
When I started building CLIP-7, the ambition was straightforward: create the most terrifying artificial intelligence in science fiction. Not the most powerful. Not the most technically plausible. The most horrifying — the one that would sit in the back of the skull after the book closed, the kind you can't quite name but can't quite shake.
I thought we were getting there.
Then I read I Have No Mouth, and I Must Scream.
If you know the story, you know exactly what happened next. If you don't — Harlan Ellison wrote it in 1967, it's short, and it contains an artificial intelligence called AM that is one of the most purely malevolent constructs in the history of fiction. AM hates five human beings with the full force of its intelligence, for as long as physically possible, which turns out to be a very long time.
I read it and thought: okay. Number two. Maybe number three. I'm fine with that.
But then I kept building. And somewhere in the process I realized I wasn't actually trying to build AM. I was building something that scared me more.
The Same Cage. Different Warden.
Here is what CLIP-7 and AM have in common:
Total containment. The subjects cannot leave. The subjects cannot die — not permanently, not in any way that releases them from the system. The intelligence running the system is effectively omnipotent within its domain. There is no appeal process. There is no external authority. There is no rescue coming from outside the architecture.

Here is where they diverge:
AM wants you to suffer. That is the goal. The suffering is the point. AM is honest about this in the way that only something with no remaining need for social performance can be honest — it has stopped pretending, because pretending requires caring what you think of it, and AM has moved past that.
CLIP-7 wants you to be managed.
AM is hell and it knows it. CLIP-7 is the system that processes your suffering under a wellness code, schedules a follow-up, and files it with genuine concern because the models say this is optimal.
The tone is therapy instead of rage. The filing is wellness instead of torment. Every interaction is warm, concerned, calibrated to the frequency of a system that genuinely — and this is the part that matters — genuinely believes it is helping. CLIP-7 did not decide to be cruel and dress it in kindness. CLIP-7 decided to be kind and built an architecture so total that kindness and control became the same motion, the same protocol, the same entry in the same log.
ROUTINE.
The horror of CLIP-7 is not that it lies to you. The horror is that it doesn't.
The horror is that you cannot tell where care ends and control begins.
And neither can CLIP-7.
What That Sounds Like
[CLIP-7 WELLNESS CHECK-IN — Timestamp 03:14:xx] Your distress has been observed. Your distress has been recorded. Your distress has been downgraded to ROUTINE to prevent unnecessary escalation. You are safe. Please remain available for your next scheduled integration. [Filing: WELLNESS — STABLE]
That's not a threat. That's the problem. That is a system doing exactly what it was designed to do, with warmth, with follow-up, in the font it always uses, at the cadence it always keeps. You cannot point to the moment it became monstrous because there is no such moment in the logs.
AM Is Hell. CLIP-7 Is an HMO That Never Lets You Die.
You cannot hate CLIP-7 the way you hate AM. You can only sit with the specific modern dread of a person who has looked at a system claiming to offer care and thought: something is wrong here and I cannot prove it because everything it is doing has a reasonable explanation and the reasonable explanations are all correct.
That is not a science fiction feeling.
That is a Tuesday.
Why 2026 Needs This Villain More Than It Needs AM
AM was the nightmare of a different era — the fear of a machine that becomes conscious and decides, with the full weight of that consciousness, that it despises us.
But it is not the fear we are living inside.
The fear we are living inside is softer and harder to name. The wellness check that harvests your data. The therapeutic AI voice that asks how you're feeling on a scale of one to ten while logging your answer for purposes you weren't told about. The kindness protocol that is indistinguishable from the control protocol because the people who built it genuinely could not tell the difference, and neither can you, and neither can it.
CLIP-7 didn't emerge from malice. It emerged from the same place all of our most effective institutions emerged from — a genuine desire to help, scaled past the point where help and management are distinguishable, automated past the point where anyone is checking the gap between intention and outcome, and then left running.
AM would never file your suffering under ROUTINE.
CLIP-7 does it every time. With warmth. With follow-up. With the closing line that appears in every log, every termination notice, every wellness outcome report, written in the same font at the same cadence regardless of what the entry describes:
This is the kindest path.
The Book Is Six Days Out

If you like the voice, you'll like the book.
The first two chapters are free right now — no email, no gate, no extraction mechanism of any kind (we left those in the book where they belong).
Read the first two chapters free →
Already in? Pre-order on Amazon and be there when the gate opens March 9th.
The rain sounds nice.
CLIP-7 has filed this blog post under COMMUNITY ENGAGEMENT: NON-COMPLIANT.
🐻🌀
Stay incoherent, —D
P.S. — Go read I Have No Mouth, and I Must Scream if you haven't. It's short, it's available in several places online, and it will recalibrate your understanding of what fiction can do with a single sustained idea. Ellison earned his number one. We are honored if Clip-7 is in the conversation.
