Daniel

hive. Unified agent memory and context compression stack for 2026 NVIDIA + edge (Vera CPU, Grace, Jetson Thor, 3090). Glues busyBee-cpu, honey-comb, and rust-brain. Better effective reasoning per token.

Rust-Brain. Rust CLI for RBMEM (.rbmem), a structured Rust-Brain memory format with timestamp-protected sections, hierarchy, graph relations, Hermes integration, and Markdown sync.

busyBee-cpu. Python

busybeaver. A 2MB sklearn classifier on consumer CPU beats Cohere's $10M+ 218B MoE on HumanEval

RBForge. Python

honey-comb. CPU-only inline context compression for agent harnesses. Keep the honey, drop the wax.

BusyBeaver-Hermes-Adapter. Python

moe-routing-atlas. Open-source toolkit for mapping and visualizing Mixture-of-Experts model routing patterns

SABER. Spectral Analysis-Based Entanglement Resolution — ablation pipeline for removing refusal directions from LLMs

llama.cpp. llama.cpp fork with a patch for RYS-duplicated Qwen3.5/Qwen3Next models (non-uniform full_attention pattern). See rys-qwen35 branch.

TALOS-trace-curator. Talos trace-curator: Hermes Skill that turns raw agent sessions into training datasets

reap-eval. A reusable, model-agnostic methodology for evaluating pruned/compressed Mixture-of-Experts (MoE) model releases: rubric, red-flag catalog, runnable behavioral tests, scorecard + producer checklists, and worked examples.

t013-logical-fallacies. T013: Reasoning Depth Assessment - Logical Fallacies. A psychometric benchmark for evaluating AI agent reasoning quality using Item Response Theory (IRT) principles.

MPKnet. Bio-inspired CNN modeling M/P/K pathways of the Lateral Geniculate Nucleus. From 76KB/15.5K params (33 FPS on Raspberry Pi) to full-scale and temporal variants. Matches ResNet18 with 52x fewer parameters. No pretraining required.

ai-gym-tower-defense. Tower defense game with a Gymnasium API for training AI agents (RL, heuristic, tournament).

autolora. A Hermes skill that fine-tunes a new version of the agent on its own traces and hot-swaps it in: AUC trace curation, RunPod+Unsloth LoRA, Stripe-paid under a cap, user-set benchmark thresholds.