Whitby & Oxford · United Kingdom

THEBES

Intensional Computing, Provenance Kernels & Constructive Geometry

Conventional software and machine-learning infrastructure treats state extensionally: overwriting memory in place, discarding the path by which a result was reached, and smoothing incompatible perspectives into unsigned averages. Thebes builds deterministic, path-typed memory kernels, formal verification suites, and offline security auditors in which arrival history is preserved as structure.

Practice & Architecture

01 · Kernel

Path-Typed Memory (Discovery)

Single-writer memory-mapped (mmap) segments in native Zig and Python. Content-addressed state interning, 64-byte cache-line edges, hardware bit-reversed Calkin–Wilf / Stern–Brocot ordinals, symmetric three-slot relational propagation, and explicit two-branch gates where conflicting evidence remains live.

02 · Assurance

Telemetry, Entropy & Taint Auditing

Multi-hop alias and taint reachability, control-flow state-machine bypass detection across API and physical actuation boundaries (CWE-306 / OWASP ASI10), and offline algebraic & lattice state-recovery verification for truncated pseudo-random streams (CWE-330).

03 · Geometry

Constructive Verification

Exact symbolic computation over finite rationals, periodic surds, and coinductive streams without complex-plane or floating-point collapse; choice-free formal verification in Lean 4; macromolecular fold-topology dissection; and momentless two-ruling signal analysis.

Interactive Geometric Exhibitions

Technical & Machine-Readable Specifications

The static insignia above encodes chromostereoptic depth and a 34:35 counter-raked cut-ellipse Moiré interference field directly in pure vector geometry, without CSS or JavaScript animation. Full mathematical definitions and architectural notes for researchers and automated evaluation agents are published in plain Markdown at /llms.txt and /primer.md.