Repos

Shows how paging the embedding table to NVMe lets the 180B-class checkpoint run on a single DGX Spark.

A faithful recreation where every sprite, tile, battle effect and song is generated by code.

Dual-3090 build with 128GB RAM reporting about 4,200 tok/s prefill and 111 tok/s decode on long prompts.

Rooms hold people and bots; mention one to wake it, or let a scoring model choose who replies, with a local daemon running each turn.

A browser action game set in a martial-arts world, with every line of code written by Opus 5.5.

A native VR port built on the Petari decompilation using Opus 5.5 assistance.

A video with a custom GPU brushstroke renderer, made by Opus 5.5 as a visual companion to the interpretability paper.

Open-source project persona-voice: Custom output voices for ChatGPT, Codex, and Grok Bot — local and near real time.. Extends or complements Grok Bot for builders.

Optimised AMD toolchain by a llama.cpp contributor.

Open-source take on Grok Bot where persistent AI teammates each get a cloud computer, powered by Runta.

Serves turboderp's EXL3 quant of the model on one Spark using exllamav3.

A native C++ skateboarding game from one model, shared as an open repository.

Typed-decision server on NVIDIA GPUs or Apple Silicon that speaks the Jev wire format, with coding-agent integrations. Built after a cat woke the author at five on a Sunday.

Step-by-step repo behind JJ Englert's demo: a touchscreen gadget with push-to-talk, spoken answers and an animated face, built on the Muse Gadget SDK for non-coders.

Quantisation and launch recipe focused purely on decode speed at 4-bit.

Intel GPU route with experts streamed from NVMe and 32GB host RAM.

Open-source project awesome-grok-bot: 3387 live x.ai/bot shares for Grok Bot — curated catalog with shelf tiers, searchable Pages, bilingual README. Not a pro. Extends or complements Grok Bot for builders.

Runs Cloudflare's clef-flash decision model on your webcam stream locally, answering typed questions about each frame. The most-starred community build of the week.

Open-source take on named AI teammates with desktop, web and mobile clients and per-bot tools, instructions and memory.

A bespoke inference engine for the Qwen4-style architecture that runs Flash-Next on a 16GB card.

Hand-tuned ROCm inference path for AMD RDNA4 cards running the 125B model.

Open-source project Hermes-Mobile-App: A mobile app for Hermes, built around Bots! Similar to Grok Bot but completely Open Source and runs on your own computer. Extends or complements Grok Bot for builders.