./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.

Saturday, 2026-10-24·last updated 2026-10-24 20:38 EDT·9/9 feeds ok

[ hardware ]12

• CMP170HX inspiration unlocks 30x/40x also, up to 2X Prefill speedup --> NVIDIA RTX 30 &… — r/LL
• b11869 — llama.cpp
• b11868 — llama.cpp
• b11867 — llama.cpp
• b11866 — llama.cpp
• b11865 — llama.cpp
• b11861 — llama.cpp
• b11860 — llama.cpp
• ADT R27A-BK3 U.2 to PCIe Adapter Review — STH
• b11858 — llama.cpp
• b11857 — llama.cpp
• 10 tok/s Deepseek V4.1 Flash 8K with Strata on a 5090 with 64GB DDR5 Ram — r/LL

[ new models ]3

• Are there any good finetunes for Blender and/or 3D modeling? — r/LL
• Meta trained small AI models to read raw bytes instead of tokens. They start out worse… — r/LL
• LitPivot: Developing research ideas through dynamic contextualization within the literat… — HF

[ local ai ]12

• Is it just me or Kimi K3.1 Pro don’t look much better than DeepSeek 4.1 Flash? — r/LL
• UK to invest £100m in 'world first' AI Biosecurity Lab — r/LL
• Small model good for reverse engineering that can run on a 3090 and 32GB RAM. — r/LL
• High-Performance llama.cpp fork for Legacy GPUs: Kepler Support, CUDA Graphs & Windows N… — r/LL
• Made a rhythm game where 4 small neural nets listen to your song right in the browser ta… — r/LL
• Raon is better than Flint. — r/LL
• RecursiveSI Flint Benchmark: unsloth/flint-GGUF flint-UD-Q3_K_XL.gguf — r/LL
• Silero VAD v6 being used by Chatgpt voice mode — r/LL
• Just update of ik_llama.cpp fork and seek of feedback — r/LL
• Kimi K3.1 removes cyberattack training data from open weight release — r/LL
• big cloud models like opus 5.5 get all the motion graphics praise, but i made this with… — r/LL
• AbliteratedOutTheBox Strata Announcement — r/LL

[ benchmarks & papers ]4

• M6 Mac Mini owners (32GB): now you can run Qwen3.8-27B in high quality on the Apple Neur… — r/LL
• KLD is misunderstood — r/LL
• Guidelines for evaluating structure-based methods for protein–ligand binding prediction — NatMI
• Efficient equivariant framework for universal charge density prediction — NatMI

[ tools ]2

• Spark X2.6 4B is the best of it's size — r/LL
• Using LLM for confidence scoring is a bad idea — HN

[ healthcare ai ]12

• STAT+: The ESMO presentation everyone’s been waiting for — STAT
• Opinion: Christina Applegate puts a face on one of the most consequential challenges in… — STAT
• Opinion: STAT readers on Elizabeth Holmes, cancer screening, and more — STAT
• STAT+: Highly anticipated Moderna-Merck melanoma cancer vaccine data arrived early — STAT
• STAT+: Diabetes association says ‘threat’ preceded members’ ejections, but admits polici… — STAT
• STAT+: Patient dies in Bristol Myers Squibb autoimmune CAR-T trial — STAT
• STAT+: Most cancer drugs approved by European regulators lack a high clinical benefit, a… — STAT
• STAT+: Novartis multiple sclerosis treatment cuts rates of relapse in pivotal trials — STAT
• Opinion: It took me 13 years to be diagnosed with MS. Christina Applegate’s misdiagnosis… — STAT
• STAT+: Up and down the ladder: The latest comings and goings — STAT
• Remembering STAT’s Helen Branswell — STAT
• STAT+: Tango builds on Revolution Medicines’ pancreatic cancer success — STAT