I built FearCutter, an AI interview prep tool, in just 5 days
Built completely on my own with Windsurf, an emerging AI coding agent
I built FearCutter in just 5 days. It’s live, production-grade, and self-serve. This process would have typically taken an engineering, product, design team ~2 months, but with the help of new AI tools like Windsurf, I was able to build and ship it completely on my own.
The start of the project was challenging and scary, especially if you haven’t been in the game for a while like me. I was rusty. I don’t have much front-end engineering experience. I don’t know React or TypeScript well, and I haven’t coded much in a while. I haven’t directly dealt with full stack app deployment. I never implemented an authentication feature before. I certainly never built an entire app by myself that was production-grade, self-serve, and generally available, ready to sign on customers.
Yet I was able to achieve this on an accelerated timeline. Mostly due to 2 factors: 1) AI code agents can make multi-file edits across your codebase, speeding feature implementation and reducing time making changes that affect other files; and 2) because I’m a one-man team, I am the vision, command center, and foot soldier, all at once, with no information loss at each communication link.
You’d think that taking on multiple roles would be a monumental task because it requires different specialized skills. But it’s actually incredibly efficient: no communication overhead, no context switching. As the product manager, I define the what and why. As the designer, I shape the user experience down to the pixel. As the developer, I build it directly. The tight feedback loop between product vision, design, and development more than compensates for juggling the roles. It is actually way more fun and fulfilling compared to my experience as a CEO/founder where I was more hands-off in day-to-day implementation of most features.
Using Windsurf felt like having the productivity of a small team, connected via mind meld. My take on this: One developer + an AI agent will outperform a team of 2-3 people, especially for MVPs and early-stage features. I can only imagine what one developer + an army of AI agents will be able to do.
For the motivated and competent solo entrepreneurs with big ideas but capital constraints, this new way of building software is game-changing. It will help experiment and ship MVPs faster at minimal cost, eliminating the inefficiency and frustration of working with misaligned teams, shitty contractors, and unreliable dev shops. AI coding agents bridge vision and product delivery like no other tool — you can build what you want, exactly the way you envision it, extremely fast.
While FearCutter only focuses on a niche market, it is just a start. The same approach and tools I used to build FearCutter can be used to build countless other products for bigger markets in the future, and the process will only get more efficient.
This post is split into multiple parts.
The main post (the one you are on), with reflections and takeaways on building with AI coding agents
My detailed product building journey, including defining, designing, and implementing the product over a 5-day coding sprint
Full notes on Windsurf, including observations, bugs, and feature requests
Journey
FearCutter is an app that helps product managers (PMs) practice job interviews. I decided to work on this side project because I noticed how hard it is to get high-quality interview practice. I also wanted to try out new AI code editors in the market and bring my product-building skills up-to-date.
About a week prior, I stumbled upon a new AI code editor called Windsurf, created by Codeium. Their “Cascade” feature looked intriguing so I decided to give it a try. I wanted to build something useful, and as someone currently preparing for PM interviews, I thought:
“I can build an app much better than anything else in the market for this”.
My initial goal was to build and deploy a working app faster than it takes to schedule a single mock interview session — which usually takes a week or longer and costs hundreds of dollars! (I kid you not). While building this, I wanted to focus on the fun stuff, like product design, UX, systems architecture, and let AI handle everything else.
Here’s roughly what the 5-day sprint looked like. Each day was a different adventure and I’ve made a more detailed post about the challenges, frustrations, and joy that I went through here.
Day 1 - Foundations: Defined the problem, outlined features, and designed the initial UI/UX and data model. Made foundational tech stack decisions and laid the groundwork for development. 0 lines of code.
Day 2 - MVP: Setup my app and created a basic chat interface with a real-time markdown editor.
Day 3 - Features: Built user authentication, registration, and interview session management.
Day 4 - Quality: Improved the AI interviewer's prompt to improve question quality and personalization.
Day 5 - Deploy: This was a day of intense debugging and deployment. The app was deployed to production!
For a more detailed and colorful version, check out my full post on the entire 5-day journey.
Reflections and Takeaways
Do you really need a team to prototype?
Conventional wisdom says you do. If you start out as a solo founder, you may struggle to find people to build with you, and end up picking anyone who is willing to work with you despite various misalignments. However, with the help of AI agents and sufficient motivation, this initial struggle and team set-up will become unnecessary. A one-person team will be significantly better than a small team, especially at the very early stage for building MVPs to validate a market.
How will AI agents impact founders and developers?
Founders++, Dev shops--.
Huge win for bootstrapped founders, with low capital who want to build out MVPs and experimental early features. AI agents will help prototype and ship extremely fast. No software experience? Learn some web foundations and get Windsurf. No money to hire a dev shop to build it for you? Same answer. I only backend and don’t know React? Same answer again. There’s really no excuse now. As for dev shops and unreliable contractors? Tough luck, although the same answer might apply here too. As for developers working at companies, if you don’t have access to these tools, you will be less competitive so get your managers to buy it for you.
Are Windsurf or other AI coding agents worth it?
Windsurf’s Cascade is incredible. It’s a game-changer. Its Write mode, in particular, was effective at making multi-file code changes across my entire codebase, which is great for grunt work, like updating code following a database schema migration or refactoring a React component used in multiple places. Of course, there were limitations, but the value it provides is totally worth it. For devs, read my post on my more detailed notes in Windsurf in my post here.
Will the role of the PM change as a result of AI code agents?
Definitely. The output of a PM will evolve beyond traditional PRDs (product requirement documents and slide decks. Instead of just conceptualizing, defining, and prioritizing, they will actively ship the first basic working version of a feature or product to secure internal buy-in. Currently, a PM might gather user feedback, conduct market research, draft a PRD, and collaborate with design and engineering to release an initial product version. In the future, I expect PMs to perform the tasks of an entire small Engineering, Product, and Design (EPD) team during the early iterations. Instead of writing a PRD, they might write an "ARD" - an AI Requirements Document - that guides AI agents to design and engineer products based on user needs.
Will new roles emerge?
Yes. These emerging roles don't have names yet; let's call the role X for now. The person filling this role doesn't necessarily need to be a PM. It could be a tech lead, an engineering manager, or even a member of the sales team advocating for a feature that's crucial for securing a million-dollar enterprise deal. The number of individuals and functions capable of shipping a v1 to gain buy-in will significantly increase. The formal title is less important than the required skill set, which includes deep user empathy, product vision, technical systems design, and taste. And, of course, AI.
How can PM interviews gauge candidates deeper?
Using FearCutter, I realized something: typing out my answers actually improved my product thinking. This makes sense: the best product thinking happens with nth-order thinking, which requires time and writing, and not by blurting out the first thing you can think of. This made me question the current state of PM interviews, which are almost entirely verbal. We need more written assessments that test product sense and judgment, in a deeper, more thoughtful way, instead of the current quick and shallow verbal way. Both approaches have value, so combining them could result in better hires.
Caveat
There have been lots of early reaction posts about AI code editors on X. Recently, someone posted about their frustration with AI agents, which is a common roadblock/downfall for a builder without technical foundation:
While AI agents like Windsurf’s Cascade help you get pretty far, the last 30% could indeed be frustrating. You still need some good foundation. You need to at least be able to read basic code, understand how web apps and the internet generally works, how clients and servers talk to each other, and how to generally design systems.
Why is the technical foundation pivotal in determining how much boost you can get from AI? Because AI doesn’t have the vision, motivation, taste and craft that you have. It doesn’t deploy and it cannot ship on its own. The process isn’t a sequence of AI deploying the first 70%, then switching to you complete the remaining 30%. Instead, it’s more of a constant refining process of directing and guiding AI to do the work, but faster, more diligently and tirelessly than if you had done it yourself.
Evidently, the future isn’t quite here yet, but it isn’t too far away either. AI doesn’t do everything and you still need to be technical enough to direct it. AI coding agents like Windsurf and foundational models that are great at coding will lead us there.
Conclusion
In just 5 days, FearCutter went from concept to a live, production-grade AI interview prep tool, a feat typically requiring months and a small team.
This accelerated timeline was only possible with an AI coding agent like Windsurf. New AI coding tools create a new paradigm for software building: a solo builder can achieve extraordinary efficiency by having a unified product vision, design, and implementation, and do it all with no communication overhead.
Windsurf, particularly its Cascade feature, is a delightful UX for code editing across your codebase. It greatly accelerates feature implementation. While a solid technical foundation remains crucial, AI agents like Windsurf will empower motivated people to build and ship MVPs at an unprecedented pace and minimal cost.
FearCutter is likely the first app deployed in production that’s self-serve and ready for users and paid customers that is fully built with Windsurf. The future of building is here, and it's faster, more efficient, and more accessible than ever before.
Other related posts:
