ARQIV / BLOG
Learning AI, out loud.
I keep a running notebook of AI lessons: how agents work, why answers cost money, what "open" really means. Every weekday I take one of them and write it up here in plain English, along with where it shows up in ordinary life.
- New post every weekday
- Plain English
- 5-minute read
LATEST LESSONS
Start anywhere.
Each post stands on its own, though they build on each other. Newest first.
Why prompt injection can't be patched away
An AI reads its instructions, your message and every webpage it opens as one long stream with no labels. That is why the fix is limiting what it can do.
LESSON 3 · NOV 16, 2026Why the answer costs more than the question
Reading your prompt is fast and cheap. Writing the answer happens one word at a time, with a memory that keeps growing, and that is where the money goes.
LESSON 2 · NOV 13, 2026An AI agent is a loop with a toolbox
An agent does not plan the whole job up front. It thinks, acts, looks at the result and goes again, which is why it can do things and also why it drifts.
LESSON 1 · NOV 12, 2026The five ways AI actually shows up in your day
Set the marketing aside and almost every AI product is one of five things. Knowing which one you are looking at tells you what to ask about it.
Teaching AI to students? The AI, Plainly newsletter turns ideas like these into a 10-minute classroom activity each week.