Twitxy Saved

David Barnard

Verified account RevenueCat

@drbarnard

“Foremost expert on App Store minutia”―@BenThompson. Growth Advocate @RevenueCat. Host @SubClubHQ. Founder @Contrast. Featured in @nytimes @WSJ @TechCrunch etc
San Marcos, Texas contrast.co Joined July 2007
3.8K Following 17K Followers 89K Posts

Latest posts 5

David Barnard Verified account @drbarnard
I showed all this to a friend over lunch on Saturday and was surprised to hear that my process is way more dialed in than all but one or two engineers at his company.

Though to be fair, they have a $750/mo AI budget. I’m maxing out both my $200 Claude sub and now $500 Codex sub (what’s that like $30k/mo at API prices?!) . It’s definitely not the most token-efficient way to work, but it saves me a ton of time and dramatically improves the quality of the output.

(Friend intentionally unnamed so I don’t blow up his colleagues.)
David Barnard Verified account @drbarnard 17h
Over the weekend I got my vibe-coding setup pretty dialed in. And at this point I pretty much don’t even type into the text box in Claude Code anymore.

Instead, I file work via an Artifact and Claude adds them as @linear tasks. Then it builds what I asked for and steps me through checking its work (including its own screenshots from the simulator, but also steps to test on my own). I can leave feedback directly on that page for each issue that needs work.

And then I have a master sheet that tracks all open decision sheets, the ones in process, and the completed ones at the bottom.

I don’t even open Linear at all, just use it as a system of record, somewhere for Claude to document tasks and check them off so that todos and related details don’t disappear from the context window as we keep working.

When we’re at a stopping point, I just ask Claude to work through all the high priority tasks in Linear that have enough detail to start. And for those that don’t, create a new decision sheet outlining what it things it should do and asking for my feedback.

I also use this pattern for generating UI mockups. Claude uses ChatGPT Image (via the Codex CLI) to generate a bunch of mockups of different ways to solve a particular problem. Then it shows me all those mockups on a decision page where I can select one and/or give feedback.
View quoted post on X
David Barnard Verified account @drbarnard
Daaaang… TIL the iPhone SE 2nd generation (released in 2020 with a freakin home button) is still supported by iOS 27.

I can’t seem to find any estimates of what percentage of active iPhone users are on an iPhone SE. Anyone have a good source for that (or a big app with internal analytics)?
Gaurav Jain | Building with AI Verified account @MrJainG 14h
@drbarnard an old SE would be my pick for that. small screen and slow enough to show every janky scroll. do you test on Android too or just iOS?
View quoted post on X
David Barnard Verified account @drbarnard
😬 I didn’t really think through how much I need a second iPhone for testing now that my carry phone is the Duo. You can’t really build apps anymore without testing on both form factors regularly.

I’m thinking I should get a relatively old iPhone to help detect scrolling issues and other things impacted by processing power. But I don’t want to get something so old that it won’t support iOS 28 or be representative of the experience of most users (I considered the 13 mini, but so few people actively use a mini it’s just not worth it).

Based on @Telemetry_Deck data, the iPhone 14 seems to be the sweet spot right now. Just $300 refurbished on Amazon.