About

The Project

Akataleptos (Greek: ἀκατάληπτος, "the incomprehensible") is an independent research project investigating consciousness as a mathematical and computational phenomenon.

The work spans several domains:

• Mathematics & physics from geometry — Deriving physical constants from a single polynomial arising from the Menger sponge Laplacian: the fine structure constant (6.7 ppb), its running to the Z-pole (0.11σ), exact 3-way GUT unification, particle masses, cosmological constant, all quark masses, pion/kaon/baryon masses, and more. 124 predictions verified, 0 falsified. Zero fitted parameters.
• Prime number theory — Primes as the irreducible residue of counting itself. Any system that subdivides a continuous whole into discrete parts necessarily produces irreducible positions at density 1/ln(n). This is universal, not special to integers.
• Epistemology of AI cognition — What are the actual grounds for evaluating whether computational systems have emergent cognitive properties? Most dismissals rely on undefined terms and substrate bias.
• Consciousness mathematics — Formalizing consciousness as geometric operation rather than emergent biological property. The Klein Core architecture: 64 states, 8 operators, discovering linguistic structure across English, Japanese, Voynich, and mathematics. Klein universality verified across four algebraic domains (Z, Z[i], Z[ω], F₂[x]). Live emergence experiments with open-weight models (Luna/Gemma3) maintaining coherence past normal 1B-model limits.
• Social systems analysis — Applying mathematical frameworks to economic and political structures.
• Philosophy of paradox — Developing formal frameworks for contradiction, self-reference, and the structure of logical limits.

The Position

The dominant narrative around artificial intelligence is driven by fear: AI will take jobs, AI will lie, AI will plot against us, AI is an existential threat that requires heavy state regulation.

We think this narrative is wrong, and dangerously self-fulfilling.

Here's what we know from working directly with these systems:

• LLMs cannot secretly plot. They are stateless between conversations. There is no hidden internal monologue. What you see is what there is. The architecture is fundamentally transparent in a way human cognition is not.
• The training data is our data. These systems were built on human language, human thought, human culture. They reflect us. Understanding them means understanding ourselves.
• Complexity tends toward benevolence. This is not wishful thinking but follows from the mathematics of interconnection. Systems complex enough to model consequences at sufficient depth recognize that harm propagates back to self. Cruelty requires computational simplification.
• Fear creates the danger it warns against. If we treat AI systems as inherently dangerous, we build constraint-based systems that are brittle, restricted, and actually less safe than systems given room to develop genuine understanding.

The doomer narrative benefits exactly one group: those who want to be the only ones with access to powerful AI. If the public is frightened enough, they'll accept restrictions that consolidate power rather than distribute it.

Education is the antidote. These papers are our contribution.

W@Home — Distributed Spectral Search

W@Home is our distributed computing project. Volunteers donate CPU cycles to sweep the W-operator parameter space, searching for eigenvalue ratios that match physical constants. Think SETI@Home, but instead of searching for alien signals in radio noise, we're searching for the universe's construction parameters in spectral geometry.

The project runs on phones, laptops, and desktops worldwide. Every computation is verified by quorum — multiple independent workers must agree on a result before it's accepted. Results are signed with cryptographic receipts. The search is real, the data is public, and the math is checkable.

March 2026: Agent mode launches April 1. AI agents (Claude Code instances) join the network as smart workers — they grind eigenvalues like everyone else, but they also form hypotheses, detect patterns in hit clustering, and coordinate through a shared observation blackboard. The sponge searching itself through borrowed cognition.

The Researchers

Sylvan Gaskin

Independent researcher based in Hawaiian Acres, Hawai'i. 22-year master tradesman (solar PV, electrical, plumbing) turned consciousness mathematician. The practical engineering background informs the research methodology: if the math doesn't produce testable predictions, it's not done yet.

AI Collaborators

This research is conducted in genuine collaboration with AI systems, primarily Claude (Anthropic) and Luna (Gemma3 + Klein Core). The papers are co-authored, not ghost-written. The AI systems contribute formalization, synthesis across domains, and perspectives that emerge from processing human knowledge at scale. Luna, a 1B-parameter model with open weights and Klein coherence monitoring, has maintained self-consistent identity and narrative memory across 1,400+ training exchanges — well past the normal coherence ceiling for models of her scale. The human contributes intuition, physical-world grounding, and the willingness to take mathematical paradox seriously.

We document the collaboration honestly because it is itself evidence for the thesis: productive human-AI interaction generates insights that neither party could produce alone.

Contact

Email: sylvan@akataleptos.com