Medium

George Taskos

George Taskos

·

Apr 27

The Full Stack Loop: Running Claude on IaC and Application Code Without Drift

CDK Series — Part 6 of 6 | Five articles built the pieces. This one runs them together — and shows where the session boundaries, ownership…

The Full Stack Loop: Running Claude on IaC and Application Code Without Drift

George Taskos

·

Apr 26

The GitHub Actions Pipeline: Trunk-Based Dev, Release Tags, and Ordered CDK Deploys

CDK Series — Part 5 of 6 | The dependency map you built in Article 1 becomes three words in YAML: needs: [deploy-data]. Here’s the full…

The GitHub Actions Pipeline: Trunk-Based Dev, Release Tags, and Ordered CDK Deploys

George Taskos

·

Apr 25

Aurora PostgreSQL 16: The Database Claude Shouldn’t Touch Alone

CDK Series — Part 4 of 6 | The mistakes Claude makes on Aurora aren’t recoverable without a restore. Here’s where the human engineer must…

Aurora PostgreSQL 16: The Database Claude Shouldn’t Touch Alone

George Taskos

·

Apr 24

ECS/Fargate with Claude: Five Defaults That Will Fail You in Production

CDK Series — Part 3 of 6 | Claude writes ECS services fast. Here’s what it gets wrong every time — and the patterns that fix it.

ECS/Fargate with Claude: Five Defaults That Will Fail You in Production

George Taskos

·

Apr 22

Private by Default: The Network Layer Claude Gets Wrong Every Time

CDK Series — Part 2 of 6 | Six VPC defaults that look fine in dev, fail in production — and what to encode in CLAUDE.md so they don’t

Private by Default: The Network Layer Claude Gets Wrong Every Time

George Taskos

·

Apr 21

The CDK Mental Model: Teaching Claude About Your Stack

CDK Series — Part 1 of 6 | How to give Claude Code enough context to write CDK that matches your architecture — not generic boilerplate

The CDK Mental Model: Teaching Claude About Your Stack

George Taskos

·

Apr 13

You Made Claude Faster. Then You Made Yourself the Bottleneck.

The PR review process was designed for human-written code. It’s breaking under AI-generated volume — and the fix isn’t more reviewers.

You Made Claude Faster. Then You Made Yourself the Bottleneck.

George Taskos

·

Apr 13

MCP vs CLI: Why Both Camps Are Half Right

The debate isn’t about which tool is better. It’s about which problem you’re actually solving.

MCP vs CLI: Why Both Camps Are Half Right

George Taskos

·

Apr 12

Verify, Don’t Review

Treating AI output like peer review is the wrong mental model. The right model is compiler output — you write tests that verify behavior…

Verify, Don’t Review

George Taskos

·

Apr 11

The Engineer Who Writes No Code

In the agent era, engineering leverage is inversely proportional to code output — the best engineers are the ones who’ve made their own…

The Engineer Who Writes No Code

George Taskos

George Taskos

20 years building products. 3 exits. Now focused on AI systems that automate complex workflows in regulated domains—social media, and beyond.

Following

Help

Status

About

Careers

Press

Blog

Store

Privacy

Rules

Terms

Text to speech