Build July 28, 2026

The Solo Developer Stack: One Person, Four Product Lines

The barrier to running multiple products alone was never skill — it was the cost of context-switching. AI as a teammate changes the math entirely.

← All articles

The old math

A solo developer used to face a brutal tradeoff: go deep on one product, or spread thin across many. The constraint wasn’t ideas, capital, or even skill. It was attention. Every product needs design decisions, infrastructure, content, SEO, customer support, and bug fixes. Switching contexts has a real cost — studies put it at 23 minutes of recovery per switch.

Running four products meant spending 75% of your time just switching gears.

The new math

AI as a teammate doesn’t remove all of that friction, but it changes the economics enough to matter:

  • Code generation handles boilerplate that used to eat hours. Your job shifts from writing to reviewing.
  • Content drafting takes a rough outline to a passable draft in minutes. Your job shifts from blank-page paralysis to editing.
  • Research that used to mean 20 browser tabs now takes one conversation.
  • Architecture decisions can be sanity-checked against a second opinion that happens to have read every framework’s documentation.

The key insight: you’re not outsourcing judgment. You’re outsourcing the mechanical work that surrounds judgment.

What the stack looks like

Here’s the actual setup running four product lines in parallel:

Infrastructure: Cloudflare for everything — Pages, D1 (SQLite), Workers, KV, R2. One platform, one deploy pipeline, one bill. No microservices sprawl.

Framework: Astro for content-heavy sites (SEO, speed, static-first). Next.js only where there’s heavy interactivity (dashboards, real-time data).

LLM: One model provider for everything — product features, content, code review, research. DeepSeek for cost efficiency, with a fallback when it’s overloaded.

Analytics: Umami (self-hosted, privacy-friendly) as source of truth. Google Analytics as secondary signal.

Payment: Creem as merchant of record. Handles tax compliance globally. One integration, multiple products.

The parts you can’t automate

What AI doesn’t solve:

  1. Taste — knowing which of five generated options is actually good
  2. Domain expertise — recognizing when the AI is confidently wrong about your niche
  3. Positioning — deciding what the product isn’t
  4. Customer empathy — understanding why a real person would care

These are the parts that compound. Everything else is becoming a commodity.

The real constraint

It’s not time anymore. It’s decision fatigue. Running multiple products means making hundreds of micro-decisions daily. The bottleneck isn’t execution speed — it’s how many good decisions you can make before your judgment degrades.

The fix isn’t more AI. It’s fewer, more deliberate choices. Ruthless prioritization. Saying no to features that don’t move the needle. The solo developer’s real skill isn’t building fast — it’s knowing what not to build.

This is a field note, not a playbook. Your context is different. Steal what’s useful.