LLMs.
06 posts
- 01
Taking RAG to Production: What Changes After the Demo Works
A RAG demo that answers 20 questions is not a production system. Evaluation gates, tracing, permissions, injection defense, and safe reindexing close the gap.
AIRAGArchitectureMLOps25 mins read - 02
The Anatomy of RAG: From Document to Cited Answer
Ten named parts of a RAG system, each with its real data shape, what it connects to, and how it fails, traced through one policy question.
AIRAGArchitecture25 mins read - 03
System One Models and Jev: How to Read the Benchmarks
TypeSafe's Jev returns typed decisions with confidence scores instead of prose. The architectural idea is sound. Every performance number is vendor-published.
AIAI AgentsSoftware ArchitecturePlatform Engineering19 mins read - 04
When Coding Gets Cheap, Judgment Gets Expensive
AI is collapsing the cost of writing code. That doesn't shrink the job, it moves the scarce part: problem selection, architecture, and verification.
AISoftware ArchitectureCareerEngineering Leadership16 mins read - 05
Why the Humanities Matter More in the Age of AI
AI generates fluent answers. It does not, on its own, understand context, weigh values, or catch its own bias. That's why literature, philosophy, ethics, and rhetoric are becoming load-bearing skills for anyone working with these systems.
AIEthicsHumanitiesPrompt Engineering8 mins read - 06
Autonomous AI Agents: How They Work and What They Cost in 2026
How autonomous agents actually work: the ReAct loop, memory architectures, real production examples with sourced numbers, current API pricing, and the failure modes you'll hit before the demo becomes a product.
AIAutomationFuture Tech10 mins read