./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.

Tuesday, 2026-09-15·last updated 2026-09-15 03:13 UTC·9/9 feeds ok

[ hardware ]12

• b10970 — llama.cpp
• Nvidia's RTX 5090 vanishes from online retail in the US — third-party sellers now demand… — r/LL
• b10969 — llama.cpp
• NVIDIA Unveils RTX PRO 5500 "Blackwell" Workstation GPU with 84 GB GDDR7 Memory — r/LL
• Dell Pro 5 16 (AMD) Laptop Review — STH
• b10964 — llama.cpp
• b10956 — llama.cpp
• b10955 — llama.cpp
• b10952 — llama.cpp
• Running Qwen 3.8 next on 16vram+32ram - A useful/fun post for the gpu poors — r/LL
• b10951 — llama.cpp
• b10950 — llama.cpp

[ new models ]10

• I think Muse Glimmer is slept on — r/LL
• DeepSeek engineer relections on RSI - burying my talent to yesterday — r/LL
• Running Qwen3.8-Flash-Next locally on a 12GB VRAM card — r/LL
• Animated transition from AA Intelligence Index v4.1 to v4.3 — r/LL
• v0.4.1 — llama.cpp
• For the GPU poor. K2 Horizon 7B ranks between qwen 3.6 27B and qwen 3.6 35BA3b on the Ar… — r/LL
• UkisAI Swift-Qwen3.8-27B / -58.3% thinking, x1.95 speed while keeping the accuracy of xh… — r/LL
• K2 Horizon lineup is out on AA, and once again AA plots are misleading. — r/LL
• b10964-b10964-b29c606 — llama.cpp
• Is there a better small model than Qwen3.5 4B for a fast local AI assistant? — r/LL

[ local ai ]12

• Will we always have to rely on companies with the funds and resources to give us open mo… — r/LL
• Are there any organizations that are lobbying in favor of open source AI? — r/LL
• Xi promotes open source AI zone among BRICS countries — r/LL
• llama: add Maple 20B-A1B ternary MoE architecture (CPU) by AlexGabbia · Pull Request #27… — r/LL
• What actually makes you trust a local coding agent enough to leave it running unattended… — r/LL
• The new k2 horizon models seem like an absolute beast — r/LL
• Intern-S2-397B (multimodal, reasoning, coding, and scientific agent capabilities) — r/LL
• Interesting Video by Asionometry on the state of SF Chips — r/LL
• What are the current best retail GPUs for max VRAM at a reasonable price? — r/LL
• RTX PRO 5500 Blackwell (84GB) released — r/LL
• CachyLlama archived? — r/LL
• 3D viz of how Deepseek Flash v4.1 is different from a typical decode only transformer — r/LL

[ benchmarks & papers ]1

• Publisher Correction: Quantum neural operators with implicit quadratic frame and express… — NatMI

[ tools ]1

• OpenArch – PyTorch implementations of modern LLM architectures — HN

[ healthcare ai ]12

• Judge blocks Trump administration plans to limit visa length for grad students, postdocs — STAT
• Abbott agrees to settlement over closure of largest baby formula plant in the U.S. — STAT
• Kennedy to deliver keynote for anti-vaccine group he once led — STAT
• STAT+: How $500 explains Republican health policy, and the GLP-1 Rx wave — STAT
• Pennsylvania officials report state’s third death involving measles — STAT
• STAT+: GSK touts lung cancer data from two drugs — STAT
• STAT+: Pharmalittle: We’re reading about another AstraZeneca trial failure, a Novo name… — STAT
• Pennsylvania records third measles-related death amid outbreak — STAT
• STAT+: Definium’s LSD therapy eased anxiety in second large trial — STAT
• STAT+: What’s in a name? Novo looks to a rebrand — STAT
• Opinion: Am I a hypocrite for loving football despite CTE? — STAT
• Opinion: Restoring public trust in science doesn’t have a simple solution. That’s OK — STAT