Practical lessons on defining missions, assigning work to AI agents, keeping human oversight, validating results and measuring whether any of it actually helped.
Insights
Practical AI, from the work itself
Six Mistakes to Avoid When Using AI for Decisions
Unsupported recommendations, weak assumptions, missing context and too much confidence: the six most common ways AI-assisted decisions go wrong, and how to avoid each.
How to Build a Simple Decision Matrix with ChatGPT
A step-by-step way to make structured comparisons with weighted criteria using ChatGPT, so important choices rest on more than a gut feeling.
How to Use ChatGPT for Better Everyday Decisions: A Practical Guide
A repeatable method for using AI to research alternatives, weigh trade-offs and make better-informed everyday decisions, with you still making the call.
Claude Usage Limits: 3 Years Later and Still No Transparency (February 2026)
Claude Usage Limits: 3 Years Later and Still No Transparency After three years of using Anthropic’s Claude across multiple product cycles, one issue remains unchanged: opaque usage limits that feel arbitrary, inconsistent, and poorly communicated. ? The Core Problem: Unpredictable Cutoffs You’re working. You’re mid-flow. Then suddenly: “You’ve reached your plan’s message limit.” No clear […]
AI Usage Limits, Renewal Friction, and the Trust Problem No One Talks About
AI Usage Limits, Renewal Friction, and the Trust Problem No One Talks About This morning I hit a familiar wall: “You’ve hit your usage limit.” The subscription renews today. Auto mode was enabled. The interface implied availability. But the system hard-stopped. This is not a rant. It is a product design observation. ⚠️ The Invisible […]
Cloud Agents with Computer Use: When to Use Them (and When Not To)
☁️ What Are Cloud Agents with Computer Use? Cloud Agents extend the agent model beyond your local machine. Instead of running inside your laptop environment, they execute in isolated cloud virtual machines with full build, test, and desktop control capabilities. They can: • Run in parallel • Build and test software independently • Interact with […]
The Automation of Automation, Accelerated
The Automation of Automation, Accelerated In November 2023, I wrote about The Automation of Automation and since then this has grown exponentially, fundamentally changing our world in ways we don’t even understand yet. Conceptually this was just that it was becoming easier and easier to connect automations together. This was being accelerated tremendously by Generative […]
Fixing Windsurf Auto-Memory: How Hidden Memories Broke My Projects and How I Recovered
## The Hidden Problem With Windsurf Auto-Memory Windsurf’s auto-memory system silently creates .pb memory files for every project—architectural rules, debugging notes, even stale temporary fixes. Over months, this accumulated into more than 100 hidden memories across 10+ projects. When these memories mixed outdated assumptions with live development, the AI began producing inconsistent, conflicting, or entirely […]
Cursor vs Windsurf Both Add Automated Agentic Capabilities
The Automation of Automation for Software Development: Agents, Control, UX, and Cost in Windsurf vs Cursor (June 2025) ? Agent Control: Windsurf vs Cursor Windsurf: Implicit & Flow-Based Cascade Mode runs agent workflows automatically based on file detection and intent. Plan Mode allows you to preview and edit generated workflows. Pros: Zero-friction automation. Great for […]