The PaaSocalypse: When Every App Lands on a Different Cloud
The SaaSocalypse is the easy half of the story. When generating software gets this cheap, the argument goes, you stop renting the app and start making it. Why pay seat-based SaaS pricing for a tool an agent can rebuild in an afternoon? Whole categories of buy-it software get hollowed out as teams generate what they used to subscribe to.
Fine. But nobody asks the obvious follow-up: once you've generated all that software, where does it run?
That's the PaaSocalypse — and it's the bill that arrives right after the SaaSocalypse party.
The dirty secret of cheap software
The SaaSocalypse makes creating apps nearly free. It does nothing about operating them. Every app you generate still needs a database, auth, storage, a queue, secrets, a domain, TLS, backups, logs, and a place to actually execute.
So here's what actually happens. You vibe-code a thing in Lovable. The next one starts in Bolt. A teammate prototypes in v0 and ships it on Vercel. The data layer is a Supabase project — no, two Supabase projects. Marketing's microsite is on Netlify. The cron job nobody remembers is a Render service. The "real" backend is three AWS accounts with a Firebase project bolted on for the mobile app.
The apps got cheap. The sprawl got expensive.
A map of the fragmentation
Look at where agent-built software lands today. It's not one cloud — it's a gradient, from "I described it and it appeared" all the way down to raw hyperscaler primitives:
| Layer | Where your apps end up |
|---|---|
| App generators | Lovable, Bolt, Replit, v0 |
| Deploy PaaS | Vercel, Netlify, Heroku, Railway, Render, Fly.io |
| Backend / data | Supabase, Firebase |
| Edge & infra | Cloudflare, DigitalOcean |
| Hyperscalers | AWS, Azure, GCP |
Every row is a different login, a different mental model, a different lock-in shape. Pick any one and you've made a bet: that this vendor's pricing, region support, and roadmap will still suit you in two years. Spread across all of them and you've made seventeen bets — and inherited the integration tax of stitching them together.
This is the PaaSocalypse in one sentence: the cost of software collapsed, and the cost of running it fragmented.
Why it gets worse with agents, not better
You might think agents fix this — let the agent manage the platforms too. But an autonomous agent spinning up infrastructure across five providers isn't consolidation, it's the sprawl with a faster engine. More projects, more databases, more orphaned services, more blast radius, and no single place where a human can see what exists or pull the plug.
The thing that makes agent-built software safe to run isn't a smarter agent. It's a substrate underneath it: one platform where every app — hand-written or agent-generated — shares the same login, the same security model, the same databases, and the same operational floor. (That's the argument behind the k8gentic manifesto: real infrastructure under the agents, not a Lambda and a prayer.)
The way out is consolidation, not another platform
The answer to too many platforms is not an eighteenth platform with its own lock-in. It's to collapse the gradient onto infrastructure you actually control.
That's what Paddington is for. Consolidate your Lovable, Bolt, Replit, Vercel, and Supabase apps into one sovereign, portable cloud:
- One platform. Shared login, security, and database infrastructure across every app — not seventeen accounts.
- Agent-safe by default. Generated apps deploy into managed, isolated infrastructure with quotas and policy, so autonomy can't quietly exceed its blast radius.
- No lock-in. Pay only for what you need, and keep the freedom to move to your own hardware — or to another Miniscaler running the same Kilter stack.
The mechanics live in the Kilter Platform: a Portal to see and manage everything in one place, and a Runtime that wires up your databases, auth, and services and keeps them running — on battle-tested Kubernetes you never have to touch.
Surviving the PaaSocalypse
The SaaSocalypse is going to keep making software cheaper to create. Good. But every app you generate is a small, permanent operational liability until it has a home. The teams that come out ahead won't be the ones who generated the most apps — they'll be the ones whose apps all run in one place, under their own control.
Don't let the cheap-software boom turn into a fragmentation hangover.
- Read the k8gentic manifesto for why real infrastructure has to sit under the agents.
- See how the Kilter Platform consolidates the sprawl.
- Start your free trial and bring your scattered apps home.