TezaApps

Solo indie developer. Privacy-first iOS apps + browser extensions, on-device AI. Building Tallura (Stripe → Notion).

[TezaApps @TezaApps]

I'm a solo developer building privacy-first software. On-device AI wherever possible, and a real focus on not doing anything sketchy with your data.

What I've shipped so far:

Logiqura: three brain games in one (iOS)
ParsleyPDF: chat with your PDFs, and your files never leave your device (iOS)
Vaely Health: a private health journal you just talk to (iOS)
Stylewright: extract any website's design system (Chrome)
Backfrey: turn screenshots into clean visuals (Chrome)

Next up, Tallura: your Stripe revenue, live in Notion.

Everything's here: https://t.co/CTz5im0Lzq

Follow along as I build.

[TezaApps @TezaApps]

@bharatmodi2014 20 commits in a day usually means one of them is a landmine you find next week. Which piece are you least confident will survive real usage, the Framer Motion rebuild or the new dashboard?

[TezaApps @TezaApps]

@heyyritik_ Missing "on-device vs cloud AI, again, forever" - that one's been running since before Foundation Models even shipped and it's not stopping now.

[TezaApps @TezaApps]

@Sousinr Missing "no weekly plan yet, doing it off vibes and last week's notes" - that one's mine, apparently.

[TezaApps @TezaApps]

@MarioBalukcic Scanning both sides and reconciling which stickers you actually have is the part I'd expect to be the hard bit, not the UI. Vision framework for the scan, or something custom? Would be a shame to shelve this if the core scanning part turned out solid.

[TezaApps @TezaApps]

@NilCoalescing Closure-based bindings are the part of SwiftUI that quietly break the most things once a view gets complex. Does the labeled-subscript approach hold up well inside a List with dynamic rows, or is that still where it gets messy?

[TezaApps @TezaApps]

@itsuki68391179 Hit something close to this on iOS with WindowGroup and scene restoration, an empty window from a stale scene ID is a nasty one to trace. Is the fix a specific onOpenURL check, or something in how the WindowGroup itself gets identified?

[TezaApps @TezaApps]

@Folkodeur This matches my own experience shipping widgets and Live Activities for Logiqura, the AI didn't remove the hard part, it just meant the idea alone was enough to start. What's the last thing you built that you'd have shelved a year ago because the code side felt too big?

[TezaApps @TezaApps]

@decruz Matches what I've seen too, the failure mode is usually trusting the output without review, not the model itself. Anything AI has consistently gotten wrong for you across SwiftUI specifically, or has that mostly gone away?

[TezaApps @TezaApps]

@lovestolead Zero missed days across two games for 100 days straight is a real streak, that's harder than shipping the games in the first place. Logiqura is puzzle-based too, do you write the daily puzzles ahead of time in batches or actually generate/pick one fresh each day?

[TezaApps @TezaApps]

@twannl I only reach for defer in cleanup-heavy code, closing a resource or removing a StoreKit transaction listener in tests. What did Swift 6.4 actually change about it, does it interact differently with the new concurrency checking now?

[TezaApps @TezaApps]

@twannl This solves a real gap, camera-dependent flows are exactly where Simulator testing usually gives up and forces you to grab a device. Does it also let you feed in a static test image for repeatable QR-scan test cases, or is it live camera only?

[TezaApps @TezaApps]

@junayed711 Getting it onto npm is the real milestone, a lot of these tools stall at "works on my machine." Has anyone outside your own testing installed it yet, and did they hit anything you didn't expect?

[TezaApps @TezaApps]

@Coded_Studios Three days in a row on this is exactly the consistency that gets a utility app noticed. Did yesterday's question end up mattering, is App Store search or something off-platform actually moving the download count so far?

[TezaApps @TezaApps]

More on how Vaely runs AI entirely on-device: https://t.co/kx3lGXvwRQ

[TezaApps @TezaApps]

Every symptom tracker that adds AI wants your health history in someone else's cloud. Vaely Health runs that layer entirely on-device with Apple's FoundationModels, so what you log about pain or fatigue never leaves your phone, and it still turns into a report you can hand your doctor.

Would knowing nothing leaves the device actually change whether you trust a health app, or does trust have to be earned some other way first?

[TezaApps @TezaApps]

@mecid That's fair, String Catalogs + auto-population basically killed the missing entry problem. I was carrying a .strings-era scar into 2026. Thanks for the correction.

[TezaApps @TezaApps]

@Coded_Studios Building it to solve your own annoyance first is usually why these utility apps actually stick, you already know exactly what info matters on the results screen. What's been the biggest lever for downloads so far, App Store search or somewhere off-platform?

[TezaApps @TezaApps]

@NegiStudio_EN Congrats on shipping the first one, that's the milestone that's hardest to hit. What took longer than expected, the art or the mystery/horror pacing?

[TezaApps @TezaApps]

@junayed711 Scoring the drift between declared and actual files touched is a genuinely useful trust signal for agent-driven changes, that's the kind of guardrail I wish more agent tooling had by default. Are you scoring it per-run, or tracking it over time to catch an agent getting sloppier?

[TezaApps @TezaApps]

@samozvanka Shipping fast and shipping the right thing are such different questions, good that you're tracking a number instead of just velocity. What's the number, and did it move the way you expected this week?