Infrastructure as Code

Featured analysis

The most pleasant way to write AWS infrastructure: constructs give you reusable, higher-level abstractions in TypeScript or Python, and the developer experience is a genuine leap over hand-writing CloudFormation. The catch is that it compiles down to CloudFormation, so you inherit that engine's deployment speed and drift behavior underneath. It is also firmly AWS-only, which is the point and the ceiling.

Featured analysis

A large ergonomic upgrade over raw ARM JSON: cleaner syntax, modules, and type checking, all first-party and free, which makes it the natural choice for Azure-native teams. It removes most of the reasons people dreaded ARM. But it is a thin DSL over ARM rather than a multi-cloud tool, so its horizon ends at the Azure boundary — outside that estate it is simply not in the conversation.

Featured analysis

The pragmatic answer to Terraform-at-scale sprawl: it keeps configurations DRY, wires up remote state, and orchestrates dependencies across many accounts and environments, which large estates genuinely need. Where teams have dozens of stacks, it earns its keep. The cost is conceptual — it is a wrapper with its own patterns layered on top, adding indirection that newcomers have to learn before it pays off.

Featured analysis

The ubiquitous choice for configuring systems and deploying applications: agentless, readable, and with a module for practically everything, which is why it lives in nearly every ops toolbox. It overlaps with but does not replace state-managed provisioners — it is procedural and push-based, so teams frequently pair it with Terraform rather than choosing between them. For the last mile of server and app configuration, nothing is more entrenched.

Featured analysis

Born from the BSL backlash, it is the governance insurance the Terraform-compatible world needed: fully open, community-steered, and compatible enough to inherit the entire provider ecosystem. It is no longer just a mirror — features like state encryption and provider-defined functions arrived here first. For teams that like Terraform but want an open long-term bet, it is now arguably the more prudent default, with the caveat that it will slowly diverge from upstream.