./daily — ← back to field notes

Daily links

One-line pointers to today's local-AI, model, hardware, and healthcare-AI news. No commentary — just the headline, the link, and where it came from.

Sunday, 2026-09-13·last updated 2026-09-13 23:04 UTC·9/9 feeds ok

[ hardware ]12

• Infineon RISC-V for Automotive at Hot Chips 2026 — STH
• CUDA-for-AMD-Windows: Run CUDA-targeted Windows applications on AMD GPUs with ZLUDA + RO… — r/LL
• b10948 — llama.cpp
• b10947 — llama.cpp
• b10946 — llama.cpp
• 3k$ 128GB VRAM + 256GB RAM DDR4 Server — r/LL
• My experience building 64GB VRAM AI SWE assistant/agent PC — r/LL
• b10941 — llama.cpp
• b10938 — llama.cpp
• b10937 — llama.cpp
• b10936 — llama.cpp
• b10935 — llama.cpp

[ new models ]9

• I have just moved from MacBook M5 pro 48 GB to RTX3090 — r/LL
• Dear 24G owners, try VLLM you might be able to run Qwen3.8 27B INT4, 144K FP8 KV on RTX… — r/LL
• Hoping for Optimized Smarter Upcoming Models .... Like DeepSeek-V4.1-Flash( KVCache + En… — r/LL
• Qwen 3.8 27B UD-IQ4_XS even faster on 16GB CUDA — r/LL
• Qwen3.8 flash next - untrained svg generation — r/LL
• internlm/Intern-S2 · Hugging Face — r/LL
• How does Qwen 3.8 27B compare on low thinking mode to the older 3.6 models? — r/LL
• What's the Story with Agnes-3.0-Flash? — r/LL
• Rebuilding AUTOMATIC1111 with Gradio Workflow — HF

[ local ai ]12

• Are there any experimental small models (9B or less) which are out or currently being tr… — r/LL
• Plot twist — r/LL
• Aurora1.0-150M Releases! — r/LL
• Talk me out of buying a 3rd Spark — r/LL
• Migration from Claude Code to a private local harness. Questions. — r/LL
• The rhetoric is really heating up! — r/LL
• What are the top AI Models that are still relevant today from 2024 and 2025? — r/LL
• The Local LLM community feels like the golden era of the internet all over again — r/LL
• Is there still strong interest in a dense 9b model? — r/LL
• Should I sell my RTX 5090 for a Mac Studio M5 Ultra 96GB? — r/LL
• huggingface_hub silently fingerprints which AI coding agent you're using and sends it as… — r/LL
• DS 4.1 and the new Harness — r/LL

[ benchmarks & papers ]1

• A collaborative agent with two lightweight synergistic models for autonomous crystal mat… — NatMI

[ tools ]2

• b10944 — llama.cpp
• Training a 3.8B LLM to 0.384 CORE for $998 — HN

[ healthcare ai ]3

• STAT+: Trump’s secretive pharma deals may undermine ‘most-favored nation’ pricing, an an… — STAT
• Healthcare AI's next test is integration — HN-HC
• Inside The Medical AI Factory — HN-MED