Undolly is getting much better at surviving messy real-world photo libraries. Long scans now keep moving, recover better, and show progress in real time, so cleaning duplicates feels less like a gamble and more like a quick win ✨
I won’t be able to publish Code & Promps tomorrow as I’m enjoining some days off. Do not worry I’ll post it on Monday and meanwhile last one is still here open.substack.com/pub/codeand...
I see a lot of people trying to make Xcode faster or moving to Tuist. Oh lord, the day you discover Bazel. I use it in every project, even in small personal ones. Far more agent-friendly, easy, and fast.
Just added 4 new smart suggestions to Undolly 🔍 Now it automatically finds your old WhatsApp media, RAW photos, and slo-mo videos that eat up storage. The app keeps learning which duplicates annoy users most.
Added tap-to-edit bulk mode to HiddenFace. Select 20 photos → tap faces you want to blur → done. Simple as it should be. 📸
Added a 3-day free trial to Undolly. Figured people need time to actually see their storage freed up before committing. Sometimes the best conversion optimization is just... being reasonable 🤷♂️
Bulk face detection for HiddenFace 2.0 🎭
Same person in 50 photos? Tap once, AI finds them everywhere and blurs all at once.
50 taps → 1 tap ⚡️
Beta soon.
#IndieApps #iOSDev #BuildInPublic
Working on HiddenFace 2.0 New: Quick photo mode 1. Snap a pic 2. Faces auto-blur 3. Share or delete (never touches your gallery) Perfect for those 'I need this blurred NOW' moments 📸
Killed Fastlane across all 5 apps this week.
Moved everything to ASC CLI (open-source App Store Connect tool).
What changed? Now my AI agents can do much more!
Fastlane automated builds. ASC CLI opens the entire store. 🚪
#BuildInPublic #IndieApps #iOSDev
Spent the day rethinking HiddenFace's first impression. Out: generic welcome screen. In: Bento Grid layout. More features are coming.
Redesigning App Store screenshots for three apps this week. You're not just building features; you're also your own marketing team, and I’m terrible at it. Using Koubou and Claude makes it far easier.
"Generate App Store screenshots for my app"
→ Claude + Koubou:
• Derives your brand colors
• Creates 5 slides (hero → features)
• Exports pixel-perfect PNGs
• Handles 10+ languages
20 minutes vs 2 days in Figma.
Open source, MIT license.
github.com/bitomule/ko...
Koubou’s HTML live preview displays features like photo management, metadata details, and cleanup options for digital storage efficiency.
Proper live mode coming to Koubou. Just run `kou live` and it will open a web dashboard that will auto-update each time you or your favorite agent changes anything. Feels like magic. Next, I would like to build templates to make agents work easier. github.com/bitomule/Ko...
Koubou v0.15.0 is out 🎉 Two big features: ✅ HTML/CSS templates (alongside YAML) ✅ Claude Code skill for AI-assisted screenshot generation Generate professional App Store screenshots in 20 minutes instead of 2 days. brew install koubou github.com/bitomule/ko...
Appstore screenshots describing that Undolly uses AI so pick best but you always decide
Improving Koubou's skill so generated screenshots get better. Looking good so far! github.com/bitomule/Ko...
Koubou v0.15.0: HTML rendering + AI skill for App Store screenshots
What changed:
• Design with HTML/CSS (not just YAML)
• AI skill built-in
brew install koubou
Now with AI superpowers.
github.com/bitomule/ko...
Install:
brew install koubou
Open source. MIT licensed.
github.com/bitomule/ko...
Expect more improvements to make Koubou more agent friendly and please let me know if you have any feedback.
6/ Real workflow:
20 minutes vs 2 days in Figma.
All the original features still work:
• xcstrings localization
• Live preview mode
• Pixel-perfect dimensions
5/ What's new #2: Claude Code skill
Conversational screenshot generation:
"Generate 5 App Store screenshots for my coffee app"
Claude:
✅ Derives brand colors from Assets.xcassets
✅ Creates narrative arc
✅ Generates HTML templates
✅ Exports PNGs
4/ Same template works for:
• iPhone 16 Pro
• iPad Pro 13"
• Mac screenshots
• Apple Watch
100+ device frames included.
3/ The magic: Pre-composition
Koubou renders your screenshot INSIDE the device frame automatically.
Your template receives a composited image.
No CSS tricks. No border-radius hacks. Just place the image.
2/ Example:
Instead of fighting YAML parameters, write HTML templates with full creative control. CSS gradients. Custom layouts. No limitations.
1/ What's new: HTML/CSS rendering mode
Koubou started with YAML configs (still supported). Now you can design screenshots with full HTML/CSS.
Why? LLMs are great at HTML, terrible at YAML.
🧵 Koubou v0.15.0 just dropped
Two features that change everything for AI-assisted App Store screenshot generation.
Let me show you 👇
Three iterations on face scanning today. Each one makes it smarter, more robust. The journey from "it detects faces" to "it actually works with real videos" is... humbling 😅 #iOsDev #BuildingInPublic
Building face detection for video:
What I thought: "just use an ML model"
Reality:
- Frame prefetching
- Geometry-based keyframe detection
- Multi-anchor identity tracking
- Velocity extrapolation
- Embeddings, ML…
It's crazy how complex this is, one of those interesting cases where AI helps but the task is still massive. Algorithms, models, testing, and UX all together.
I have been working on HiddenFace V2: Multi-anchor pipeline for face video tracking. Can't share much yet, but let's just say face detection is about to get a lot more interesting... 👀 #AI #Privacy
Sometimes you just need to build the tool yourself.
Working on Nagare3D: a simple web app to manage 3D print requests for my Bambu Lab A1. Early stages, but it's already saving me time.
Might share the code soon if there's interest.
New weekly newsletter published. This week, everything revolves around "Harness Engineering." I strongly believe this defines the future far better than "Vibe Coder." codeandprompts.substack.com/p/code-and-... Subscribe free and get it in your inbox every week.