Closer to the Metal Beats More Abstraction
AI collapsed build cost. That changes which infrastructure layer is worth paying for.
For twenty years, the rational response to expensive development was abstraction. Managed platforms, workflow builders, low-code layers, and hosted frameworks traded control for speed. When building was slow and expensive, that trade often made sense.
AI changes the equation. When a senior builder with AI assistance can produce infrastructure code, deployment scripts, and workflow logic quickly, the speed premium of some abstraction layers shrinks. What remains is the cost: vendor lock-in, execution markup, opinion constraints, and dependencies you do not control.
The Abstraction Tax
Every platform layer between you and the infrastructure can become:
- A dependency: the platform roadmap becomes your constraint
- A rent extraction point: you pay for convenience even after the work is understood
- An opacity layer: you debug the abstraction instead of the underlying system
- A portability trap: the deployment is not fully yours
Cloud primitives — state machines, event buses, serverless functions, AI inference — have three properties that matter now:
They’re durable. State machines deployed years ago still run today. A fragile workflow builder can break when its abstraction changes.
They’re composable. Each primitive does one thing. You compose them declaratively. A state machine orchestrates. A function executes. An event bus routes. The composition is visible in files that AI agents read, modify, and deploy.
They map to reality. A state machine is the business process, not an approximation of it. Application code with background jobs is a developer’s model of a state machine, with all the drift and maintenance that modeling implies.
What Changed
Three capabilities converged: infrastructure-as-code made cloud resources declarative and reproducible. AI coding assistants absorbed much of the complexity that justified platforms. API-first cloud services became composable primitives, not just products you integrate with.
The old bottleneck was “cloud infrastructure is hard.” It is still real engineering work, but it is less of a reason to accept weak ownership by default.
Implication
The right response to AI making development faster is not always more abstraction. Often it is less. Get closer to the metal. Compose primitives. Declare infrastructure.
A workflow automation system built on cloud primitives can run in the customer’s account, use cloud-provider durability guarantees, and remain maintainable by a competent engineer with AI support.
Build the factory on primitives you can own.
Contrarian To
“Use another abstraction layer to move fast.”
Sometimes those tools are the right choice. But do not confuse speed with ownership. In many cases you can move quickly with primitives now, and what you build is more durable, portable, and yours.