bongoDev

The AWS Cloud Development Kit (CDK) lets you define AWS infrastructure in a general-purpose programming language and then deploy that infrastructure through AWS CloudFormation.

The most useful mental model is:

Text

TypeScript / Python / Java / C# / Go / JavaScript
                    |
                    v
               CDK App
          constructs + logic
                    |
              cdk synth
                    |
                    v
        CloudFormation template
            + asset manifest
                    |
              cdk deploy
                    |
                    v
             CloudFormation
                    |
                    v
              AWS resources

CDK is not an alternative AWS control plane. It is a developer-focused way to produce and manage CloudFormation deployments.

Why cloud engineers use CDK

CDK is attractive when infrastructure needs real software-engineering techniques:

reusable classes and modules
loops and conditionals
type checking and IDE autocomplete
unit tests
package management
code review
shared internal infrastructure libraries

Instead of copying 500 lines of YAML, you can model an internal platform primitive such as SecureApi or ProductionBucket once and reuse it.

CDK vs CloudFormation

Question	CloudFormation	AWS CDK
Authoring model	JSON/YAML templates	General-purpose language
Deployment engine	CloudFormation	CloudFormation
Reuse	Nested stacks/modules/macros	Constructs, classes, packages
Logic	Intrinsic functions/conditions	Language-native logic at synthesis
Output	Desired resources	Synthesizes desired resources

CDK does not eliminate CloudFormation knowledge. Strong CDK engineers understand enough CloudFormation to debug failed stack operations, logical IDs, replacements, rollback behavior and resource policies.

CDK vs Terraform

They overlap, but their operating models differ.

Text

CDK -> synthesizes CloudFormation -> CloudFormation manages resource state
Terraform -> Terraform configuration -> Terraform state maps resources

Do not pick a tool because its syntax looks nicer. Consider team language skills, AWS-only vs multi-cloud requirements, existing state, governance, deployment model and ecosystem.

Infrastructure as software does not mean application code

A common mistake is to write huge CDK programs full of hidden side effects. Infrastructure code should still be predictable.

Prefer:

TypeScript

const app = new cdk.App();
new NetworkStack(app, 'Network', { env });
new ApplicationStack(app, 'Application', { env });
app.synth();

Avoid synthesis-time code that mutates cloud resources, calls random external APIs or depends on undocumented developer-machine state.