lxe
Formerly @OpenAI, @Netflix, @opendoor-labs, @uber
Built from lxe's public GitHub profile, San Francisco, CA. Not on Take My Repo yet, so there are no posts or updates here.
See and operate an unmodified iPhone from a Linux PC: AirPlay mirror in, Bluetooth HID mouse/keyboard out, plus a CLI and skill for agents.
DeepSeek-V4.1-Flash-1017 (552B MoE + 196B Engram) on 2x RTX 3090 + 128 GB RAM. Rust + CUDA, T16 quantization. 40+ tok/s measured with expert deferral + FP8 projections.
LoRA weights for Cerebras-GPT-2.7b finetuned on Alpaca dataset with shorter prompt
Allows LLM agents to control a local chrome instance without taking screenshots
Mobile web app for audio "push-to-talk" + TTS chat interface with OpenAI-like APIs
Batched DeepSeek-V4.1-Flash inference on 8x V100 (sm_70): low-bit lookup formats, split experts, NVLink collectives
Krea 2 + MiniMax H3 R2V workflow, prompts, seeds, and reference assets for the Skythread short film
A ComfyUI custom node for integrating OpenAI-compatible LLM APIs with support for both text-only and multimodal (text + image) interactions. Works with OpenAI GPT models, local Ollama/LLaVA models, and any OpenAI-compatible endpoint.
Playground LLM chatbot app using stable-lm (but you can use any model)
Qwen3.5 inference server on hand-written CUDA kernels with continuous batching
Experimental ComfyUI nodes and frontend extensions developed for LXE workflows.
Clear, connected prose for AI agents: a portable skill with contextual examples and reversible Codex / Claude Code hooks.
Do you need Claude Opus to teach, grade and judge? Six code-graded tasks, eleven models, every answer published.
Rust toolkit for inference fleet simulation, scheduling, routing, and validation