21 hours ago
- Robin Sloan's 2020 concept of 'home-cooked software'—an app built for a small, loved audience, like a family messaging app with only four downloads—is now more feasible with AI.
- In 2026, AI tools make personal computing truly personal, enabling rapid creation of custom apps (e.g., baby tracker, sleep app, fitness app) without unnecessary features.
- The author built several personalized apps: a sleep app for a consultant's plan, a fitness app that adjusts smoothie portions to daily runs, a marathon plan from personal races, a jazz quiz app, and a medical records tool.
- The sleep app (shared by family and nanny) and fitness app (integrating multiple data sources) demonstrate how AI can combine data from separate platforms for hyper-personalized recommendations.
- The author's tech stack: Next.js, Tailwind, shadcn/ui, Better Auth, Postgres/Neon with Drizzle, Resend, GitHub Actions, Claude Code, Vercel AI SDK, Warp, and Umami; costs about $160/month (mostly Claude subscription), with free tiers available for simpler apps.
- Key learnings: building costs have dropped dramatically (weekend builds), maintenance is easy via AI, ephemeral apps are fine, AI unlocks small niche markets, good APIs are crucial, and the building process itself is rewarding and addictive.
- The future points to non-technical users building personal software via plain English prompts (e.g., Sam Altman's example), making truly personalized apps the new baseline expectation.