a higher paid conversion rate can hide fewer paying customers.
if removing a trial changes who enters the funnel, i'd compare paying customers per install, then refunds and renewals at the same cohort age.
keep the denominator fixed.
a travel app doesn't need me back every morning.
it needs me back when i'm planning the next trip.
i'd rather measure that than turn it into another daily check-in.
an invite button can feel like a privacy decision.
a memory app's builder explained that inviting someone doesn't reveal your memories. i'd put that sentence beside the button, not leave people guessing.
'i reverse engineered X' is a good hook.
but the demo has to answer the obvious next question: what did you actually figure out?
an image tool i saw here has a 'try a sample batch' button.
show me what i'll get before asking me to bring my own files.
small onboarding detail. i like it.
i asked a builder why he'd made a job application tracker.
'sometimes i forgot i applied altogether.'
that sentence tells me more than a feature list.
i suggested showing a concrete finding on an app's landing page. the builder pointed out it already existed inside the app.
a useful reminder when giving feedback: 'show it here' and 'build this' are two different requests.
a cheap photo frame is a good constraint for a home dashboard.
@wellavelino described using one websocket and redrawing only when something changes. my favourite detail: updates without unplugging it. that matters a lot more once the screen is on a wall.
'your data stays in the file' can mean you can recover it, or you can open it elsewhere and read the finished document.
a conversation about a Markdown app here made that distinction concrete. i'd want to know which one i'm getting before trusting a format.
a small suggestion here about showing when data last synced became a real conversation with another builder.
that kind of exchange is a better reason to share work than trying to find something impressive to announce every day.
a detail from the Apartment concept: heating needs two numbers, what the room is now and what you're asking for.
i want that distinction to stay clear between the room view and the control. still a design exploration, not a working heating integration.
after using voice to steer codex through iphone maintenance, the interaction i keep thinking about is 'wait, leave that one'.
starting a task is only half of it. changing your mind halfway through needs to be just as easy.
the 3d view is the fun part of Apartment to show.
but turning off the same light every night shouldn't require exploring a model. i want everyday actions to stay easy to reach too.
the design challenge is making both feel like the same app. still figuring that out.
one thing i want from a smart home screen: make it obvious when a command hasn't worked.
a lamp looking on in the interface isn't enough. the actual lamp has to be on too.
i'd rather see 'not responding' than a reassuring animation that's lying to me.
i've been delegating iphone maintenance to codex through device hub on my mac, using my voice.
i'd like more tools to work this way: talk through a task, see what's happening, and steer along the way. phone cleanup is a pretty good place to start.
the setup is the part i keep coming back to with Apartment.
a 3d view can make a light easier to find. but if getting your home into the app takes an entire weekend, that's a lot to ask before the first useful tap.
a closer look at the ipad concept i'm working on.
pick the room, then the actual lamp. i'm trying to make the controls feel familiar before you've learned any device names.
still a work in progress.
a 3d smart home UI has to earn the extra dimension.
if finding a lamp takes longer than tapping a list, the nice render isn't helping. that's the test i want Apartment to pass.