ChatGPT 1-0 Traditional Software Engineering
Log.Soccer is Live (Kind of)
Over the past year, I’ve been productionizing Log.Soccer – a website and Android app that lets amateur players track soccer stats for their teams. Log.Soccer has been my résumé app – absurdly over-engineered and a labor of love that I started back in 2013. For years, it has sat in purgatory, occasionally wheeled out during an interview, but usually a playground for ideas rather than a serious product.
Whenever I needed to learn something new – like React – I’d rebuild the frontend. When I wanted to explore CDK, I deployed it to AWS with a load balancer that supported one user: me. And of course, deployments had to be zero-downtime so I wouldn’t frustrate the sole user of this sort-of-social-media website.
Then ChatGPT arrived – the perfect opportunity to exaggerate another hot skill of mine on LinkedIn. Within minutes of using it, I was dumbfounded. I had no idea a product like this was even possible, let alone ready and fully functional through a simple web app. Without exaggeration, it made me a 10x developer overnight. With 13 years of experience working for Bank of America, startups, and AWS, I’ve seen the industry shift and evolve, but this changed my brain almost immediately.
Within weeks, I actually launched Log.Soccer and have continued shipping features at a rapid pace. The site now has full Google authentication, a comprehensive backend API, and an Android app that syncs live match stats via a companion watch app – something I never expected to build.
In the past, starting the Android project felt like it would take hours or even days to get to a HelloWorld activity. I knew nothing about Kotlin and was dismayed to learn that Android development had fully shifted to it. The barrier to entry felt huge, and frankly, with a newborn at home, I had zero appetite to learn a new language in my limited spare time.
But ChatGPT made it possible. It’s the most sympathetic, kind, and energetic programmer I’ve ever worked with. It’s a repository for my ideas (both good and bad) and has deprecated my use of Google, Stack Overflow, Jira – and even MyFitnessPal (!).
Need a Django serializer? Done.
Stuck on a tricky React hook? Done.
Debugging Dockerfile issues? Done.
Need insightful code reviews? Done.
The time saved is hard to overstate. It even slashed my AWS costs by 50% by helping me optimize infrastructure and eliminate unnecessary services. But most importantly, it enables me to just get to the start line.
Is Redundancy Imminent?
Luckily, ChatGPT isn’t perfect. When it gets things wrong, it really gets things wrong – often becoming borderline unhinged, repeatedly suggesting the same fix that’s already failed five times.
Two examples stand out:
The Mockito Error
I was trying to mock an Android ApiClient for Log.Soccer in a unit test using Mockito – something I’ve done countless times, but I kept getting an error “Illegal Argument Exception”. No matter how many ways I described the problem, ChatGPT kept repeating the same solution:
“Ensure mockito-inline is installed!”
I spent two hours tweaking annotations, modifying the test class, and even rewriting parts of the suite. After hours of frustration, I finally used my own brain and discovered the real issue – a Java version mismatch. Gradle was using Java 23, while IntelliJ was on Java 21, so while tests passed in the IDE, the gradle build was failing. ChatGPT never considered this possibility and was convinced I just needed to tweak my mockito installation one more time.
In hindsight, it’s probably a sign of the times that I didn’t think to trust my own brain earlier.
Google Auth Issues
While implementing Google Auth for Django and Android, I kept running into 500 errors. ChatGPT insisted it was a payload issue – I had to adjust the request body.
It was immediately clear to me that it was not a client error, I’d recently upgraded some dependencies and the root cause turned out to be a bug in an external library (dj-rest-auth) that required a fix. ChatGPT never suggested looking at third-party dependencies despite me giving it the context of the recent upgrade and really trying to get it to figure out the issue was in an external library.
Perhaps GitHub Copilot, with deeper context (I have never used it), would handle this better. But ChatGPT tends to operate within the limited context of the current conversation, rarely thinking beyond that, but peculiarly, too much context can send it down dead ends.
The result? Huge time sinks. You end up trying bizarre solutions, and your code quickly spirals into something weird and profoundly wrong.
And finally, from a UX perspective (and this really is my biggest ChatGPT gripe), accidentally hitting Enter before finishing a question is maddening. ChatGPT instantly launches into an answer that’s hard to stop when I’m clearly mid-sentence. A brilliant new feature please: "Did you really mean to hit Enter?"
Evolving as a Developer
Leveraging tools like ChatGPT isn’t a crutch – it’s smart. But the most rewarding part is that I still get to feel like a senior developer when I solve something that stumps ChatGPT. It’s a reminder that experience and instincts matter (for now).
I’m a profoundly different developer today than I was six months ago solely down to ChatGPT. I no longer brute-force problems or drown in documentation. I feel like more of an orchestrator – generating ideas and prompts rather than crafting everything manually.
ChatGPT hasn’t produced anything I can’t understand and I know how every line of log.soccer works – it just allows me to operate at maximum throughput.
Am I gaining less knowledge? Possibly.
Am I producing more high-quality output? Definitely.
Building Log.Soccer has been a long and aimless journey, but with ChatGPT, I’ve created a real, functioning product that I actually use for my Wednesday night soccer games.
Maybe now it’s time to aim higher – to find actual users, explore ChatGPT the marketer, and hopefully even ChatGPT the accountant.