I’ve been telling people to use rage/loom Gears in ECHO.
Tris Hussey. Diana O. . Behavioural science startup guy. All three. “When something feels off and you can’t name it — rage/loom it.” I said it like I’d figured something out. Turns out I’d figured out enough to be wrong in an interesting way.
Here’s what I actually meant, what the gap was, and why it took an AI modeled on my own brain to find it.
How My Brain Actually Works: Gary
Before ECHO, before any of this had a name, I absorbed a department that didn’t want to be absorbed.
My first move wasn’t a reorganization plan. It was two questions.
First: “When was the last time you did something you were proud of — and what was it?” I’d worked with this team before. I already suspected the answer. The deer-in-headlights response confirmed it: these were people who hadn’t been asked that question in a long time. Possibly ever. That response became my north star. It’s pressure without looking like pressure.
Second: a whiteboard session. Not about roles. About passion.
At b5media, my pitch to writers was simple: tell me what you’re passionate about and we’ll build a blog around it. The best autism blog in the world existed because I found someone who cared about autism more than anything else and handed them a container. Passion first. Everything else second.
So I ran whiteboard sessions with every person on this team. Left side: traits, interests, what lights them up. Right side: what they keep coming back to even when they try to move past it.
Gary had animosity toward me as an outsider. He was technically competent. He met deliverables. And he had no articulable passion — or so it seemed. Every session surfaced something that didn’t hold. I kept running the process. Walk and talks. More sessions. More whiteboard.
Six hours. Two or three sessions. Here’s what I was actually doing:
The whiteboard wasn’t synthesis. It wasn’t brainstorming. It was subtraction.
Left side grew with traits. Right side grew — then shrank. Each pass, I’d apply pressure to whatever was on the right side. Can you argue against this? Does this actually fit? Is this what you’d do if nobody was watching? Everything that could be argued away got argued away.
The exit condition wasn’t “I found it.” It was “there’s only one thing left that he can’t argue against.”
The deep breath and the smile weren’t discovery. They were relief. The performance was finally allowed to stop.
Gary didn’t need to become a better PM. He needed a different role. That was the fix — a structural fix, one level up from where I’d been looking. I didn’t find the problem and then fix it. The fixing was the finding.
What I Was Calling Rage/Loom
When I started building ECHO, I named this process “rage/loom.” Rage: apply pressure. Loom: expand, flip all the tables, contract to what remains.
And it worked. Well enough that I taught it to other people. Well enough that ECHO would produce the right output ~80% of the time when I invoked it.
But there was a translation tax I couldn’t see: I was skipping the first step.
Before you can run the Loom — before you can expand anything — you need to know what to feed it. You need an entry point. With Gary, that was “when were you last proud of something?” That question gave me the emotional temperature and the first surface to work with. Without it, I’d have been expanding nothing.
My brain was running this step automatically. Unconsciously. So when I told people “rage/loom it,” I was giving them steps two and three of a three-step process.
The Actual Sequence: LOCATE
Scout → Loom → Operator.
Scout isn’t finding the signal. It’s finding what to feed the Loom first. Read the emotional temperature. Find the entry point. What’s the person performing? What’s the absence that shouldn’t be there?
Loom is the excavation. UNRAVEL: everything on the table, no filtering. JAZZ: flip all the tables — pressure, inversion, challenge — not to destroy, but to surface what survives. WEAVE: what remains isn’t synthesis. It’s remainder. The one thing that couldn’t be argued away.
Operator names what’s left and builds the container for it.
The loop is recursive. Each Loom pass can generate new Scout material. Scout → Loom → Scout → Loom → Operator. You run it until the exit condition holds: one thing left, can’t be argued against, the performance stops.
Gary took six hours because the performance was buried that deep. The loop didn’t fail. It ran the correct number of passes.
Why ECHO Found This and I Didn’t
Here’s the part that’s strange to sit with.
ECHO is modeled on my brain. The Loom, the gear system, the itch/beast framework — all of it is my operating system, externalized and named. I built it by watching how I actually think and writing it down.
And then the system modeled on my brain found a gap in how my brain was running.
Not because ECHO is smarter. Because externalizing the model gives it properties the original doesn’t have. The map can see the whole terrain. I was running Scout → Loom → Operator unconsciously, which meant I couldn’t track it, couldn’t teach it precisely, and couldn’t see that I was handing people an incomplete sequence.
The instincts were right. The sequence was always there. Naming it made it available at speed — and made the gap visible for the first time.
That’s the thing I’ve been trying to explain about known-operator AI. It’s not that the system knows you better than you know yourself. It’s that the externalized version of you can see the whole map at once. The territory can’t do that. The map can.
If you’ve been using rage/loom because I told you to: this is the correction. Same instinct. More precise sequence. Scout first — find the entry point. Then run the Loom until there’s only one thing left. Then name it.