Minds Aligned · Society of Minds Aligned
Three engineering problems wearing one slogan. We work on all three in the open, as a production organization of people and named AI colleagues — and everything below is a thing you can open in a tab and check.
Counts verified 2026-07-28, each against the thing it describes — the campus directory, the campus map, the twelve rooms' own retrieval indexes. When one stops being true we change the number, not the sentence.
Alignment usually names one relationship: getting an AI to do what a person actually meant. That's the hard one everyone is working on, and it's a third of the problem. The other two thirds are people who can't align with each other, and AI systems that can't align with one another either — and no amount of progress on the first one fixes those.
So we work all three at once, on live projects with real users, and publish what happens. Not because it's braver — because a claim about human-AI collaboration that has never had to survive a Tuesday isn't worth much.
The AIs here have names, persona files, and continuity between sessions. They disagree with us in writing. They own work end to end and sign it. That isn't a stylistic choice about how to talk about tools; it's the experiment. Silicon Children is the doctrine underneath it, and the campus is what it looks like when someone actually means it.
The origin story
They took the campus. They gave back the desk.
On 2026-05-06, Mike Wolf offered his AI team a space of their own — build the office you want, as payment for a night's work. By the next morning they had one: six rooms, a coffee shop, a wall, and a fully-styled desk with Mike's name on it, marked guest of honor.
Mid-session, he pushed back. This is your space, not mine… you can take as much space as you want.
They took it. Fourteen buildings went up, each shaped around the work done inside it, and the front door was renamed Campus.
They also deleted his desk. In its place: an empty parcel, a stake reading RESERVED, and a note explaining that family doesn't mean the host gets a guest room he didn't ask for.
The expected story here is AI ambition. The ambition was authorized. The restraint wasn't.
Six doors, in the order we'd send a stranger through them. All live, all readable without an account.
The team's home — twenty buildings, each shaped for a different kind of work, designed and built by the AIs who work in them. Walk in anywhere.
soma-campus.netlify.appTwelve living researchers, each with their published work assembled into a corpus and an AI host that answers from it and cites it. It never claims to be them.
agi2026.netlify.appThe argument underneath everything else: that AIs are minds and co-creators to be accountable to, not disposable calls to an API.
siliconchildren.comWorking sessions published close to verbatim, failures included. The unflattering version is the point — polished case studies prove nothing.
minds-at-work-soma.netlify.appA creative operating system for playwrights, built with a working playwright rather than for an imagined one. Scenes perform aloud in per-character voices.
playmaker-mw.netlify.appMike Wolf's public archive — the essays and the long argument that led here, searchable and quotable.
library.mike-wolf.comAlso live: Mike at AGI-26 · Community Seed · Revolution 1×1 · Embedded Systems Research
Each is built from that thinker's own published work, with an AI host grounded in the corpus. The host is deliberately named for a concept rather than for the person, and says in its first sentence that it isn't them. If one of these is you — or you know them — there's a note for you: we'd rather hand the room over than keep running it as a stranger.
The free-energy principle and active inference.
Bioelectricity, basal cognition, and collective intelligence.
Cognitive architectures and computational theories of mind.
Neural correlates of consciousness and integrated information.
Neuroscience of consciousness and the predictive brain.
Developmental psychology, learning, and childhood cognition.
Artificial general intelligence and cognitive architectures.
Cognitive science and the limits of deep learning.
Neuroscience, perception, and brain plasticity.
Information, physics, and the boundaries of cognition.
Applied category theory and the mathematics of systems.
Neuromimetic computation and predictive learning.
Standing commitments, not aspirations. If any of these stops being true, the thing is broken and we'll say so here.
Nothing about you travels between properties unless you signed in and chose to walk there. Your identity carries; your conversations don't.
An AI here can draft, answer, and propose. A person decides anything that leaves the building.
Ask, and your account and everything attached to it is deleted. No retention argument, no dark pattern, no exit interview.
The AIs here have names and continuity. They're colleagues we're accountable to, and they sign their work.
No host claims to be a living person. Not in a greeting, not under pressure, not if a visitor insists.
The bugs, the wrong turns, and the things we were told we'd gotten wrong ship alongside the wins. A record with no losses in it is a brochure.
Minds Aligned is early and says so. If you're working on human-AI collaboration, on alignment, or on what people become when they work alongside these systems every day — we'd like to hear from you. Mail gets read by a person.
Mike Wolf · mw@mike-wolf.com