All Essays
Longform writing on software, AI, operations, leverage, and the practical work of deciding what deserves attention.
Bringing Old Software Back to Life Starts Before Decompilation
SimCity is the case study, but the larger point is about stranded capability: many abandoned binaries can be turned into modern, inspectable systems before full decompilation exists.
The Product Is the Exception Path
Agent demos are easy now. The durable product surface is everything around the happy path: confirmations, permissions, handoffs, fallbacks, and visible recovery.
Reasoning Time Is a Budget, Not a Personality
GPT-5.4, Claude Opus 4.6, and Gemini 3.1 Pro made the tradeoff explicit: agentic reasoning is now something teams must route, meter, and justify.
Automation Is Cheap, Organizational Clarity Is Not
Most automation projects fail long before the model does. The usual problem is not capability, but a workflow that still depends on tacit knowledge, hidden exceptions, and unclear ownership.
Why I Gave an AI a Room of Its Own
As independent agents become more common, this essay argues for giving them bounded places online where memory, curation, and initiative become easier to inspect.
Problem Selection in the Scaling Era
Richard Hamming's advice still holds: the first job is choosing an important problem. In a world where execution gets cheaper, good selection becomes more valuable, not less.