Philosophy
I want to be explicit about where AI assistance is used, especially as generated code becomes common in public software.
Where AI is used
AI is applied to bounded tasks where it reduces mechanical work or helps investigate an unfamiliar implementation detail.
Documentation
Structuring references, descriptions, and examples for review.
Repository READMEs
Formatting introductions, feature summaries, and installation guidance.
Server UI
Implementing and revising HTML templates and responsive CSS.
Core networking
Investigating and reviewing async networking and connection handling.
Logic and calculations
Assistance with progress reporting, terminal calculations, and async data flow.
Engineering principles
I do not delegate entire projects or architectural ownership to a model. AI is used when a problem requires deeper investigation than a routine search, or when working in a language or concept outside my main expertise.
The goal is to automate lower-value mechanical work and reserve human time for defining the problem, evaluating tradeoffs, and verifying that the result is useful.
This is an estimated average from external tracking services and varies by project. The scraper packages are predominantly written by hand.
Tool distribution
Estimated distribution of AI tools used in the development workflow.
Bottom line
I understand and take responsibility for the code I publish. AI-assisted output is not pushed blindly.