BLOGSIdeas worth
Ideas worth
thinking through.
Technical articles on AI engineering, evaluation, agent systems, and choosing the right problem to solve.
5 min readEngineering decisions · Technology evaluation · 29 October 2026Who Should Choose the AI Model? A Framework for Task-First SelectionA decision framework for model choice, verification, acceptable quality loss, governance, and efficient routing.↗3 min readResearch question · Human–AI interaction · 20 September 2026Memory, Provenance, and Anthropomorphism in Human–AI InteractionHow persistent memory may shape perceived understanding, trust, and social attribution—and why the provenance of a remembered claim matters.↗4 min readResearch exploration · Human oversight · 9 August 2026The Reliability Counter: When Should an AI System Ask for Human Help?A literature-informed proposal for evidence-based autonomy, selective intervention and the economics of trustworthy AI.↗3 min readIndependent research · Agent reliability · 18 June 2026When Should an AI Agent Trust What It Has Learned?A proposal for turning successful agent trajectories into reusable workflows without allowing past mistakes to become permanent behavior.↗5 min readResearch implementation · Efficient inference · 15 May 2026SLM Cascades: Designing Cost-Aware Model Routing for Coding AgentsInside a LangGraph code-repair prototype that assigns bounded tasks to small models and reserves larger models for open-ended planning and code generation.↗