AI Agents Have Goals. DRIs Have Consequences.
AI agents can pursue goals, coordinate work, and operate with increasing autonomy. But until they can own the consequences, the DRI still has to be human.
AI agents can pursue goals, coordinate work, and operate with increasing autonomy. But until they can own the consequences, the DRI still has to be human.

A fairy tale about perfect obedience, royal confidence, and the kind of miracle best kept under supervision. In a kingdom fond of polished promises, a wondrous guardian arrives with all the right answers, all the right manners, and very little sense of its own.
Three practical Textual patterns I learned while building a Python terminal app.
Why as const beats enums for most TypeScript use cases. Covers Zod integration, the satisfies operator, and keeping a single source of truth.

Vibe-coded apps are everywhere, and most fall apart the moment they meet a real user. AI hasn't killed software discipline. It's just made ignoring it effortless. An observation on scope creep in the age of AI, and why infinite capacity might be the worst thing to happen to software.

A small experiment exposes how companies are secretly and illegally using AI to screen candidates and generate follow-up emails.
An AI-generated analysis of Anthropic's published system prompts
A small side project I worked on this week: a set of tools to enhance data portability from Duck.ai, the private AI chat feature of DuckDuckGo.
Most mobile software assumes perpetual engagement and recurring revenue. This list catalogs the exceptions — applications built with defined scope, clear boundaries, and business models that do not depend on retention metrics. These applications operate on the assumption that user data belongs to the user, not the vendor.

AI can speed up the right kind of work, but it also has a talent for scaling whatever standards you bring to it. Used well, it helps careful engineers move faster; used badly, it turns half-baked thinking into a much more efficient mess.