| <!DOCTYPE html> | |
| <html lang="en"> | |
| <head> | |
| <meta charset="UTF-8"> | |
| <meta name="viewport" content="width=device-width, initial-scale=1.0"> | |
| <meta name="theme-color" content="#07080c"> | |
| <title>The Schema Engine — why imageless thinkers and LLMs converge</title> | |
| <meta name="description" content="A field guide to the imageless mind: how aphantasic thinkers and large language models arrived at the same architecture — representing the world without rendering it."> | |
| <link rel="preconnect" href="https://fonts.googleapis.com"> | |
| <link rel="preconnect" href="https://fonts.gstatic.com" crossorigin> |
| ``` | |
| 𝔸5.1.formal-paper@2026-05-17 | |
| γ≔paper.aisp.formal.v3 | |
| ρ≔⟨inversion,assumptions,protocol,semantics,modalities,empirics,failures,roadmap⟩ | |
| ⊢ND∧CAT∧ΠΣ∧μ∧Ambig<0.02 | |
| ;; ── Ω:FOUNDATION ── | |
| ⟦Ω:Foundation⟧{ | |
| 𝔄≜{⊤⊥∧∨¬→↔∀∃∃!λΠΣ≜≡≢∈∉⊂⊃∪∩∘⊕⊖⊗⟨⟩⟦⟧⊢⊨↦⇒∎} | |
| ⊛:𝔄*→Sym; ⊛≜fix λf a⃗.a⃗≡ε→ι|hd(a⃗)⊗f(tl(a⃗)) |
Written to be spoken. Read it with the pauses left in. This one goes dark before it goes bright. Stay with me through the dark. The bright is worth it.
I want to start tonight by making you uncomfortable, because I think you came here to be comforted, and I respect you too much to do that.
Written to be spoken. Read it with the pauses left in.
I have to begin by giving away the ending, and confessing something it took me years to be able to say out loud.
A retrospective essay, written 2076. Speculative throughout — a thought experiment about where one architecture might lead, not a prediction.
There is a particular kind of mistake that only becomes visible from a distance, and the early 2020s were full of one of them. We believed, almost without noticing we believed it, that intelligence lived inside the model. We measured minds in parameters. We said a system was smarter because it had more weights, the way you might say a library was better because it had more shelves — never mind whether anyone could find a book, never mind whether the books connected to each other, never mind whether the library could think a thought it had not already been shelved to think.
Bradley Ross
Harvard University, ALM in Digital Media Design 2026
Toronto, Ontario, Canada | Cambridge, Mass, USA
bar181@github · linkedin.com/in/bradaross
Version 3.0 (arXiv submission draft) — 2026-05-17
Companion artifacts. Reference implementation, full glossary, replication evidence, and toolchain (aisp-converter, aisp-validator) at github.com/bar181/aisp-open-core. Authoritative protocol source: AI_GUIDE.md (AISP 5.1 Platinum). Canonical AISP-native version of this paper at formal-papers/aisp-protocol.md; the present document is a derived human-readable view per the directional translation principle of §6.7.
Bradley Ross
Harvard Extension School, ALM in Digital Media Design
Toronto, Ontario, Canada
bar181@github · linkedin.com/in/bradaross
Version 3.0 (arXiv submission draft) — 2026-05-17
Companion artifacts. Reference implementation, full glossary, replication evidence, and toolchain (aisp-converter, aisp-validator) at github.com/bar181/aisp-open-core. Authoritative protocol source: AI_GUIDE.md (AISP 5.1 Platinum). Canonical AISP-native version of this paper at paper/AISP-formal-paper.aisp.md; the present document is a derived human-readable view per the directional translation principle of §6.7.
Version: 4.0 (Final)
Date: January 22, 2026
Core Insight: AISP and Coherence Engine are orthogonal complements — AISP ensures precise communication (spec-time), Coherence ensures consistent state (run-time)