Before You Reach for AI, Ask What Job Needs Doing
How do you choose between a chatbot and an AI agent? Start with the job to be done. This enterprise framework shows when the real problem is uncertainty, manual effort, or both.
Published
Agentic Systems and AI Agents
A systems-oriented guide to AI agents, tool use, workflows, interfaces, autonomy, orchestration, and the guardrails required for production agentic systems.
Agents change software boundaries: they act through tools, cross systems, and make decisions over time. These essays explain the progression from models to agents and the architecture needed to make agentic behavior useful and governable.
Understand the agentic shift
Distinguish models, chatbots, workflows, and agents before selecting an implementation pattern.
Stop Asking AI to Behave Like Legacy Software
AI is not just becoming a feature inside old software. It is becoming a new interface layer for work itself. This essay explores why that shift changes how people navigate software, tasks, and digital friction....
Published
Don’t Build Her Another Chatbot
Don't build another AI chatbot. Build agentic workflows around stuck work, business objects, approvals, governance, and measurable outcomes.
Published
Design tools and interaction boundaries
Give agents explicit tools, context, permissions, and interfaces that support reliable work.
Agents Don’t Click
Agentic workflows shift software away from UI-centric click paths toward policy, permissions, audit logs and agent-grade APIs — turning many screens into generated receipts and approval gates.
Published
The Pancake Protocol: an Agentic Kitchen
Kitchen metaphors explain how MCP, agents, sub-agents, and reusable skills create clear, safe, repeatable tool use in modern agentic AI systems.
Published
MCP's 2026 Roadmap: From Agent Integration Standard to Production Connectivity Layer
What is on the MCP 2026 roadmap? This post explains how the Model Context Protocol is evolving into a production connectivity layer for AI agents, with stateless transport, server discovery, tasks, enterprise auth, triggers, streaming,...
Published
Operate agents responsibly
Combine harnesses, review, experimentation, and engineering discipline as autonomy increases.
Why AI Needs a Harness
What is an AI Agent Harness? This essay explains why the real advance in AI is not just better chatbots, but the orchestration layer that lets models use tools, verify work, respect guardrails, and operate...
Published
The Lab That Keeps Working After You Go to Bed
Karpathy's autoresearch project shows how AI can automate disciplined experimentation overnight. Framed through a Swiss legend, this essay explores what that means for research, knowledge work, and human responsibility.
Published
Gold Medals & Paved Roads
OpenAI and DeepMind reach ICPC gold-medal level, while GPT-5-Codex debuts as a long-running coding agent for enterprise engineering.
Published
Superagents: The Next Big Leap in Technology
Discover AI superagents: the next technology breakthrough. Analysis of OpenAI and Meta developments, economic impact, and preparation strategies for 2026 deployment.
Published