Latent Intent
2026AI-safety interpretability work toward detecting hidden objectives and scheming, studying how latent deception signal manifests when an LLM's outward behavior is held fixed.

AI-safety interpretability work toward detecting hidden objectives and scheming, studying how latent deception signal manifests when an LLM's outward behavior is held fixed.

Framework for visualizing, decomposing, and analyzing layer-wise latent trajectories during generation.

Modular agentic framework for automation, learning, and research.

Latent-space guidance system for Meta's Coconut model.

Pure Python/PyTorch GPT-2 reimplementation extended with QLoRA personality tuning.

A local Apple Silicon CLI that turns EPUB chapters into MP3s and audiobooks.

Semi-objective conversational insights from exported chats and screenshots.

Full-resolution OpenScreen fork with lower memory use and faster MP4 exports.

Automatically reorders Spotify playlists so similar songs sit next to each other, to avoid akward transitions.

A bunch of miscellaneous small automation scripts and bots.

Natural language to shell commands, years before Claude Code and Codex!

An enhanced personal Japanese dictionary TUI, with automatic Anki card creation and export.

A tiny Android journal for tracking mood, habits, and the correlations you only notice after a few weeks.