智练 Zep — Your AI Driving Manual
An immersive AI learning product. 30 minutes of guided storytelling lets a non-technical adult complete a real work task — and walk away with something shareable.
Brand Palette
Mint
#BDE6D9
White
#FFFFFF
Cool Gray
#ECEFF1
Overview
ChatGPT exists, but most non-technical adults still feel like passengers — they use AI without ever feeling in control. Existing AI courses are either too theoretical (LLM mechanics) or too gamified for kids (Google AI Quest). Nothing addresses the 22-to-32-year-old worker who has AI anxiety and wants to actually drive, not just ride.
Process
- 01
Brand & narrative architecture
Built around one extended metaphor: AI is the car, you are the driver, Zep is the manual. A Quest is a road trip. The deliverable is a destination reached. The narrative system is locked before any code — every UX decision flows from it.
- 02
Persona design: 安小白
Zep's guide character — calm, slightly underdressed, has already done the math. Detailed brand-persona document: how she speaks, what she'd never say, what she finds funny. Getting the persona right is more leverage than any feature, because she IS the product's voice.
- 03
First Quest: 30-min competitive analysis
The pilot Quest is a 30-minute guided session that produces a one-page competitive analysis the user can post to LinkedIn or send to their boss. Concrete shareable output is non-negotiable — every Quest must end with something the user can show.
Result
In pre-build narrative-and-brand phase. Three handoff documents (brand persona v0.2, narrative system v0.1, project handoff v2.0) form the brand bible — any new collaborator gets the entire context in 30 minutes. Visual identity locked: deep space green + warm orange + cream, drawing from Studio Ghibli's Princess Mononoke night palette. Engineering starts when the first Quest script is tight.
By the numbers
30 min
Quest length
22 — 32
Target user
3 / 3 locked
Brand docs
Next project
Quant Systems — Computational Finance in C++
A modular C++ options-pricing library covering the three canonical methods — Black-Scholes exact solutions, Monte Carlo, and finite-difference — built across six homework groups in Baruch's Pre-MFE Level 9 program.