On-device AI: what it is genuinely good at now
Five categories where local models are genuinely sufficient, the honest ledger of what you give up, and a five-second test for whether an app is really running on your phone.
Tag
#Evaluation
An iPad app is not a bigger phone app
The differences are not about pixels: how the device is held, how long the session runs, whether it shares the screen, and whether a keyboard is attached.
Reading an app's privacy label properly
Three separate questions hide in every category, and the one people skim past is the strongest signal on the page.
Choosing a kids app for a child who cannot read
Four questions, and only one is about content. The others are about what happens when a child taps something unexpected — which is most of the time.
Voice recorders that transcribe: where they still fail
The same four failures across every tool, because they are properties of the audio rather than the app — which is why comparing on a clean sample tells you nothing.