tools went from "nice autocomplete" to "genuine productivity multiplier."

Cursor (~$20/month) is my primary editor. I use it for:

• Multi-file refactors where I need to see the changes visually
• Component development where I'm iterating on UI
• Quick edits where Tab completion saves keystrokes

Claude Code (~$20/month) is my CLI companion. I use it for:

• Complex multi-step tasks ("add i18n support to all glossary pages")
• Debugging where I need the AI to read many files and reason about the system
• Autonomous changes that span 10+ files

The combo workflow: I plan in Claude Code, execute in Cursor for small changes, and hand complex migrations back to Claude Code. They complement each other well because Cursor is visual and fast while Claude Code is thorough and autonomous.

Content system: MDX + gray-matter + Zod

fromscratch is a content-heavy site with 100+ pages of comparisons, reviews, glossary terms, and blog posts. All content lives as MDX files in the repository, validated by Zod schemas at build time.

Why MDX in the repo instead of a CMS? Three reasons:

1. Version control — every content change is a git commit with full history
2. Type safety — Zod schemas catch broken frontmatter before it reaches production
3. Speed — no API calls to fetch content, everything is local at build time

I use gray-matter to parse frontmatter and next-mdx-remote to render MDX with custom React components. The content system supports dual-locale (English and French), automatic reading time calculation, and structured data extraction.

The trade-off: non-technical collaborators can't edit content without git knowledge. For a solo founder, that's fine. If I hire a content writer later, I'd add a simple CMS layer (probably Keystatic or Tina) that commits to the same git repo.

What I spend per month

Here's the real breakdown. No affiliate links, no sponsored numbers.

The $0 items aren't bait — Neon's free tier gives you 0.5 GB storage and autosuspend compute, which handles thousands of users. Netlify's free tier includes 100 GB bandwidth and 300 build minutes. Both are genuinely sufficient for an early-stage SaaS.

What I'd change if I started over

I'd start with Clerk instead of Better Auth. The time I spent building auth UI was real. Clerk's pre-built components would have saved me 2-3 days, and I could migrate to self-hosted auth later when cost optimization matters.

I'd skip the dual-locale system initially. Building French translations before having English SEO traction was premature. Lesson learned: nail one language first, then expand.

I'd add Fathom from day one. I wasted the first month without analytics, which means I have no baseline data for my earliest traffic. $14/month is worth it from launch day.

I wouldn't change the core stack. Next.js + Neon + Drizzle + Netlify is the right foundation. It's boring, it works, and it gets out of my way so I can focus on content and features instead of infrastructure.

verdict

The best solo founder stack in 2026 isn't the most technically impressive one — it's the one that lets you ship every day without thinking about infrastructure. Next.js + serverless Postgres + managed hosting gets you there for under $50/month. Pick tools with generous free tiers, strong TypeScript support, and large communities. Save the architectural astronautics for when you have revenue to justify the complexity.

Where each tool fits in the bigger picture

If you're evaluating tools for your own stack, I've written detailed comparisons and category guides for most of the tools mentioned here:

• Databases: Neon vs Supabase · Supabase vs Firebase · Best database tools
• Hosting: Vercel vs Netlify · Railway vs Vercel · Best deployment platforms
• Auth: Clerk vs Auth0 · Best authentication providers
• Payments: Stripe vs Lemon Squeezy · Stripe vs Paddle · Best payment processors
• AI Coding: Cursor vs Claude Code · Cursor vs GitHub Copilot · Best AI coding tools
• Analytics: Best analytics tools

How many customers do you need?

Plug in your price point and income goal to see exactly how many paying customers your SaaS needs.

Try the calculator