Catalogue
Every repository, read
22 repositories, 96,271 characters of README run through Vera itself — the same loaders, the same splitter, the same segmentation it uses on anything else.
Every line below is verbatim. Summaries are the first real sentence of the README; topics are the cores the engine extracted, ranked by mass within that document. Nothing was rewritten to read better.
ARC-AGI2 solver — 84.0% on training (840/1000). Hybrid: 30+ hand-crafted solvers + Claude Sonnet 4.5 program synthesis with deterministic verification. No fine-tuning, no GPU.
A native stereoscopic VR container for Mac and Apple Vision Pro. Bypasses D3DMetal IPC constraints to stream x86 PCVR engines (Source 2) via low-level C++ openvr_api.dll memory hooking and direct visionOS hardware compositing.
BM25 tool search for MCP — find the right tool from 50+ without context collapse. Works with any LLM.
Zero-Trust Enterprise AI IDE for macOS. Gatekeeper Mode ensures 100% security by using Local SLMs to convert raw code into JCross IR, letting Cloud LLMs (Claude/DeepSeek) refactor logic without leaking any proprietary semantics.
Verantyx Pure-Through - AI Memory Refresh System. Prevents context pollution and enables continuous use of clean intelligence.
6次元記憶空間 - Claude AIの能動的記憶操作システム
樹木型空間記憶（Dendritic Memory Space）を編集するためのWebサイト
GitHub Desktop tutorial repository
Paku Paku Eating
Paku Paku Fishing
This application generates 1930s-style newspaper articles based on modern news.
立体十字構造体 (Stereo-Cross Container) — reorganizing LLM knowledge into a shared cross-layer coordinate system with Matryoshka granularity and puzzle-style inference. Pre-registered experiments, personal-scale compute.
This repository contains the inference engine and mathematical proofs for Generative Spatial Compression, a revolutionary technique that shrinks Large Language Models by transforming dense weight matrices into low-rank generative spatial coordinates.
Pure program synthesis solver for ARC-AGI-2 — 74/1000=7.4%, no neural networks, no LLMs
By incorporating a proprietary model into the routing process, we can overcome uncertainties such as those associated with harnesses. The proprietary model thinks in terms of vectors, and another AI translates those vectors.
Context-window problems solved with 35 files. MCP-native spatial memory for any LLM.
On-device long-term memory benchmark - LongMemEval on iOS with llama3.2:1b via MLX
Verantyx-Logic
Verantyx Pure-Through (日本語版) - AI記憶リフレッシュシステム。コンテキスト汚染を防ぎ、クリーンな知能を継続的に使用。
Verantyx Vera α — deterministic, LM-free knowledge & reasoning engine with typed refusal (no hallucination by construction). CLI + MCP server.
Verantyx Vera β — deterministic, LM-free knowledge & reasoning engine + fingerprint-derived leak-attribution watermarking. CLI + MCP server.
Advanced visionOS app for Apple Vision Pro - Attach virtual tools to real-world objects with magnetic snap and tracking