# Boilerplates Are Dead. Starter Kits Are Dying. What's Next.

> Boilerplates sold starting lines; starter kits sold training wheels. Solo founders need neither — they need a shipping kit: parity, guardrails, flat bills. The obituary plus the replacement.

Source: https://boringpush.com/blog/boilerplate-starter-kit-dead

## Cause of death #1: the boilerplate (a starting line in a world of free generation)

A boilerplate's entire value proposition was "skip week one." Auth wired, Stripe connected, folders arranged — you paid to not type the boring parts.

Then AI learned to type. What cost a week now costs a prompt. The starting line is free, so the product built on selling it is worth zero. Worse than zero, actually: boilerplates were written for humans typing every line by hand. Point an agent at one and it hallucinates packages, breaks schemas, and introduces memory leaks — the codebase has no fences, because fences were never the product. What's left is a pile of code that generates fast and rots faster.

Boilerplates aren't evolving past this. Their architecture *is* the absence of architecture: files arranged for reading, not for constraining. You cannot fence that retroactively without rebuilding it — at which point you didn't buy a boilerplate, you built a kit.

## Cause of death #2: the starter kit (training wheels, no engine)

Starter kits saw the problem and added tutorials, docs, and onboarding checklists. Better — but training wheels, not an engine. They still assume a human at the keyboard making every decision the docs describe. The moment you hand the repo to an agent for a weekend, the same rot sets in: no machine-checkable rules, no parity contracts, no gates. A checklist the AI never reads is decoration.

The starter kit dies the day you stop typing. Which, if you're honest about 2026, already happened.

## What survives: the agent-first shipping kit

The replacement isn't a better starting line or better training wheels. It's a different species, defined by three properties neither ancestor has:

**1. True database parity.** Write the schema once, write the queries once, flip one variable: SQLite locally, Supabase (or MongoDB) in production. No dual maintenance, no rewrite quarter. Boilerplates weld you to one vendor; starter kits document the migration you'll dread. Parity deletes the category.

**2. Guardrails, not guidelines.** Strict files the agent *must* obey (conventions), gates that *fail the deploy* when disobeyed (lint, parity checks, thin-route limits), least-privilege database roles so even a disobedient agent can't destroy data. If the protection isn't machine-enforced, it doesn't exist. This is the single biggest difference between a kit and everything that came before it.

**3. Flat economics.** $199 once, $5 VPS, no seats, no meters, no "contact us" pricing page. Boilerplate-era pricing smuggled subscriptions into starters (per-seat "teams" tiers for solo founders). A shipping kit charges once because there's nothing to meter — you own the code, the database, and the server.

Plus the boring parts, finished: auth, billing across three rails, marketplace, analytics without cookie banners, one-command deploy and rollback, Docker under 300 MB.

## The top-down verdict

From the top: boilerplates are dead (generation made starting lines free), starter kits are dying (humans stopped typing), and the thing that replaces both is the agent-first shipping kit — outcome insurance, not code volume. Every comparison on this blog scores individual rivals bottom-up; this is the frame they all sit inside. BoringPush is built as that replacement: one switch between SQLite and Supabase, same code, full marketplace, guardrails so AI doesn't break what's already paying you.

An agent-first shipping kit with true SQLite ↔ Supabase parity. One switch. Same code. Full marketplace. Built so AI doesn't break what's already paying you.