How I AIShe vibe coded an iPhone app and launched it to the App Store with zero coding knowledge
CHAPTERS
Meet Bryce and the “Daily Hundreds” idea: a non-technical builder goes to the App Store
Claire introduces Bryce Rattner Keithley—career talent/recruiting leader and self-described non-technical person—who nevertheless shipped an iPhone fitness app. Bryce previews what the app does: a daily 100-rep exercise with logging and (later) fun AI-generated demo videos.
Origin story: pandemic “100 reps a day” need turns into a product prompt
Bryce explains the personal problem that sparked the app—wanting lightweight movement during long at-home workdays. The key insight: variety matters; she’d do 100 reps daily if someone simply told her what to do.
From prompt to MVP: choosing Replit (and why Lovable didn’t win)
Bryce describes opening Lovable and Replit the same day and prompting for a basic MVP, then sticking with Replit for momentum and support. Claire frames Replit as strong for quickly building functional apps with backend needs.
How Bryce actually worked in Replit: preview panel, iteration loops, and “plan mode”
Bryce shares her day-to-day workflow: making requests, checking the preview panel, and undoing when changes go sideways. A major unlock was using “plan mode” to align on a step-by-step approach before letting the tool modify the app.
Beginner’s mindset as a feature: trusting tools, asking ELI5, and buying infrastructure blindly
Instead of being blocked by what she doesn’t know, Bryce leans into it—asking for idiot-proof instructions and sending screenshots when stuck. Claire highlights the broader shift: dev tools now get adopted by non-developers who can ship outcomes without understanding every layer.
Why the app needed videos: users asked ‘what is that exercise?’
As Bryce shared early versions, people texted questions about how to perform unfamiliar movements. To remove herself as the “single point of failure,” she decided the app needed clear demos—but didn’t want it to just be videos of her.
Building the animal-video pipeline: Sora → Gemini images → Higgsfield + Kling motion control
Bryce walks through the production workflow she landed on after trial and error. She generates a still animal image in Gemini (NanoBanana), films herself doing the exercise, then merges image + motion video in Higgsfield using Kling (noting quality differences vs using Kling standalone).
Prompting like a coach: precision cues, retries, and “no extra characters”
In a live example (leopard crunches), Bryce demonstrates highly literal prompting: body orientation, limb placement, and scene constraints. She iterates when the model misses details, sometimes rewriting from scratch rather than tweaking copied text.
From still to video: matching orientation, controlling scene, and waiting for renders
Bryce shows how she pairs the still image with her exercise clip in Higgsfield’s Motion Control. She chooses image-based scene control (to avoid her living room) and ensures orientation matches so the motion transfer looks natural.
Anthropomorphic outtakes and wins: the floating genie, surprising turtle, and edge cases
They review prior generations to illustrate both failure modes and delightful surprises. The ‘floating genie’ Superman demo shows how wrong things can go, while the turtle examples prove even unlikely animals can look convincing and friendly.
Turning a web app into an iPhone app: the App Store as the “holy grail”
Bryce explains why a Replit-hosted web app isn’t enough for daily consumer habits and credibility. She recounts early advice that she’d “need someone technical,” and how that guidance changed as models improved—prompting her to attempt the full App Store path herself.
Claude + Claude Code + terminal: the step-by-step App Store submission workflow
Bryce describes a “two-agent” workflow: Claude for planning and guidance, Claude Code for implementation, and her own terminal actions for privileged steps. She compressed the work into ~25–30 hours over a weekend and got approved on the second submission.
App Review feedback and fixes: checkboxes, Sign in with Apple, and account deletion
Bryce shares the concrete reasons for her first rejection and how she resolved them by pasting Apple’s feedback into Claude. The fixes were practical and compliance-driven: correct parental settings, ensure Sign in with Apple works (especially for iPad compatibility), and add account deletion.
Lightning round: hiring, career relevance, and “robot karma” prompting habits
Bryce reflects on what building changed about her views on teams and hiring—especially adapting roles as AI shifts what humans should optimize for. She recommends books about evolving beyond past strengths and shares her prompting style: be literal, use screenshots, and balance politeness with firmness.
