./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.

Monday, 2026-10-12·last updated 2026-10-12 23:06 UTC·8/9 feeds ok (HN)

[ hardware ]12

• b11577 — llama.cpp
• b11576 — llama.cpp
• b11575 — llama.cpp
• Intel’s Crescent Island AI Server GPU Gets a Name and Specs: Meet the Intel GPU D-Series — STH
• Marvell Structera X 3406 and 3506 CXL 3.2 Memory Expanders — STH
• b11574 — llama.cpp
• b11573 — llama.cpp
• b11572 — llama.cpp
• b11571 — llama.cpp
• b11570 — llama.cpp
• graphsignal-top: a `top` for inference GPUs in the open-source Graphsignal profiler, ker… — r/LL
• b11569 — llama.cpp

[ new models ]8

• Qwen3.8-27B vs Frontier vs Others — r/LL
• Qwen released QwenGyre - RL framework for training AI agents on extremely long complex t… — r/LL
• b11568: examples : fix draft sampling and rejection handling (#30095) — llama.cpp
• I made an open-source AI leaderboard that shows the best models your machine can run — r/LL
• QLoRA on ExLlamaV3 Quants - Train @ 2-8bpw — r/LL
• Decision 3.0: open, multimodal decision models from 0.6B to 27B. Ranks #1 on both Jev De… — r/LL
• strix-llama new branch: prompt processing +21–54% over Vulkan and up to 2.7× over stock… — r/LL
• Impactful scheduling for GPU clusters — HF

[ local ai ]12

• Once again anthropic doing free marketing for open weight models — r/LL
• It's been 2 months — r/LL
• openbmb/MiniCPM-V-4.7-1B — r/LL
• openbmb/MiniCPM-V-4.7-35B-A3B — r/LL
• In 2030 owning a gpu cluster will be seen the same as owning a gun for home defense — r/LL
• Boole-1.7B: A 1-bit Qwen3-1.7B model that outperforms Bonsai-1.7B — r/LL
• BrokenML: can agents fix broken ML jobs? — r/LL
• NVIDIA RTX 5080 24GB reportedly launching in Q1 2027, board partners still waiting for d… — r/LL
• A "free" multi-modal "embedding" model (already) included with your LLM: vLLM plugin + u… — r/LL
• Any takers for local llm? NVIDIA’s CUDA Moat Is Accelerating RTX Spark Laptop Sales, Hig… — r/LL
• Building strata from source — r/LL
• Receipt scanner that runs a fine-tuned 0.8B vision model on the phone (free) - detailed… — r/LL

[ benchmarks & papers ]2

• Emotional optimization by newsroom AI needs behavioural evaluation — NatMI
• Characterizing the role of reasoning tokens in large language models as state over token… — NatMI

[ tools ]3

• LLC: lightweight Open WebUI alt now with Agents+MCP+RAG+Image Editing — r/LL
• LSearch - Spotlight for windows with embedding — r/LL
• VectorDash: A full Windows search replacement incorporating Embeddinggemma-2 — r/LL

[ healthcare ai ]3

• Americans trust science more than peers around the world, survey shows — STAT
• Healthcare AI needs a different model-development flywheel — HN-HC
• Inside The Medical AI Factory — HN-MED