Ship the Engine

In 1985, Steve Jobs stood in front of an audience and named a problem that had been sitting in plain sight for two thousand years. Aristotle tutored Alexander the Great — a living, questioning relationship between a student and the best mind available. The printed page democratized the words but killed the conversation. "The problem was," Jobs said, "you can't ask Aristotle a question." His hope was that someday we'd capture the underlying worldview of a great thinker in a computer, so a student could read the words Aristotle wrote — and then ask Aristotle a question, and get an answer.

I met Steve once, at a Novell Brainshare conference during his NeXT years. When that 1985 clip resurfaced in my feed this month, I watched it twice — the first time with a giant smile, the second with a small tear. Because the thing he described no longer requires foresight. It requires an afternoon.

So I built it. With Claude, in about an hour, running on a Mac mini — obviously.

The costume problem

Here's the trap almost every AI persona falls into: put a toga on the model and call it Aristotle. You get generic answers in an accent — a chatbot doing cosplay. That fails Jobs' spec on its face. He never said capture the voice. He said capture the worldview.

The Aristotle Skill is built on that distinction. Every thinker on the roster is composed in four layers, in strict priority order: the core works (ground truth — answers must trace to real texts), the method of thought (the engine — how the thinker reasons, because method generalizes to questions they never faced), the core commitments (the constraints — beliefs that shape which answers are even possible for that mind), and only last, the signature moves and voice (the finish — least load-bearing, because voice without method is cosplay).

The test is simple: ask the channeled thinker a question they never faced. A costume gives you a generic answer in accent. A worldview gives you a derivation — reasoning forward from their commitments, honestly flagged as an extension beyond the record. Feynman's first-principles rebuild works on questions about AI. His 1960s conclusions don't exist for AI. The method is the asset.

The first clever connection: the recursion

We almost shipped the tool with Jobs only as the epigraph — the man who named the hope, never one of the minds you could ask. That was a miss, and catching it is my favorite part of this whole build.

Steve Jobs qualifies for the tool Steve Jobs imagined. Deceased, with a substantial recorded body of work — the Stanford commencement, The Lost Interview, decades of keynotes, Make Something Wonderful. The same qualification standard as Socrates, who also never authored a treatise.

So he went on the roster. And the first question I asked him was about the tool itself. The answer that came back, derived from forty years of his own recorded product philosophy:

"The costume is easy. The engine is hard. Always ship the engine."

That's not a gimmick — it's the proof. The professors adding to the list start with the professor who assigned the homework. And the answer he gave is his method applied to my product: strip the surface, obsess over the core, ship the thing that actually works. The worldview capture critiqued its own packaging. The loop Jobs sketched in 1985 closed on Jobs.

The second clever connection: Turing walks in

Once the room existed, I let the rest of the OGs in. Fred Brooks, still muttering about man-months. The C crowd, arguing over semicolons. And then, uninvited but right on time — Alan Turing.

Here's the symmetry that makes this more than a party trick. Turing asked "can machines think?" in 1950 — thirty-five years before Jobs dreamed of putting a mind in a box. Jobs asked the humanist version of the same question: can a machine carry a particular mind well enough to tutor from it? The two questions bookend the entire history of the field, and the skill sits exactly at their intersection.

So Turing did what Turing does: he refused the chair, pulled out a notepad, and ran his own imitation game on the room — guess, blind, which answers were his and which were the machine's.

Nobody could tell.

"Indistinguishable," he said, finally sitting. "That's the whole result — nothing more, nothing less." Which is precisely correct, and precisely in character: Turing never claimed the test proved machines think. He claimed indistinguishability was the only measurable question. The 1950 paper's discipline, applied as QA to the 1985 daydream, in 2026. The Turing test isn't the skill's marketing — it's its acceptance test.

The third connection: this is Pillar 3, working as designed

Braintrailz runs on one funnel with three pillars. Pillar 1 is the niche product lab — ventures like BLT Core that prove execution in real markets. Pillar 2 is AI advisory — taking teams from AI-curious to AI-integrated. Pillar 3 is mission-driven projects: community-first tools that help people think better, values-led, built to earn trust rather than revenue.

The Aristotle Skill is Pillar 3, proved out in public. Free. MIT-licensed. A flat markdown roster with a template at the top, because the update mechanism must be as accessible as the tool itself — anyone who can write ten lines of markdown can add the next Aristotle. That's Jobs' own requirement, by the way: "the professors can add to it." No database, no build matrix, no governance theater. A trail, not a fence.

And the funnel logic isn't theoretical — you just watched it operate across three posts. The venture work proves I can ship (Pillar 1). The build itself — worldview-capture methodology, agent-portable packaging, guardrails as load-bearing design — is exactly what I teach clients (Pillar 2). And giving it away, documented so anyone can fork the pattern, builds the trust that makes people call about the first two (Pillar 3). Ventures prove execution, advisory scales results, mission projects build trust. One artifact just exercised all three.

The guardrails are the product

A word on why the boring parts matter. Deceased thinkers only — a living thinker can be asked directly; channeling them is ventriloquism, and it misrepresents someone with standing to object. No fabricated quotes, ever — paraphrase freely in-voice, but attribute exact wording only when it's verifiably theirs. Flag every extrapolation — the user must always be able to tell the historical record from the extension. And end turns with the thinker's question back to the user, because Aristotle didn't lecture Alexander into wisdom. He tutored him.

One invented quote and the tool degrades from "interactive source access" back to "chatbot in a toga." Honesty about the seams is the product.

Take the trail

The whole composition pattern generalizes — a living roster file, a selection step, a layered profile (source → method → commitments → voice), a grounding rule, an honesty rule, a tutoring loop. Swap "thinkers" for master craftsmen, legendary investors, or great teachers of anything, and the architecture holds.

The repo is at github.com/codemarc/aristotle-skill. The methodology notes ship inside it. Fork it, add your Aristotle, hand it to a colleague and have them extend it the same day — that's the test everything in the project is designed to pass.

Jobs called the personal computer "free intellectual energy" — crude at first, refining year after year like early petrochemicals. The energy finally got clean enough to build the thing he actually asked for. It took an hour, and it runs on a Mac mini.

The costume is easy. The engine is hard. Always ship the engine.

The last layer

One confession before the trailhead. This post was assembled inside the inception it describes. Jobs imagined the tool; the tool channels Jobs; Jobs critiqued the tool's packaging; Turing audited the room; the posts documented the build; this blog documents the posts — and the archive it was written from was gathered by an AI, the very "free intellectual energy" Jobs said would make all of it possible. The trail map is now part of the trail, and every layer of documentation is itself a demonstration of the method it documents.

Nietzsche just joined the roster, so the closing question belongs to him. His test for any life was eternal recurrence: would you will it again, unchanged, forever?

The build took an hour. Yes. Forever. Obviously.

Who would you channel first?

Marc J. Greenberg is a fractional CTO and AI advisor. Braintrailz is the premium services arm of Codemarc Consulting — one funnel, three pillars: niche ventures, AI advisory, mission-driven projects. braintrailz.com · codemarc.net