This Pharmacist Created A Fitness Tracker that Adapts As Your Body Changes
Yim Yoon-seok (임윤석) spent his evenings hunched over a spreadsheet, toggling between Excel tabs and KakaoTalk windows. Each tab represented a client who needed a custom macro breakdown and a calorie target recalculated weekly. He was a pharmacist four days a week, a bodybuilding nutrition coach the rest of the time, and a YouTuber somewhere in between. The coaching side was growing, but the work behind it was entirely manual.
For every new client who signed up, he opened a fresh spreadsheet, typed in their weight and body fat estimate, ran the formulas by hand, and messaged them their plan on KakaoTalk. When they sent back photos of their meals or screenshots of their scale, he logged those too. Multiply that by dozens of active clients, and the math stopped working. He was spending more time on data entry than on actual coaching.
The Problem: Knowledge Without Delivery
He needed a web application that could handle intake, tracking, and adaptive calculations automatically. Something his clients could log into on their phones, input their data, and get real-time feedback without him manually crunching numbers every night. He had never written a line of code.
Before February 2026, his experience with AI was limited to asking chatbots questions. They were useful for research and brainstorming, but they were conversational tools. They could not actually build software. He found Manus through an internet advertisement in mid-February. It was his first encounter with an AI agent, a system that could write and deploy code on his behalf. He decided to try building the app himself.
Building Without Code
The first week was a blur of trial and error. He had no mental model for how software development worked. He did not know what a framework was, what an API call looked like, or how a database connected to a front end. He learned by doing, feeding instructions into Manus and watching what came back. When something broke, he described the problem and tried again. When he encountered complex development bottlenecks, he learned how to refine his instructions, providing the agent with clearer context to resolve the issue.
By the end of that first week, the core framework of Alyak Coach was live. Clients logged in on their phones and entered their meals directly. The system tracked calories, weight progression over time, and workout history. It synced with Strava and HealthSecret so that step counts, exercise data, and calorie logs flowed in without manual input. At the center sat an AI-powered TDEE (Total Daily Energy Expenditure) calculator that recalculated every week based on each user's actual weight changes and intake data. For a bodybuilding nutrition coach, this was the feature that mattered most. Most calculators estimate TDEE once using a static formula. His version adapted dynamically week over week, responding to real changes in a client's metabolism rather than relying on a number set months ago.
Learning to Work With an AI Agent
Over time, he developed a workflow that other Korean power users had independently arrived at: use conversational AI tools for initial brainstorming and structuring his ideas, and then leverage Manus's advanced agentic capabilities to execute and deploy the actual code. The chatbots helped him articulate what he wanted, and Manus turned those articulations into working code.
He learned that specificity made the difference between a wasted session and a working feature. When the AI hit a wall on a bug, the fix was always to provide more context about what the feature should do and what data was flowing where. With enough context, the agent was consistently able to resolve even the most stubborn bugs.
"It was comparable to working with a PhD-level researcher-developer hybrid," he said. The system grasped specialized terminology like metabolic adaptation and progressive overload protocols without him needing to translate his coaching knowledge into layman's terms. He could describe a feature the way he would explain it to a fellow coach, and the output matched what he meant.
Growth and Revenue
He started charging users in the third week. His existing coaching clients migrated over first, grateful to stop sending KakaoTalk messages and start using a proper interface. Then his YouTube channel kicked in. He had an existing audience interested in bodybuilding and nutrition, and when he mentioned the app in his content, signups accelerated. Word of mouth carried it from there, spreading through gyms and group chats organically.
At the time of his interview with Manus, roughly two months in, the app had around 300 active users. By April 20, the backend showed 3,900 registered users and millions of KRW in revenue. The numbers were still climbing.
From Spreadsheets to Scale
An external developer had estimated the project would require six months of full-time work. That would have meant hiring someone, paying them for six months, and watching his coaching business stay stuck in spreadsheets the entire time. He built it in weeks instead, on his MacBook at home and on his smartphone in his spare time. During break times at the pharmacy when there were no patients, he would pull out his phone, notice a bug, jot it down, and feed it to Manus that evening. The mobile-first habit mirrored how his own users interacted with the product, which made him a good tester of his own app.
Today, Yoon-seok manages a growing user base of nearly 5,000 people without opening a single spreadsheet. Every one of those users is someone tracking their weight, hitting their macros, and working toward a body composition goal that used to require a personal coach messaging them directly. Now the system does it for them, adapting every week automatically. The spreadsheets are gone, the coaching finally scales beyond his own hours, and Yoon-seok gets to focus on what he actually cares about: helping more people get stronger.
Ready to build something amazing?
