Tech

Running Lemonade on Strix Halo (EVO-X2): Vulkan Shared Memory Leaks and ROCm Stability

Real-world testing of AMD Lemonade v10.0.1 on Ryzen AI Max+ 395. LLM, image generation, speech recognition, and TTS running simultaneously, NPU Hybrid execution, Vulkan vs ROCm benchmarks, and discovering shared memory leaks.

AMD's Lemonade Local AI Server Bundles GPU, NPU, and Multi-Modal Inference Under One Roof

Lemonade is AMD's open-source local AI server that manages multiple backends like llama.cpp and FastFlowLM across GPU/NPU/CPU, serving text, image, and audio generation through an OpenAI-compatible API.

Google's Gemma 4 launches in four sizes (E2B–A4B), publishing Gemini 3–derived reasoning under Apache 2.0

Google DeepMind has released Gemma 4: four models—31B dense, 26B MoE (A4B), E4B, and E2B—with a 256K context, multimodal input, tool calling, and support for 140 languages.

SwiftLM is a Swift-based LLM inference server that integrates TurboQuant and SSD streaming into Metal shaders

SwiftLM, an Apple Silicon–only MLX inference server, provides a native Metal implementation of TurboQuant V2+V3 hybrid KV‑cache compression and NVMe SSD expert streaming.

Automatically Decomposing an Anime Character Illustration into 23 Layers with See-through and Exporting as PSD

Automatically decomposing a single anime illustration into front hair, back hair, clothes, and other layers with inpaint completion of hidden areas. Testing the LayerDiff + Marigold-based implementation.

Adobe Creative Cloud is rewriting the hosts file without permission

Adobe CC's WAM component silently adds a detect-ccd.creativecloud.adobe.com entry to the Windows hosts file and uses it to detect installations from the browser. A breakdown of the mechanism and the broader pattern of major software taking control away from the OS and the user.

Claude Code full source exposed via npm; OpenAI Codex token-theft flaw disclosed at the same time

A summary of how source maps bundled in the Claude Code npm package made over 510k lines of TypeScript visible, and how a branch-name command injection in OpenAI Codex could have allowed theft of GitHub tokens.

Ollama Moves to MLX Backend, Dramatically Speeds Up Local Inference on Apple Silicon

Ollama 0.19 switches the Apple Silicon backend to MLX, achieving 1,810 tokens/s prefill and 112 tokens/s decode. NVFP4 quantization support and cache improvements landed at the same time.

Scaling Qwen3.5-35B-A3B from 4K to 65K Context with Only 800MB Extra VRAM

Qwen3.5-35B-A3B is an SSM+Attention hybrid where only 10 of 40 layers use KV cache. Expanding ctx-size from 4096 to 65536 on llama-server added just 800MB VRAM with zero speed loss. Includes q8_0 KV quantization benchmarks and TurboQuant status.

Radeon 8060S (gfx1151) Vulkan Broke Again After AMD Driver Update

After updating to AMD Software 26.3.1 on a GMKtec EVO-X2 (Ryzen AI Max+ 395), Vulkan backend fails to allocate device memory properly and falls back to CPU. Investigation and workaround by changing BIOS VRAM allocation from 48GB/16GB to 32GB/32GB.

Tracking Down Why Qwen Image Edit Started Taking 10 Minutes After a ComfyUI Update

Diagnosing the speed regression caused by MPS BF16 being 2x slower than FP16, combined with an FP16 Attention bug — and the fix.

Hypura’s NVMe Streaming and TurboQuant’s KV Cache Quantization

Hypura breaks away from llama.cpp’s mmap design and streams even dense models with a three-tier NVMe placement, while TurboQuant eliminates quantization-constant overhead via a polar-coordinate transform. Includes a design comparison with Flash‑MoE and a review of scenarios where KV‑cache compression actually helps.