Dispatch Unit · Articles
Field notes on Apple platforms, Swift, AI and everything else.
AI is breaking down the barriers to software development, but it also risks turning creativity into mass production. Much like postwar architecture, building faster doesn't necessarily mean building better. What if the real AI revolution isn't about producing more, but about giving us the freedom to imagine something different?
Agentic coding is not only lowering the cost of building software. It is also lowering the cost of replicating it. As implementation becomes easier to reproduce, competitive advantage shifts toward distribution, proprietary data, community, reputation, and the ability to decide what to build next.
For years we dismissed them as "stochastic parrots". Then we looked inside the black box. Here's what we found.
A data-driven portfolio with type filtering and pagination, a résumé HUD that links into it — and two scroll-related bugs (a pinned-scroll cancellation and a classic-script function collision) that were more instructive than the features.
Foundation Models and the @Generable macro make it possible to build interfaces an on-device model assembles at runtime. Here is what that actually means, and where the wall is.
How a minimal play/stop/volume player streams a YouTube playlist's audio on loop — using the IFrame API, a hidden iframe, and view transitions to keep the music playing across page navigation.
I dreamed of a cow falling off a mountain. The scene was entirely made up, yet it held together. That led me to an uncomfortable question about AI: is predicting the next word really the opposite of understanding? Maybe not, and maybe the stochastic parrot is a fair critique but not a sufficient one.
The first post written in Notion and published to monoidx.dev by the automated sync — covering how the pipeline works end to end.