Log Records
Field notes on local AI, system architecture, and making complex things work.
Strata on a Mac Studio M5 Ultra (96 GB)
Strata first claimed Qwen3.8-Flash-Next, a 125B model, could run on a machine with just 12 GB of VRAM. My new Mac Studio M5 Ultra with 96 GB just arrived, so I turned it into a headless local AI inference machine.
- Tag
- /llm /local-ai
- Date
- Read
- 9 min read
Dense vs. MoE Models Explained
Dense models reuse the same network for every token. MoE models choose a few experts from a larger pool. Qwen3.8-27B and Qwen3.8-Flash-Next show why less work per token can still require much more memory.
- Tag
- /llm /local-ai
- Date
- Read
- 9 min read
Prefill vs. Decode
Prefill, or prompt processing (PP), reads the input. Decode, or token generation (TG), writes the output. GPU compute matters for ingesting repository context; memory bandwidth helps explain generation speed.
- Tag
- local-ai
- Date
- Read
- 6 min read
Local Coding Models by Memory
From 4 GB to 512 GB, the right local coding model is the one that leaves enough memory for context, KV cache, and the runtime—not the largest file that barely loads.
- Tag
- /llm /local-ai
- Date
- Read
- 5 min read
LLM Quantization Explained
What BF16, FP8, Q4, and GGUF mean, how quantization saves memory, and what to check before choosing a local model.
- Tag
- /llm /local-ai
- Date
- Read
- 5 min read
Local Agentic AI Machines for Qwen3.8-27B
A practical look at local agentic AI hardware for Qwen3.8-27B, from Apple Silicon and the RTX 5090 to DGX Spark, Intel Arc Pro, and AMD Radeon AI PRO.
- Tag
- /local-ai
- Date
- Read
- 11 min read
One Mile, Part 2: Scope Setup
Part two of my one-mile shooting journey, documenting scope setup, rail cant, eye relief, and alignment.
- Tag
- /precision-shooting
- Date
- Read
- 7 min read
One Mile, Part 1: The Build
The beginning of my journey into long-range shooting, covering the rifle, optics, and accessories I chose for a one-mile goal.
- Tag
- /precision-shooting
- Date
- Read
- 9 min read
MRAD vs MOA for Long Range
My comparison of MRAD and MOA, from angular measurements and range corrections to communication in precision rifle competition.
- Tag
- /precision-shooting
- Date
- Read
- 4 min read
AKS Part 3: Cloudflare TLS
Configuring Cloudflare and Azure Application Gateway for end-to-end TLS, with origin certificates and Azure Key Vault integration.
- Tag
- /kubernetes
- Date
- Read
- 4 min read
AKS Part 2: Key Vault Secrets
Integrating Azure Key Vault with AKS through the Secrets Store CSI Driver, managed identities, and automatic secret rotation.
- Tag
- /kubernetes
- Date
- Read
- 8 min read
AKS Part 1: Private Networking
The first part of my Azure Kubernetes Service series: deploying a private cluster with Azure CNI networking and exposing an application through Application Gateway.
- Tag
- /kubernetes
- Date
- Read
- 9 min read