Worth your attention
Recent research is pending.
No validated story met this window. There is no quota to fill.Last week in AI / ML / Strix
Edition archive →Lemonade v2026.40.0: AMD iGPU Streaming and Agent Updates
This edition covers the v2026.40.0 release, highlighting new AMD integrated GPU support for model streaming and the addition of the Junie launch agent. It also notes temporary removals of specific ROCm components due to ongoing bug fixes.
Published late. Short evidence-backed edition.
- AMD Integrated GPU Support
- Agent and Component Changes
Explore the knowledge base
Machine & Omarchy
Hardware, unified memory, kernel, drivers, Mesa, firmware, storage, power, cooling and suspend.
Unified memory
Research backlog: evidence is being gathered.
Evidence is being gathered for this branch.
Kernel
Research backlog: evidence is being gathered.
Evidence is being gathered for this branch.
Mesa
Research backlog: evidence is being gathered.
Evidence is being gathered for this branch.
Omarchy
Research backlog: evidence is being gathered.
Evidence is being gathered for this branch.
Storage
Research backlog: evidence is being gathered.
Evidence is being gathered for this branch.
Power & suspend
Research backlog: evidence is being gathered.
Evidence is being gathered for this branch.
Inference
llama.cpp, ROCm/HIP, Vulkan, serving, batching, concurrency, context, KV cache, attention, quantization, speculative decoding, MoE and model fit.
- Neural Network Driven Quantization Aware Optimization for Low Latency Large Language Model Inference
- AMD ROCm 10.0.0 Software Stack Overview
- llama.cpp Backend Documentation
llama.cpp
Research backlog: evidence is being gathered.
Evidence is being gathered for this branch.
Vulkan
Research backlog: evidence is being gathered.
Evidence is being gathered for this branch.
ROCm & HIP
Research backlog: evidence is being gathered.
Evidence is being gathered for this branch.
KV cache
Research backlog: evidence is being gathered.
Evidence is being gathered for this branch.
Quantization
Research backlog: evidence is being gathered.
Evidence is being gathered for this branch.
MoE
Research backlog: evidence is being gathered.
Evidence is being gathered for this branch.
Models
Language, code, vision, image/video, speech, embeddings, rerankers, licenses and releases.
- Hugging Face Diffusers Documentation Scope
- OpenAI Whisper Documentation Overview
- Qwen3.8-27B Model Card Architecture
Qwen
Research backlog: evidence is being gathered.
Evidence is being gathered for this branch.
Vision
Research backlog: evidence is being gathered.
Evidence is being gathered for this branch.
Speech
Research backlog: evidence is being gathered.
Evidence is being gathered for this branch.
Licenses
Research backlog: evidence is being gathered.
Evidence is being gathered for this branch.
Agents & coding
Tool use, structured output, MCP, orchestration, sandboxing and coding workflows.
Tool use
Research backlog: evidence is being gathered.
Evidence is being gathered for this branch.
MCP
Research backlog: evidence is being gathered.
Evidence is being gathered for this branch.
Structured output
Research backlog: evidence is being gathered.
Evidence is being gathered for this branch.
Knowledge systems
Ingestion, chunking, hybrid retrieval, embeddings, reranking, OCR, citations and graphs.
Citations
Research backlog: evidence is being gathered.
Evidence is being gathered for this branch.
Retrieval
Research backlog: evidence is being gathered.
Evidence is being gathered for this branch.
Document ingestion
Research backlog: evidence is being gathered.
Evidence is being gathered for this branch.
Training & adaptation
LoRA/QLoRA, datasets, synthetic examples, feasibility, evaluation and regression.
LoRA
Research backlog: evidence is being gathered.
Evidence is being gathered for this branch.
Evaluation
Research backlog: evidence is being gathered.
Evidence is being gathered for this branch.
CPU / GPU / NPU
Measured support, XRT/amdxdna, Lemonade, FastFlowLM, throughput, bandwidth and offload.
- Lemonade v2026.40.0 Introduces AMD Integrated GPU Streaming
- FastFlowLM: NPU-Optimized Runtime for Ryzen AI
- AMD NPU Kernel Documentation Overview
- Lemonade Server Documentation: Kernel Requirement for Strix Halo
Lemonade
Research backlog: evidence is being gathered.
Evidence is being gathered for this branch.
amdxdna
Research backlog: evidence is being gathered.
Evidence is being gathered for this branch.
FastFlowLM
Research backlog: evidence is being gathered.
Evidence is being gathered for this branch.
Reproducible experiments
Protocols, prompts, hashes, runtime versions, thermal conditions and unsuccessful attempts.
Benchmark protocols
Research backlog: evidence is being gathered.
Evidence is being gathered for this branch.
AI / ML developments
Papers, methods, releases and practical or foundational connections.
Papers
Research backlog: evidence is being gathered.
Evidence is being gathered for this branch.
Releases
Research backlog: evidence is being gathered.
Evidence is being gathered for this branch.
- LoRA Review for Domain Transfer
- Hugging Face Diffusers Documentation Scope
- OpenAI Whisper Documentation Overview
- EleutherAI lm-evaluation-harness Overview
- LoRA Mechanism in PEFT Documentation
- Sentence Transformers Semantic Search Documentation
- MCP Tool Schema and Invocation Scope
- Neural Network Driven Quantization Aware Optimization for Low Latency Large Language Model Inference
- Lemonade v2026.40.0 Introduces AMD Integrated GPU Streaming
- FastFlowLM: NPU-Optimized Runtime for Ryzen AI
- AMD ROCm 10.0.0 Software Stack Overview
- AMD NPU Kernel Documentation Overview
- Lemonade Server Documentation: Kernel Requirement for Strix Halo
- Qwen3.8-27B Model Card Architecture
- Mesa Documentation Overview
- llama.cpp Backend Documentation
- Lemonade v2026.40.0: AMD iGPU Streaming and Agent Updates