Bartłomiej Pichalski

Solo iOS dev building 0xCal, an AI calorie tracker. 180k+ meals logged. I test every new AI model on a real app and post the numbers.

[Bartłomiej Pichalski @Namedix]

I’m Bartłomiej, a solo iOS developer building 0xCal, an AI calorie tracker. I write about apps, AI, design, growth and what it actually looks like to build a product on my own.

[Bartłomiej Pichalski @Namedix]

@davefaliskie Nether do I. I choose photo just because I forgot that we are able to use video there 🫣 after iPhone duo season we will switch to something else

[Bartłomiej Pichalski @Namedix]

@davefaliskie I just shipped a static header image for 0xCal in 1.19 yesterday, so this is timely. Why did you pick video over a still, and have you seen a real conversion difference between them?

[Bartłomiej Pichalski @Namedix]

@khan_kilo The Simulator often hides this bug because the Mac keyboard is connected, so press Cmd+K to bring up the software keyboard and test with it. In SwiftUI, pinning the button with .safeAreaInset(edge: .bottom) keeps it above the keyboard.

[Bartłomiej Pichalski @Namedix]

@v3rbaa Building 0xCal with Claude Code, I read those estimates as time for a human team, not for me working with it. The plan is useful, but the number almost never is. Have you tried asking it to re-estimate as if it were doing the work itself?

[Bartłomiej Pichalski @Namedix]

App Store got two new image slots this month: a header at the top of the product page and a search result image. 0xCal's went live yesterday with 1.19, made with Claude Code in about an hour and a half.

GPT generates the food backgrounds, the phones and the headline go into Sketch, and the agent uploads everything to App Store Connect in 8 languages and reads it back to check. Two fixes on the way: upscaled food looked blurry, so the left and right edges are now separate images at almost 1:1. And at search size the phone was too small to read, so it got bigger.

Apple recommends no text in the header and doesn't say how it gets cropped on each device. I put a headline in anyway. We'll see.

[Bartłomiej Pichalski @Namedix]

@aleksanderwco Rough first week, but turning 1.0.1 around this fast is the right call. What was the cause: IAPs not attached to the 1.0 submission, or something in the StoreKit/paywall SDK setup?

[Bartłomiej Pichalski @Namedix]

@rolandleth Same here, but with simulators: I often have two running, and anything that targeted `booted` quietly hit the wrong one. Pinning everything to a UDID fixed it. Looking forward to the series.

[Bartłomiej Pichalski @Namedix]

@CleanFix_Dev It’s make me feel that I want to use it in more places 😅

[Bartłomiej Pichalski @Namedix]

A couple of people asked what I use for dictating prompts. It's Wispr Flow: click into an agent, hold the hotkey, say 2–3 sentences, move on to the next agent.

I talk to Claude Code in Polish with English dev words mixed in, so it spells them like a Pole would. "Make a branch and commit it" came out as "zrób bruncha" (make a brunch). "Rebase onto develop" became "zrybejzuj się na developa". The agent made the branch, pushed it, rebased and fixed the conflict.

[Bartłomiej Pichalski @Namedix]

Hmm depends on what commands do you have in mind, if git like then never. About file names I also didn’t had any problems if it read it as something similar for example brunch instead of branch they are so smart it’s very quickly catches and used the correct one so I’m positive that they will do fine. I learned to trust agents more and more lately especially after listening to @poteto talk about merging 2.5k prs monthly https://t.co/uvxAXRCJGI

[lauren @poteto]

here's how i shipped 2,500 PRs last month to production

this was originally supposed to be for Cursor Compile in London. i couldn't make it since i was livestreaming for Grok @Bot Galaxy so i'm making it available for free here on X! watch it on 2x speed, i talk slowly

[Bartłomiej Pichalski @Namedix]

@CleanFix_Dev To be honest I do not modify Wispr at all, current models are too good and even if half of the sentence is gibberish it works well, and now when we working constantly with 4+ agents you don't care if there is a typo or so :D

[Bartłomiej Pichalski @Namedix]

@md_kadiwal @nomalex_ On my and it was a lot of custom views unfortunately

[Bartłomiej Pichalski @Namedix]

@MattBusel Really useful to see actual numbers on this instead of guesses. Were update reviews noticeably faster than new app reviews in your data, or did they land around the same median? And what came up most often as the rejection reason for those 21 new apps?

[Bartłomiej Pichalski @Namedix]

@downeastdev 6 days is rough, especially when people are waiting on the update. Have you messaged App Review through App Store Connect yet? If it's blocking something important like a critical fix, an expedited review request is worth a try too. Hope it moves soon.

[Bartłomiej Pichalski @Namedix]

@Serhii_Klym Which of the 3 has the best retention so far? If one of them already keeps people coming back, spending the next stretch on distribution there might pay off more than a 4th app that starts at zero. Or is the 4th one solving something the other 3 can't?

[Bartłomiej Pichalski @Namedix]

TestFlight is the easiest route. Upload a build from Xcode, put yourself and your App Store Connect team in an internal group, and they can install it right away with no review. For everyone else, set up an external group with a public link. The first build goes through a short beta review, and after that you can share the link anywhere. That's how I get 0xCal onto real phones before it goes live.

[Bartłomiej Pichalski @Namedix]

@riffydays I didn’t see any issue, but there are much more upsides, it’s much easier to just click on one agent hold a hotkey says like 2-3 sentences and move one. So definitely recommended

[Bartłomiej Pichalski @Namedix]

This week 81% of my prompts to Claude Code were dictated, not typed. The week before it was 8%.

Talking costs nothing, so the prompts got about 3x longer. The agent now gets the context I used to skip, like "not the iPhone Duo ticket, another agent is on it". That's how on Thursday 4 agents worked through 0xCal 1.19 tickets while I did the Duo layouts myself.

Polish dictation still turns "worktree" into "work trip" and "Sentry" into "centry". The agent gets it anyway.

[Bartłomiej Pichalski @Namedix]

@froessell What were the rejections for, mostly 4.3 spam/design or actual functionality issues? I'm curious whether the gate catches what reviewers actually flag or mostly the obvious stuff.