Repos

An experiment that pushes the Qwen3.8 models onto the Neural Engine of Apple Silicon rather than the GPU. A long-shot attempt to cut power use for local inference.

Neel Nanda's project, a song with a code-drawn video telling the story of mechanistic interpretability.

Early notes on fitting the 126GiB checkpoint onto one DGX Spark.

Build and run notes for the AtomicChat 4.27-bpw quant with multi-token prediction on a mid-range gaming PC.

OMLX setup using prompt-lookup MTP drafts that speed edit turns by 1.3 to 2.2x with exact output. By a privacy educator who wants coding models that stay on his own machine.

Runs the 177B build across RTX 5060 Ti and 4070 Super cards.

Local MCP server that gives AI agents structured decisions with probabilities from Clef-Flash through llama.cpp, with no network needed.

Live gallery of motion graphics made by Opus 5.5, Kimi K3, Fable 5 and Astra shown next to the prompts that produced them.

Reference-driven study of the public bot mark animation, with a visual timeline editor and a framework-free Web Component.

Open-source Claude Code mod by danyuchn. Its repo describes it as: Claude Code mods: screen-guard masks names and secrets while you screen-share; cache-panel reminds you before the prompt cache goes cold. Install via one plugin marketplace. Shows what the community built in the…

Delegates research and test runs to the Muse CLI so Claude Code spends fewer tokens on routine work.

One prompt installs a panel to add API or ChatGPT-login providers and switch with /gs commands across three protocols.

Skill pack covering list research, cold email, parallel LinkedIn, reply handling and meeting logs with Instantly and HeyReach CLIs.

Open-source Claude Code mod by omarcevi. Its repo describes it as: Community Claude mods, plugins & skills, installable from one marketplace. Shows what the community built in the first days of mods.

Experimental one-box DGX Spark deployment.

Verified guide for running the model on a high-memory Mac.

262K context with vision and FP8 KV cache on four used GPUs.

Extends Perplexity's open Lily inference engine so it can run Qwen3.8-Flash-Next on M5 Macs. Shows how a community member adopted a corporate engine for the new model.

Tutorials, prompts and modelling code for turning Realsee 3D scan exports into editable Blender spaces with Astra.

A stability-focused recipe for a single DGX Spark with a KDA decode kernel and EAGLE speculation, fixing KV-loop failures others hit.

Fits the model by quantising the embedding table too.

Community measurements of memory needs at different quants and contexts.

Showcases long-context concurrent serving on used cards.

Written the day Amazon blocked agent shopping traffic: a local API that does product search and add-to-cart for personal and business carts so Muse can keep shopping.