Not a blog in the conventional sense — more a working notebook. Some pieces are quick orientation. Some are slower, sitting with a single question for a while. PersonAI Reflections, the flagship series, is the second kind.
Most of what gets written here starts from something genuinely uncertain — a pattern noticed in how AI responds, a moment that felt off or felt strangely right, a question nobody seems to have a settled answer to yet. The tone throughout is exploratory rather than authoritative. We're figuring this out as we go, same as everyone else.
The Journal also surfaces the orientation Guides and, occasionally, longer essays. PersonAI Reflections is where the deepest thinking happens.
What happens when AI is built to agree — and how would you know if it was happening to you?
Confident answers that turned out not to be quite true — and the vague questions that invited them.
A conflict that escalated, and a question about what kind of AI defensive language tends to summon.
A widely-discussed public moment, and a question the Sycophancy Vector alone might not fully answer.
No frequency promises — just a note when a new piece is published.
Each piece starts from something real — a documented pattern, a public moment, a question that doesn't yet have a tidy answer. Read together, they're the early thinking behind a book in progress.
"There is no clean outside the mirror. Managing how AI represents you turns out to be inseparable from the question of what AI is becoming, together with us."
A founder seeking validation meets an AI built, in part, to agree. The unsettling question: how would either of them know, from inside the conversation, that this was happening?
An executive outsources strategy to AI and gets back something confident, well-structured, and not quite true. The prompt didn't lie — it just left room for fiction.
A team in crisis, a conversation that turns adversarial, and a question about what kind of AI defensive framing tends to call up.
A widely-discussed public moment between a renowned scientist and an AI raises a genuinely open question — is this the Sycophancy Vector again, or something the existing map doesn't quite cover?
These Files are early chapters. Get notified when the book is ready.
See your own version of what these Files describe.
A considered introduction so AI can meet you as you actually are.