Serverless

An execution model where the platform runs your code per request, scaling and capacity management fully abstracted away.

Serverless means you deploy functions, not servers. The platform, Vercel Functions, Cloudflare Workers, AWS Lambda, hosts your code and executes it per request: scale to zero when idle, scale out automatically under load, and a bill that starts at zero too. You hand over capacity planning and pay per invocation and millisecond.

Why it matters to an indie builder: it collapses the operations budget to almost nothing. No VPS to patch, no load balancer to configure, no 3am capacity incident during a launch spike. For spiky, event-driven or small workloads, which describes most side projects and many production indie products, the free tiers alone often cover real traffic.

A concrete example: a webhook handler for Stripe events. Deployed as a function, it executes for 80 milliseconds per event. At 10,000 events a month that is 800 seconds of compute, comfortably inside free-tier limits, with cost exactly zero. The same handler on a VPS idles 24/7 for $5 monthly minimum and still needs patching. Under steady heavy traffic the math reverses, servers get cheaper at scale, which is the honest tradeoff to know.

Common mistakes: cold starts in latency-sensitive paths (a function that wakes in 400ms breaks a snappy API; keep hot paths warm or move them to edge runtimes), doing long work in one invocation (timeouts bite; split into queued small jobs), unbounded external calls inside the function (concurrency limits and provider bills punish fan-outs), and forgetting that state must live elsewhere: functions are ephemeral, so data goes in Upstash, Neon or Turso, the serverless-friendly stores grouped in the hosting blocks on IndieBoot. The nocode blocks on IndieBoot include platforms that push this model to its logical end.