Notes from the practice
Engineering
in the open.
Architecture, infrastructure, creative development, and the details that make a system work.
All categories & topics
87 articles found
Llama 4 vs Mistral Large 2: Architecture, Licensing and Deployment Tradeoffs
Correctly compare Llama 4 mixture-of-experts models with dense Mistral Large 2, including active parameters, memory planning and release-specific licensing.
Local LLMs: Choosing a Runtime and Estimating Memory
Plan local model inference with realistic weight-memory estimates, runtime support, context costs and a verified data boundary.
Investigating Rising Memory Use in Long-Running Node.js
Separate Node.js heap usage, external allocations and RSS, correct the V8 diagnostics API, and investigate allocator fragmentation without relying on a false ratio.
Midjourney V7 and GPT Image: Comparing Image Generation Workflows
Compare documented image-generation workflows while preserving exact typography, authentic project evidence and realistic expectations for editing.
Modular Monoliths: Enforcing Boundaries in One Application
Define module ownership, public interfaces, data access and dependency checks without treating deployment size or team headcount as architecture rules.
Multi-Cloud: Define the Requirement Before Adding a Provider
Evaluate portability, recovery and provider specialization with explicit data, identity, networking and operating boundaries.