- London, United Kingdom
-
22:32
(UTC) - https://www.danieldemmel.me
- @dain@hachyderm.io
- in/danieldemmel
Stars
Complete logs of the experiments in the blog post 'LLMs can understand and write neural network weights'
Training-free language model modification: Remove censorship, suppress slop, reduce sycophancy, and more
Headless Blender-compatible 3D engine in Zig, driven by LLM agents over MCP
A python library built on top of Inspect AI to support Control research and evaluations, by Redwood and EquiStamp
Training-free feature discovery for LLM activations via streaming exemplar partitioning.
1M param OCR model - runs on CPU, 15x faster than tesseract, close to 1B+ param models in accuracy
Spotify, native and fast. One lightweight Rust app for your whole library, local playback, and Spotify Connect on Linux, macOS, and Windows.
GPU checkpoint/restore for vLLM: snapshot a warmed-up engine (compiled kernels, CUDA graphs, weights, 1-8 GPUs) into an OCI image and restore it in seconds instead of a multi-minute cold start. Scaβ¦
Code-rendered music video for "I'm Upping My P(doom)"
Turn any LLM into a Jev-style decision model: typed decisions, real probabilities, no training. (continue updating, welcome any issue and PR request)
DS Menu Bar - A Lightweight Menu Bar Control for DwarfStar Server (ds4-server)
A local inference engine for Apple silicon, built around the model.
VectorCraft: An open-source, vector art tool written by scratch in pure Rust
PhotoCraft: An open-source, photo editor written by scratch in pure Rust
A secure Python sandbox, written in Rust, for code written by AI.
Collection of evals for Inspect AI
Every Eval Ever is a shared schema and crowdsourced eval database. It defines a standardized metadata format for storing AI evaluation results β from leaderboard scrapes and research papers to locaβ¦
A framework for testing and evaluating AI agents across various task domains, designed for misalignment interpretability research.
Resources for my pure data, GEM and alien audio visual workshop
Which open-weight LLMs actually run well on which Apple silicon, and through which engine. Updated daily.
Feature-packed native macOS screenshot & recording tool: annotate, auto-redact PII, record GIFs, OCR + translate, scroll capture, beautify, and more. No Electron, no subscription.
macOS menu bar app to run and monitor local DeepSeek V4 / ds4-server β tray metrics, fans, and one-click Pi/Codex/Cursor launches, embedded DSH agent window.
Self-maintaining, Obsidian-compatible knowledge base for pi β turn raw sources into an interlinked wiki that compounds. Native Open Knowledge Format (OKF) v0.2.
[ICLR 2026] Code for the paper "Graph-of-Agents: A Graph-based Framework for Multi-Agent LLM Collaboration"