Notes from my work, collected here: iOS, legacy, automation. One topic per piece.
Local AI or a cloud model: the question comes up in every first conversation about automation, usually as a worry about the data. Four questions answer it more reliably than the worry does, and both paths carry a price worth knowing in advance.
· 7 min
Native or hybrid is too narrow a question. Between a web app and a native app lie six steps, and three questions about the product decide which one fits. I have answered two of them on my own projects: once for web, once for native.
· 8 min
An avatar presenter had to stay centred in frame after cropping. That single requirement moved the processing out of the workflow and into an existing backend. Why the limit of low-code is not about the complexity of the logic.
· 5 min
In an evening project, an agent rates the user's tone and keeps a record of it. The rating is the model's job, the bookkeeping is a Python script's. Why that boundary is the actual architecture decision.
· 5 min
Generated code rarely fails visibly against third-party libraries. It names methods that are gone and versions that never shipped. Reading more carefully does not help; a source that is not the model does.
· 7 min
The decision is not made about the code but about the knowledge inside it. Three signals that actually decide it, and the price of the path that is usually right.
· 9 min
A visual workflow is code with all the obligations of code. Five of them decide whether an automation survives a year or gets rebuilt at the first change request.
· 9 min
The line does not run between old and new, but between a value and a stream: actors where state is shared, Combine and RxSwift where it keeps flowing.
· 9 min
Generated code is rarely sloppy, it is over-engineered. That barely shows up in a diff, which is why my checkpoint sits before the first line of code.
· 8 min
Anyone getting into an unfamiliar codebase should not read it but interrogate it. The order of the questions decides whether you know something after a day or have seen everything after a week.
· 9 min
An agent does not need as much context as possible, it needs an index and a rule for fetching the right part. How I split project knowledge so it stays usable as it grows.
· 8 min
What you have to secure in the first few days so a staffing problem does not turn into losing the app — in the order that matters.
· 7 min
The order is no accident: build, dependencies, tests, release path — and only then the code.
· 8 min