Back/How I AI
How I AI

How I AI: Bryce Rattner Keithley's No-Code Playbook for Building a Fitness App with Replit, Gemini, and Claude

Discover how a non-technical talent leader built, designed, and shipped a complete iPhone app to the App Store, using Replit for development, Gemini and Higgsfield for custom AI videos, and Claude as her technical co-pilot.

Claire Vo's profile picture

Claire Vo

June 1, 2026·9 min read
How I AI: Bryce Rattner Keithley's No-Code Playbook for Building a Fitness App with Replit, Gemini, and Claude

Welcome back to How I AI! I’m Claire Vo, and I’m on a mission to help you build better, faster, and more creatively with the latest AI tools. Today, I was so excited to sit down with my friend, Bryce Rattner Keithley. To be perfectly clear, Bryce is one of the most talented people I know in the world of talent and recruiting, but she is, in her own words, “not at all” technical. She's spent her career working with brilliant engineers, but always from the other side of the table.

And yet, she somehow beat me—and a lot of other software engineers—to the App Store. She built a polished, functional, and frankly delightful fitness app called Daily Hundred. It’s not just a simple tracker; it features custom, AI-generated videos of anthropomorphic animals demonstrating every single workout. Think of a polar bear doing lunges or a turtle doing curtsies. It’s amazing.

In our conversation, Bryce walked me through her entire journey, proving that with a beginner's mindset and the right set of tools, the barrier between idea and execution has completely crumbled. She shared three distinct workflows that took her from a simple prompt to a published app. We’re going to break down how she used Replit to build the core application, her creative process for generating animal workout videos with Gemini and Higgsfield, and finally, how she used Claude as a technical co-pilot to navigate the intimidating process of App Store submission. This episode is for anyone who’s ever had a great idea but thought, “I just don’t know how to build it.” As Bryce shows us, now you can.

Workflow 1: Vibe Coding a Functional App with Replit

The story of Daily Hundred begins during the pandemic. Like many of us, Bryce was stuck in the same room for hours on end and craved a simple, consistent way to move. She started a 100-pushup-a-day challenge but found it monotonous. The core idea was born: what if an app just told you a different exercise to do for 100 reps each day? Simple, effective, and varied.

With this idea in mind, she didn't get bogged down in technical specifications. She just started building.

### From Idea to Prompt

Bryce opened two AI-powered development tools, Lovable and Replit, on the same day and gave them both a simple, direct instruction. This was the seed for the entire project.

“Hey, build me a tool called Daily Hundred that pushes a different exercise to me to log a hundred reps.”

She was amazed that both tools immediately spit out a basic, minimum viable product. She ended up sticking with Replit because she had a friend working there she could call if she got truly stuck, but the initial process was driven entirely by her and the AI. This is a perfect example of what we call “vibe coding”—translating a feeling and a goal directly into a functional product.

Iterating with Plan Mode and a Beginner's Mindset

A huge unlock for Bryce was discovering Replit’s “plan mode.” Initially, she would give direct commands like, “let’s change the progress bar to a circle,” which sometimes led to “bananas” results that she’d have to quickly undo. By switching to plan mode, she could have a more collaborative conversation with the AI.

A detailed view of the Replit IDE, showcasing a live preview of the 'Daily 100' workout tracker web application, alongside deployment logs with troubleshooting steps, and a project file structure containing numerous TypeScript files.

She’d start with a big-picture goal, like, “Okay, robot, here's what your non-technical friend wants to do. How can we collaborate on our new idea?” This approach, rooted in what she calls a “beginner’s mindset,” was her superpower. Instead of being limited by what she knew to be possible, she was free to ask for anything. The immediate feedback from Replit’s preview panel let her see if she was on the right track, allowing for rapid, intuitive iteration without ever needing to become a software expert.

Workflow 2: Generating Custom Workout Videos with Gemini and Higgsfield

Once the app was functional, Bryce started sharing it with her network. The feedback was immediate: people didn't know how to perform exercises like a “Superman” or a “reverse lunge.” She needed video demos, but she didn’t want to be the star of low-res videos filmed in her living room. The solution? Anthropomorphic animals doing the exercises for you.

This workflow is a brilliant combination of creative prompting and tool-chaining to produce something truly unique.

Step 1: Create the Character with Gemini

Bryce's process starts with generating a high-quality static image of the animal in the correct starting position. For this, she uses Gemini. She learned that being hyper-literal and precise in the prompt is the key to success. During the episode, we generated a leopard for bicycle crunches, and her prompt was incredibly detailed, drawing on her past experience as a barre instructor.

Here’s the prompt we workshopped:

Create an anthropomorphic leopard in a gym setting, wearing exercise gear. The leopard's hands should be behind its head, elbows out to the side, head resting down on the mat. The head should be to the left and the feet should be to the right. The knees should be positioned above hips and feet forward in a tabletop position. Both feet off the ground. There should be no other characters in the image.
Demonstrating specific image generation with a detailed AI prompt in Google Gemini, showing a cougar in a gym based on precise pose instructions.

The first attempt wasn't perfect, so she rewrote the prompt from scratch, adding more clarifying details rather than just editing the old one. She believes this “resets the karma” with the AI. The second attempt was much closer to what we needed.

Demonstrating AI image generation with Google Gemini: A detailed prompt for an anthropomorphic leopard in a gym, showing the AI's understanding and refinement of instructions to create the desired image.

Step 2: Record the Motion Reference Video

This is the most low-fi part of the process. Bryce simply uses her iPhone to record a short video of herself performing the exercise. The key is to match the orientation and general form of the static image she generated in Gemini. For the crunches, she filmed herself in the same tabletop position with her head oriented to the left.

Step 3: Merge Image and Video with Higgsfield

Next, she brings the two assets together in Higgsfield, a platform with an “embarrassment of riches” in AI models. She specifically uses the Motion Control feature with the Kling model.

The process is straightforward:

  1. Upload her motion reference video (e.g., crunches_cut.mp4).
  2. Upload the static character image generated by Gemini (e.g., leopard_supine.png).
  3. Set the “Scene Control” to “Image” so the final video uses the gym background from the Gemini image, not her living room.
  4. Click Generate and wait.
A detailed view of the Higgsfield AI UI, showcasing the 'Motion Control' feature with an AI-generated humanoid cougar character in a gym setting. The interface displays options for model selection (Kling 3.0 Motion Control) and scene control mode for background generation.

The result is a high-quality, seamless video where the AI-generated leopard is now performing the crunches from her video. The model intelligently animates the character based on the motion data, even adding realistic details like reflections in the gym mirror. It’s a workflow that would have required a professional animation studio just a year ago.

Workflow 3: Shipping to the App Store Using Claude as a Technical Co-pilot

Having a great web app is one thing, but getting it into the App Store is the holy grail for a consumer app. This was the final boss, and everyone Bryce asked, myself included, initially told her she’d eventually need to hire a technical expert. But between the fall and when she was ready to tackle this, the models got so good that the advice changed: “You can probably do this yourself.”

The Claude-Claude Code-Terminal Loop

Armed with her beginner’s mindset, Bryce went to Claude and laid out her situation with complete honesty:

“How do I prepare a Replit app for app store submission? I am not technical.”

This led to a powerful workflow where she used two different versions of Claude as a team:

  1. OG Claude as the Architect: She used the standard version of Claude as her project manager or technical architect. It gave her a high-level, step-by-step plan, explaining the concepts and what needed to be done.
  2. Claude Code as the Engineer: When the plan required code, she would take the specific task to Claude Code. It would act as her dedicated software engineer, writing the necessary scripts and code blocks.
  3. Back to OG Claude for QA: She would then bring the generated code back to the original Claude instance to verify it. “This is what you told me to have it do. Here’s what it did. Is this right?”
  4. Terminal for Execution: Finally, with confirmation in hand, she would copy and paste the commands directly into her computer’s Terminal to execute them. This process involved migrating her app off Replit's comfort blanket and onto a hosting platform called Railway.

Conquering the App Store Review

This workflow got her all the way to submitting the app. Her first submission was rejected, but instead of panicking, she simply copied Apple's feedback and pasted it into Claude. The AI helped her troubleshoot each issue:

  • Issue 1: A misconfigured setting for age restrictions. Claude told her exactly which box to check.
  • Issue 2: The “Sign in with Apple” feature was broken. It was user error—she had implemented it but never tested it. Claude helped her debug and fix it.
  • Issue 3: The app needed a way for users to delete their accounts. Claude helped her add the required button and functionality.

After a weekend of focused work—about 25-30 hours—she resubmitted, and Daily Hundred was approved. It's now live in the App Store, a testament to this incredible, AI-powered workflow.

Final Thoughts

Bryce’s journey with Daily Hundred is one of my favorite stories we’ve had on the show. It perfectly encapsulates the current moment in AI. The three workflows she pieced together—building the app with Replit, creating stunning custom media with Gemini and Higgsfield, and navigating the App Store with Claude—show that the biggest barriers to creation are no longer technical skills. They are creativity, persistence, and the willingness to ask for help, even if you’re asking a robot.

Her unabashed beginner’s mindset wasn’t a liability; it was her greatest asset. It allowed her to explore possibilities without the constraints of knowing how things are “supposed” to be done. I hope this episode inspires you to look at your own ideas and realize that the tools to make them real are right at your fingertips. Go download Daily Hundred, get your reps in, and then go build something amazing.

Thank you to our Sponsors

A special thanks to our sponsors who make this show possible:

  • WorkOS—Make your app Enterprise Ready today
  • Metaview—The Agentic Recruiting Platform for winning teams

Where to find Bryce Rattner Keithley:

Where to find Claire Vo:

Tools Referenced:

Start shipping
better products.

Join 100,000+ product managers who use ChatPRD to write better docs, align teams faster, and build products users love.

Free to start
No credit card
SOC 2 certified
Enterprise ready