The Real Problem: The Grind and The Dream
Like many of you, I'm tired. Tired of the corporate grind, tired of being the solo-dev bottleneck, and most of all, tired of watching my own projects languish because there's "always something to do" for the company. The dream is to break free - to work on my own things, on my own schedule, at my own pace.
That's the motivation behind DinkySub (Hopefully, my ticket out!).
I had an idea for a tool and a goal: ship a beta, fast. This time, I decided to tackle the project not alone, but with my new AI partner, Kiro. I had worked with Kiro on my previous project, WorryBox, where I experimented with letting Kiro do all the code. For WorryBox I was just the project manager. But this time I'd be in the driver's seat, using the AI as an accelerant, not a replacement.
🛠️ The Good Stuff: My AI Accelerant
Using Kiro as a highly skilled assistant, I was able to slash my project timeline dramatically. What I'd traditionally budget as a 6-month side project was beta-ready in just 2 months. Here’s where Kiro was truly invaluable:
⚡ Rapid Scaffolding: Kiro whipped up the initial app skeleton—database connection, user login, and routing—in minutes. I've wasted hours on this using my own boilerplate or heavy frameworks like Sails.js. This alone was a huge time saver.
⚖️ The Architectural Rubber Duck: When I was genuinely stuck between two or three different implementations, Kiro could articulate the pros and cons of each. It was like having a smart, impartial senior dev to talk through the trade-offs.
🔄 Unstoppable Momentum: The best part? Kiro kept the pace high. Because I retained control of the core logic, I felt confident and in sync. The speed allowed me to stay motivated and avoid the burnout that kills most side projects.
💥 The Problems: My AI's Very... Creative Mistakes
This isn't a flawless success story. Kiro's biggest flaw is that it is definitely not business-minded. Our partnership was a rollercoaster, and I had to spend a lot of time double-checking its "confidence."
💰 'Free Subscriptions For Everyone!'
I had a painful experience settling on a payment processor. After Stripe failed to verify me (for reasons unknown) and PayPal instantly closed my account after reviewing WorryBox (no details there either), I landed on Paddle. Honestly, the setup was a pleasure... until Kiro got involved.
I asked Kiro for help with the Paddle implementation. It confidently set things up, added the required keys, and declared it "production-ready!"
It was not.
Kiro had used the completely wrong API implementation. And, hilariously, it added a feature I'm calling the "Auto Accept Fallback."
🚨 Kiro's "Business Logic": If the payment method fails, just automatically accept the subscription anyway, requiring no payment.
Yes, my AI co-pilot tried to turn DinkySub into a pro-bono passion project!
I found the correct Paddle documentation and was pleasantly surprised to find Kiro could read the URL I provided and restructure the code. However, it stubbornly reintroduced the "Auto Accept" fallback, and then, after I asked it to remove the fallback, it put back the original, non-working implementation of paddle!
After several hours of manual surgery, I stripped out Kiro's code and installed the correct payment implementation myself. Payments finally worked. (Until I asked Kiro for help on the analytics page, and it managed to re-break the payment code again by trying to sneak in its favourite, incorrect method!)
🐛 The TypeScript Tantrum
The second major frustration was Kiro's penchant for creating build-breaking TypeScript errors. Time and time again, I'd ask for a feature, and the resulting code would fail to compile due to a TS issue.
The frustration was real, but the defence is simple: Kiro was also the solution. I just had to say: "You caused build errors. Please fix them."
And Kiro would dutifully burn through my credits, building, testing, and fixing its own mess until the code was clean again. A very expensive, but reliable, QA process!
✅ Overall: AI as an Employee, Not a CEO
Working with Kiro as a coding partner is certainly an adventure. It is lightning-fast and great for scaffolding and architecture, but requires a human with a healthy dose of suspicion and an eye for business logic.
The net result is undeniable: A 6-month project was finished in 2 months. And by using Kiro as an assistant instead of asking it to "do it all" (my mistake with WorryBox), I mostly managed to keep a handle on my credit usage.
If DinkySub takes off, I will definitely subscribe to the paid tier of Kiro. I might even list it as an employee on the company roster.
🙏 Your Turn: Check out DinkySub!
If you're interested in this journey, please feel free to give DinkySub a try at DinkySub.com
And if you find it useful, a subscription would help a fellow developer get one of those lives you keep hearing about on YouTube and Instagram. You can also support the project and get access via my Patreon page. (Supporters get a DinkySub subscription plus some other benefits.)
With 2 subscribers, DinkySub is self-sustaining.
With 300, I have a good side hustle.
With 1000+, I could potentially give up going to an office. :p
I welcome any comments, suggestions, or feedback! (Though if you have a bug report, please send it - I will deal with it, but I refuse to welcome it 😉).
**What are your thoughts on using AI for side-project development? Have you had any similar experiences with an AI co-pilot making hilariously bad business decisions? Which AI coding tool do you find works best for you?"
Top comments (0)