It’s finally time to catch up on what has been happening behind the scenes. First, I want to apologize for leaving some of my previous posts without updates. I’ve been deep in the trenches of learning, and I’m excited to share what I’ve discovered.
Where to start? Honestly, trying to keep pace with the rapid evolution of AI made me lose track of time. For someone who learns by doing, failing fast and fixing errors, I became deeply curious about how AI could enhance my personal skills and transform my day-to-day work in product management.
Choosing to Learn About “Vibecoding”
I wrote about AI Product Manager: A New Breed for a New Era, how the modern PM is starting to leverage AI agents to build things. Diving headfirst into that world is exactly where I lost track of time. As I tried to organize what specialty I wanted to pursue next, I landed squarely on Vibecoding.
With vibecoding, I can rapidly turn the ideas in my head into working prototypes. For a Product Manager who entered product development with a background in Graphic Design, this has been a game-changer. Being able to spin up an MVP to present to a UI designer or a leadership team makes proposing new features or optimizations incredibly seamless.
But before I could do that effectively, I had to understand how the underlying tech works. I used to struggle with engineering lingo, which sometimes created a communication barrier. Whether it was understanding how information travels from the product interface to the database or how APIs communicate, I knew I needed hands-on experience. Vibecoding provided the perfect sandbox to finally learn how frontend elements integrate with a backend database by building a product completely from scratch.
Straykin – No Stray Left Behind
To truly understand how to build a fully operational product, I started a side project called Straykin – No Stray Left Behind. It is a community-driven platform designed to help stray animals in local neighborhoods.
Initially, I envisioned an all-in-one native iOS or Android app. However, given my technical limitations, a Progressive Web App (PWA) was the much smarter route. I turned to vibecoding, guiding AI models through instructions and prompts to construct the application.
Of course, when you are starting, the initial results are rarely perfect. Success comes down to how effectively you prompt and utilize the tools. Because I needed a frictionless way to handle both the database and deployment. I chose Google AI Studio and Google’s ecosystem. The generous free tier made it a no-brainer for development, and it significantly streamlined the integration process between my front end and back end.
Goals and Purposes
Building Straykin forced me to live the entire Product Life Cycle and experience . What it feels like to run a startup from absolute zero. Revisiting core principles like the Golden Triangle of Product Ideation and Product Development 101 reminded me how crucial it is to keep foundational frameworks in mind, even when using cutting-edge AI.
My goal was to create something functional that could genuinely serve a community, not just sit as a concept file on my laptop. Despite navigating a mountain of unexpected technical challenges, Straykin is finally live.
It has its limitations, but it works. You can try it out and see the results of this vibecoding journey firsthand by visiting: Straykin – No Stray Leave Behind.
Challanges
If we talk about the challenges of bringing Straykin to life, there are too many to list, ranging from platform limitations to managing API costs.
When I first started development, I thought I could manage multiple features inside a single prompt. But I quickly realized that AI development doesn’t work that way. As your project grows, every time you ask the AI to optimize a feature, it has to re-read your existing files and previous chat history to understand the context.
If you aren’t careful, your token usage rolls downhill. I call this the “Snowball” Effect of Context:
- Prompt 1: Your Question + 1 Code File = 5,000 tokens
- Prompt 2: Follow-up Tweak + 1 Code File + Previous AI Response = 10,000 tokens
- Prompt 3: Another Tweak + 2 Code Files + Entire Chat History = 20,000+ tokens
Before you know it, you are hitting daily token limits or burning through costs just to make a minor tweak. My technical shortcomings forced me to learn a crucial product skill early on: how to scope prompts effectively and optimize one micro-feature at a time rather than dumping the whole codebase into the AI.
My Designer Advantage (and What’s Next)
Building a product from scratch requires balancing four pillars: Technicality, Creativity, Business, and Marketing. Right now, even though Straykin is live, I am heavily focused on the first two.
Why is marketing taking a backseat for now? Because my background is in Graphic Design and Creative Management. Designing user flows, thinking about visual hierarchy, and managing the creative direction it me forte. it’s my comfort zone and my secret weapon.
Because I don’t have to struggle with the creative side, I can channel my energy into the technical execution. Vibecoding acts as the perfect bridge, allowing me to learn backend logic, deployment, and database structure naturally through hands-on doing
Closing statement
That is all I can share for now. However, I am seriously committed to continuing this journey. Whether Straykin evolves from a PWA into a native Android or iOS app, or grows into something even bigger, is something I can’t predict just yet. But one thing is certain: I will keep building and developing this platform until I hit a major technical bottleneck, one that finally requires bringing in outside expertise to scale.
