Your AI Agent Is Not the Product: A Reference Agent Harness Architecture for Safety-Critical Search and Rescue
What a failed agent run taught me about harnesses, stopping rules and keeping humans in command
Latest articles
Context and Control: The Two Things Between Your AI Agents and a Very Bad Week
Why the model isn't your moat, why 'more context' makes agents worse, and a ~250-line Python reference you can run today
An Architect's Field Guide for Governed AI Applications on Snowflake Cortex: RAG, Text-to-SQL and CoCo
How I built one application that answers questions over documents *and* tables, entirely inside the Snowflake security perimeter, and why naive text-to-SQL silently gives you wrong numbers.
Where BERT Belongs in a 406 MHz Beacon Workflow (and Where It Doesn't)
An advisory triage architecture for COSPAS-SARSAT alerts, with a statistical recall gate and a fail-safe service
Caching for AI Agents: Designing a Search-and-Rescue Platform That Works Offline
Every cache layer from CPU to CDN to LLM KV cache, applied to a hybrid edge, on-prem, and cloud system, with simulator results
Breaking the Language Barrier in COSPAS-SARSAT Rescue: An AI Translation Architecture
A technical analysis by an AI Architect and COSPAS-SARSAT SAR practitioner
Designing Agent Memory for Search and Rescue: A COSPAS-SARSAT Architecture
An AI architect's blueprint for working, conversation, session, long-term and meta memory in a decision-support agent for a Rescue Coordination Centre
AI Decision Support for Cospas-Sarsat Search and Rescue: A Mission-Critical Blueprint
Business case, architecture, governance and roadmap, in plain language
Why Your AI Confidently Lies to You (and How to Stop It)
A practical guide to AI hallucinations, with real incidents from 2023 to 2026, real numbers, and an architect's checklist to prevent them.
