Programmer.ie
Technical library for the AI era
Menu
Home
Books
Solutions
Learn
Prompts
Articles
Search
About
Codeai
The Smallest Useful Model Call
One Operation, Several Model APIs
Normalize at the Boundary
The Model Is Not the Process
Context Is an Input, Not a Transcript
Externalize Working Memory
Raw Output First
Claims, Evidence, and Decisions
Let the Machine Touch Something
Capability Is Not Authority
The Agent Cannot Grade Its Own Homework
Retries Are Side Effects Too
Independent Calls
The Models Were Different. Their Mistakes Weren't
Make the Problems Harder
Diversity Without More Models
Replicate Before You Believe
What Should Happen Next?
Applied AI