Uncategorized

Five work areas – client science writing, two blogs, the channels, a household – and one AI system underneath. The map, the pattern, and the honest math.

One real day of client work with AI agents: a citation audit across live pages, a second agent auditing the first, and a dashboard of parallel sessions.

ChatGPT is useful and confidently wrong. A science writer on AI citations that don’t hold, the goading trap, and how to check before you ship.

A Claude Code subagent is Claude with a fresh context window and no memory of your chat – the constraint is the feature. When to use one, from zero setup.

I set up ChatGPT for content creation the right way – custom GPTs, voice files, drafting workflows – and wrote a whole blog with it. Here’s where it broke.

Dictation is how I think: I ramble, the AI organizes, I read it back. The voice-memo pipeline behind it, and why typing loses half the ideas.

Is using AI cheating? A science writer takes the six real fears seriously – skill atrophy, theft, disclosure – and shares the line she actually holds.

A Claude Code skill is a markdown file – that’s the whole trick. I tested a downloaded pro skill against my own system across nine posts. Here’s what each won.

A working science writer’s map of AI in research work: where it helps grants and explainers, why it fails literature review, and the line it must not cross.

Claude forgets you between chats. A science writer’s plain-files fix: one folder Claude reads every session – no plugin, nothing to rent.