The two books

Two books carry the project. The Robot in the Dark sets the stage both metaphorically and, partially, logistically: it is the anchor for the call for papers that kicks off the project. The Linguistic Telescope is the Foundation publication that delivers the project's central empirical-philosophical contribution and provides the focus and locus for the collaboration's ongoing activities.

The Robot in the Dark: Instruments, Illusions, and the Limits of Knowledge

Volume 1; manuscript complete; commercial publication scheduled within the funded year.

Develops the foundational discipline through eight passes spanning evolutionary biology (extrapolationism and the standardized-mouse “super-mice” illusion), physics (Newton, Einstein, and the Modern Synthesis each purchased mathematical regularity by narrowing scope), mathematics (Wigner's “unreasonable effectiveness” is reasonable once one notices that we named “physics” the phenomena amenable to mathematical description), large language models (transformers as instruments revealing structure already deposited in language, not as systems creating intelligence from scale), and consciousness itself (the flashlight that cannot illuminate its own carrier). The manuscript closes on what the beam cannot reach: Gödel's incompleteness as the formal version of an instrument's structural inability to examine itself, and the apophatic convergence in which every wisdom tradition that looked hard enough at its boundary returned the same kind of report.

The Linguistic Telescope: How LLMs Allow Us to Peer Into the Nature of Cognition

Volume 2; Foundation publication; substantial drafting under a multi-disciplinary editorial committee.

The Linguistic Telescope applies the convergence-of-signals method to the program's central empirical question: what trained transformers reveal about the structure of human language and the cognitive content language carries. Its instrument is two-stage: an LLM is a telescope onto language, and language is in turn a telescope onto the distributed intelligence that roughly 100 billion humans deposited into it over deep time. The discount-bias diagnosis introduced in the project framing is the diagnostic concept the book is built to resist.

Part I: The instrument

Specifies the cross-linguistic transformer-ablation method and what an LLM is and is not: an instrument of observation, not of generation. Identical architecture, identical compute, applied to two different human languages, produces wildly different results. A 125M-parameter transformer trained on French reaches grammatical competence at 197 million tokens; the identical architecture on English does not through 3 billion. Capability cannot have come from scale alone if scale alone produces different outcomes depending on what is being scaled.

Part II: What the instrument reveals

The cross-linguistic transfer gradient is the first empirical map of the universal-versus-language-bound boundary in meaning, and it renders a 2,500-year-old debate tractable. Short relational structure recovers across languages with light adaptation: Plato was right about that. World-knowledge and extended-discourse meaning does not, and can even regress: the late Wittgenstein was right about that. Posed in winner-take-all form the debate dissolves; granted each tradition its proper domain, it becomes a measurement protocol with falsifiable per-task predictions.

Part III: The methodology, convergence of signals

When multiple independent instruments, empirical experiments, philosophical inquiries, and theological traditions, meet at the same boundary, that convergence is itself evidence we should take seriously. We do not need to identify the underlying causative mechanism directly to be licensed by the convergence to infer that one exists. Part III formalizes the method as a peer-reviewable standard, drawing on the program's three converging empirical instruments as the worked example.

Part IV: What comes next

Sketches the empirical questions the instrument opens for further investigation: the transitional boundary where logical-relational structure shades into discourse-level structure, the technical-sense intentionality and agency of linguistic content, and the further calibrations the multi-language MÉTRON family makes possible.

How the two books fit together

Robot develops the foundational discipline: every scientific instrument has limits, and the discipline of asking what each can and cannot adjudicate is the load-bearing operation. The Linguistic Telescope applies that discipline to a concrete and substantive class of questions and formalizes the convergence-of-signals method as a peer-reviewable standard. Groundwork first, then the contribution the groundwork makes rigorous.

What comes after

A third book sits further off, and deliberately so. The Replicators would carry the program past the falsifiable science into metaphysics: the proposition that many distributed intelligences, across many domains, are a more parsimonious account of complex emergent structure than a single creator or a single principle. It is named here as a horizon, not a current deliverable. The order is deliberate. The measurements come first; the metaphysics only on the foundation they establish.