How to Keep Your AI Character From Becoming a Stranger
Ask an AI for the same character twice and you meet two different people. The fix is two moves: choose a canon image, and send it with every single request.
Free to read
Category archive
Thinking with Circuits
Observing AI through a human lens. How it mimics and misses the human. Tool reviews, insights, and explorations of AI like Cursor, Claude, Hedra, and more.
38 entries, newest first
How to Get Claude Code Running in Your Terminal (No Coding Needed)
I went to school for tech, so the terminal has never been the scary part for me. But I know what "just run it in the terminal" sounds like from the outside. Fifteen minutes, every command for Windows and Mac, and the exact output you should see at each step.
Free to read
Six Platforms Came for AI Slop. None of Them Banned AI.
A pineapple with abs is cheating on a strawberry and thirty-nine million people watched it happen. Then in fifteen days six platforms moved against AI slop at once. Read the policies side by side and not one of them banned AI. What they banned was volume.
Free to read
The Machine That Runs This Website
This post put itself on the schedule on a Tuesday night while I was doing something else. Let me walk you through what actually runs, because most of it is boring, and boring is the whole point.
Free to read
What Actually Happens When You Let AI Use Your Browser
Two weeks ago my terminal got a new door into my browser. Since April I had been letting AI use a separate browser built for the job, one that knows only what I signed it into. Then the extension updated and it could open the browser I actually live in. Here is what that changes, where the machine still stops and waits for me, and why the work moved rather than disappeared.
Free to read
Local Models Are Catching Up. It Depends What You Ask Them.
A model you rent and a model you keep are different kinds of things. One lives on somebody else's servers and can be shut off on a Tuesday. The other sits on your hard drive and works in a blackout. What running AI on your own computer actually gets you, what it does not, and how to tell which one you need.
Free to read
Two Frontier Models Landed This Week. I Get to Build With Both.
Two frontier models landed in one week. Anthropic extended Claude Fable 5 to July 19 and OpenAI shipped GPT-5.6 Sol, Terra, and Luna. Here is why more capable models at more price points is good news for anyone who builds, and why the hard part was never the tool.
Free to read
I Asked Fable to Build My Jarvis. Meet Veronica.
In April I wrote about Mythos, the model Anthropic said was too dangerous to release. In June they released Fable, a less dangerous version of that model. Then it built my Jarvis. Her name is Veronica.
Free to read
The Next AI Divide Is Not Access. It Is Workflow.
Everybody has access to AI now. But the next divide is not who can open a chatbot. It is who can build workflows, systems, and repeatable processes that turn AI into real work.
Free to read
The Machine Said No, 1.1 Billion Times
Workday's hiring software rejected around 1.1 billion job applications, most of them never seen by a person. Here is what the lawsuit reveals, why hiring is exactly where human oversight should be non-negotiable, and how the rules meant to check this keep falling apart.
Free to read