Human Augmented Agentic Engineering

Today I discovered sometimes you can just edit small portions of the code directly by hand rather than prompting an agent to gain efficiency when only making small changes. I'm calling this human augmented agentic engineering.

Human augmented agentic engineering is an advanced technique emerging in the field being driven forward by engineers who value speed over token spend, direct local output token creation without the need for expensive input tokens, and deterministic iterations.

Here's how it works. Let's say you're working on a large codebase and you want to change to change a padding a value of a parent vertical stack container that contains a label with the value "hello world".

The old way:

• Open Claude
• Prompt Claude with something like "Add 2pt of padding to the parent of the label that contains the value 'hello world' on @FooView.swift"
• Wait for Claude to discombogulate, gyrate, ruminate, fry onions, etc.
• Review the change
• Remedial prompt
• Repeat until the padding is in the right place

The new way:

• Open the file you want to change directly.
• Search for the text you want to change :/hello world
• Make your changes

Try HAAE in your workflow today and stayed tuned for my new javascript framework with non-optional platform as a service built around this new paradigm.