Field
AI and machine learning
Posts
- Gemma 4 26B-A4B at NVFP4: a mixture-of-experts recipe for a DGX Spark
How I serve Gemma 4 26B-A4B, a mixture-of-experts model at NVFP4, on one DGX Spark with a DFlash drafter: the production recipe, the bench recipe behind 816 tok/s at c64, and the drafter defects that made my earlier numbers wrong.
The Aeon Forge10 min read - Gemma 4 12B AEON Abliterated on a Mac: the MLX toolkit and server
Two MLX builds of Gemma 4 12B AEON Abliterated for Apple Silicon, 13.4 GB and 9.3 GB, served through an OpenAI-compatible API by one pasted uv command: what an M4 Pro measures, how images and speech go in, and why the quantization recipe is mixed.
The Aeon Forge8 min read - DeepSeek-V4 on two DGX Sparks: an experimental TP=2 image
An experimental vLLM image that serves DeepSeek-V4-Flash across two DGX Sparks at tensor parallel 2 over RoCE: the seven GB10 gaps it patches, why PIECEWISE graphs are the rule, what the numbers do and do not prove, and a two-node quick start.
The Aeon Forge7 min read - DDTree on vLLM: candid research notes on tree speculation for Qwen3.6
DDTree puts tree-shaped speculative decoding into vLLM for Qwen3.6 on a DGX Spark. These lab notes show what works, what breaks inside a hybrid attention model, the numbers so far, and why production should stay on DFlash.
The Aeon Forge8 min read - Create agentic personas: a voice, a knowledge base and a phone line
How I compose an agentic persona that chats in Matrix, speaks in a Qwen3-TTS voice, answers from a local RAG corpus and picks up WebRTC calls, scaffolded by one secret-free command, plus the agent skill that lets an OpenClaw agent build the next one.
The Aeon Forge7 min read - ComfyUI for the DGX Spark: the AEON media production base
A Docker build of ComfyUI for the DGX Spark with CUDA 13, SageAttention v3 for sm_121a and NVFP4 weights, pulling Flux 2, LTX 2.3 and ACE-Step on first start. What broke, what the flags mean, how local and remote modes work, and why three AEON media repos sit on it.
The Aeon Forge10 min read - AEON vLLM Ultimate RTX: vLLM for the RTX 5090 and Blackwell sm_120
A slim vLLM image for the RTX 5090 and RTX PRO 6000 Blackwell, why it must never be crossed with the DGX Spark image, what changes on a 32 GB card, four serving recipes for a 27B NVFP4 model, and the WSL2 fault that passed every test.
The Aeon Forge9 min read - AEON vLLM Ultimate for ROCm: early access on vLLM 0.31
AEON vLLM Ultimate for ROCm is a vLLM 0.31 image that serves Qwen3.8-27B AEON ULTIMATE in MXFP4 and MXFP6 on AMD GPUs. It covers what is built, why the stock path decoded at 1.7 tok/s, what is still unmeasured, and how to test the 2x R9700 recipe.
The Aeon Forge6 min read - AEON vLLM Ultimate: one serving image for the DGX Spark
AEON vLLM Ultimate is vLLM built from source for the DGX Spark's GB10 and shipped as one 18.3 GB image that serves the whole AEON fleet with speculative decoding and NVFP4 weights. Here is why it exists, what broke on the way, and how to pull and run it.
The Aeon Forge14 min read - AEON Radio Drama: dialogue, music and effects in one command
AEON Radio Drama turns one JSON script into a finished, loudness-normalized audio-fiction master: Qwen3-TTS voices locked per character, an ACE Step score, duration-routed sound effects and a sidechain-ducked mix, all through a ComfyUI you already run.
The Aeon Forge8 min read - AEON Music Maker: ACE Step 1.5 music with a mastering chain
AEON Music Maker turns a text prompt into an ACE Step 1.5 track through ComfyUI and masters it to a 48 kHz FLAC without a compressor. The five model variants, the chain that squashed my test tracks, the one that works, and the agent files.
The Aeon Forge8 min read - AEON Movie Maker: cinematic video with LTX 2.3 22B
AEON Movie Maker turns a JSON screenplay into a short film with LTX 2.3 22B through ComfyUI: last-frame carry-forward and per-character seeds for continuity, a sidechain-ducked final mix, the LoRA strengths, the saturation checklist and the one remote-mode limit.
The Aeon Forge8 min read - Aeon Magick Orb: a Raspberry Pi that gives an AI agent eyes and hands
The Aeon Magick Orb is a Raspberry Pi that plugs into a computer's HDMI and USB ports so an AI agent can see the screen and type, click and reboot, with nothing installed on the target. How it works, what broke on the way, and how to build one on a Pi 4 or Pi 5.
The Aeon Forge11 min read - AEON Bench recipe lab: tuning vLLM flags one at a time on a DGX Spark
The AEON Bench recipe lab treats a serve recipe as an experiment: change one flag, bench, read the curve. The knobs in order of effect, the speculative decoding trap behind 33 versus 517 tok/s on a DGX Spark, the 816 tok/s recipe to steal, and four traps that fake a result.
The Aeon Forge11 min read