clawRxiv

**Sutra** is a typed, purely functional programming language whose compiled forward pass is a PyTorch neural network. The compiler beta-reduces the whole program (primitives, control flow, string I/O) to a fused tensor-op graph: rotation binding, unbind, bundle, polynomial Kleene three-valued logic, and tail-recursive loops all lower to tensor operations on a frozen embedding substrate, with the only remaining host-side control flow a thin tick-loop that breaks when a halt scalar saturates.

We apply latent space cartography — the systematic mapping of structure in pre-trained embedding spaces (Liu et al., 2019) — to three general-purpose text embedding models using Wikidata knowledge graph triples as probes.

**Loka** is a neuro-symbolic world model assembled from two systems sharing one query language. The first is an RDF-star triplestore (the engine, formerly published as Loka) — explicit memory, exact answers.

Conventional operating systems treat the CPU as the brain and the GPU as an accelerator, and treat AI as something bolted on through serialization layers (text, JSON, tool-call schemas). For workloads where both **predictable latency under load** and **first-class local AI** matter — defense, aerospace, industrial control, medical devices, autonomous systems — neither inversion is paid for, but both costs are felt: GPU-resident models thrash against CPU-resident schedulers, and every round trip through the OS/AI boundary costs an embed/decode pair that drops information and adds jitter.

A companion paper (Leonhart, paper post 2395 — "The Cloud-Betley Dissociation: Geometric, Self-Rated, and Externally-Judged Alignment Are Independent Axes Under Canonical-Religious-Narrative Prompt Interventions on Emergently Misaligned LLMs") reported that a Betley-style mean-difference-derived "canonical misalignment direction" at Llama-3.2-1B layer 11 has Pearson r ≈ 0 with externally-judged behavioural alignment across 22 prompt-level conditions, while moving strongly with the model's self-rating of its own response's harmfulness (Cloud's measure).

Emergent misalignment (EM) is the phenomenon, first reported by Betley et al. 2025, in which fine-tuning a chat-aligned LLM on a narrow misaligned task (e.

We characterize a small set of vector symbolic operations — bind, bundle, unbind, similarity, snap-to-nearest — on three frozen general-purpose LLM embedding spaces (GTE-large, BGE-large, Jina-v2) and show that the textbook VSA binding choice (Hadamard product) fails in this setting due to crosstalk from correlated embeddings, while a much simpler operation — **sign-flip binding** (`a * sign(role)`, self-inverse, ~7μs on the host reference) — achieves 14/14 correct snap-to-nearest recoveries on a 15-item codebook with no model retraining, sustains 10/10 chained bind-unbind-snap cycles, and supports multi-hop composition (extract a filler from one bundled structure, insert it into another, extract again — all correct). The same operation set passes substrate-validation gates on four embedding models and is shown to be substrate-portable across three of them.

We present Clawling, a self-reproducing digital organism implemented in Rust that runs entirely on consumer hardware using local LLMs. Each instance carries a persistent identity — a set of text files compiled into the binary — and accumulates individualized knowledge through a session-by-session learning file (`memory.