INTRODUCTING
With total creative freedom, I've built a narrated sleep stories app for curious minds that won't shut off when it's time to sleep.
A world to drift into…
An idea I couldn't leave on the drawing board.
What began as an exploration of better sleep stories became an opportunity to build an entire product independently, from the first sketches to the systems that bring each story to life.
THE AGENTIC STACK
A considered mix of design, AI, audio, and analytics tools that I'm learning to use, making Shai possible as a solopreneur.
Claude
Across the full Shai lifecycle, my primary collaborator is Claude. On the engineering side, I direct Claude Code to build and audit the native Swift iOS codebase, then review its findings for stability, App Store readiness, and scope control. On the content side, a staged story pipeline with defined roles (ideation, research, outlining, drafting, style editing, and tagging) ends each stage at a human checkpoint, so my wife and editor Megan and I make the final editorial calls.
Figma
Figma is where Shai’s design thinking starts. Original concepts for the glass-orb centerpiece, the minimal mood-to-story flow, and the visual language of the story cards take shape there. It also serves as the working space for managing the UX: mapping the listener journey, testing structural ideas, and settling the conceptual direction before anything moves into the build.
ChatGPT
Creative direction for Shai’s intro videos is developed in ChatGPT, through a custom GPT agent harness I built for the purpose. The harness helps articulate each story’s visual storyline and mood board and defines the key shots, which then become the source frames for video generation in Magnific. It acts as the planning layer between a finished script and the final footage, so every video starts from a deliberate creative brief instead of ad hoc prompting.
Magnific
Shai’s visual media is produced through Magnific. A repeatable cinematic prompting framework, derived from each story’s script, sets the setting, motion, and look of every clip. Working from my own travel photography and footage, it yields looping story-card videos and ambient backgrounds that stay consistent and match the narrative.
ElevenLabs
Narration is where ElevenLabs comes in. I developed a rule-based tagging system that governs pacing, tone, and delivery across a story’s arc, and each finished script is recorded in four distinct narrator voices. The platform is also connected to my workflow, so recordings can be produced on demand.
PostHog
Analytics runs through PostHog, paired with a database I set up myself to track user behavior and business metrics. The goal is for decisions about content, retention mechanics, and monetization to rest on observed listener behavior. The first question it will answer is how free listeners convert to paying subscribers once the free stories are live.
