It’s my thirteenth post, and while I'm thinking on the new wall sprites design, I’d like to share my own personal AI “policy” and perception.

I saw a lot of projects, from indie-driven to large enterprises, where AI has chewed everything. No one cares anymore about thinking, engineering or debugging. You got an idea? Ask Claude to code it for you. You got a bug? Ask ChatGPT and blatantly copy-paste its generated code. Why should you care to ever review it when Copilot can do it for you? And if the AI-generated code produces bugs… just make AI to fix it!

Sound tempting, right? It shouldn’t really surprise that so many devs completely turned in to vibe coding. AI-infected code all around us became our new unspoken pandemic. I believe it’s much worse than Covid-19 in terms of contagiousness, and shall break out in much severe chaos in a not-so-distant future than any human virus we had to deal with by the day.

I truly hate this modern trend. As an indie developer, I don’t want to give up my mind and skills. Call me a luddite, but even in the end of 2025 I prefer to spend hours of my free time to write code on my own rather to take some AI vomit. I believe it lets me preserve control over my concepts, ideas, designs, architecture. I have to keep a lot of things, tiny and major architectural decisions and reasons for them in my mind. Probably it’s much less efficient, but honestly? I simply love coding and architecting.

I understand this stubborn approach shall delay my delivery for months, and I’m sorry for that price. But this is my way of working, and I hope it allows me to bring much more quality than most of the software products are created nowadays.

When I do use AI

It would be completely dumb of me to ignore the sophisticated assistants in 2025, of course.

  • ✅ I consult ChatGPT when I need multiple fresh design ideas to pick between.
    • This is how I got to know about ECS, for example.
  • ✅ I report to ChatGPT my status updates.
    • …which might sound weird for such the AI-hatred manifestation I’ve done above. But as an indie, I find it’s perfect for moral support and to keep myself on track, with “someone” gently guiding me when I stuck at some overcomplicated decision.
  • ✅ I allow ChatGPT to generate code for some boring but simple stuff which I don’t really want to waste my time on.
  • ✅ I consult ChatGPT on the language and framework specific features.
    • Most of the time it’s just quicker than googling.

When I don’t use AI

  • ❌ Even though I consult AI about my ideas, I force it to propose several ones, and prompt it hardly to find out all possible caveats. In the end of the day, I come up with some combined approach on myself.
    • I never take any AI output for granted, and never use it directly in any of my work.
    • I treat AI output as a StackOverflow or randomly googled article rather than any kind of source of truth.
  • ❌ I never allow it to write any automated test code, or even to think about test scenarios.
    • Unit tests, functional tests and other layers of automated testing comprise a quality gate. I am the gatekeeper, and take the full moral responsibility and accountability for the code I deliver. That’s why I make this part of work ultimately on my own.

Fraction of AI-generated code

Counted by lines amount:

  • Product code: 3%.
  • Test code: 0%. Zero. And always will be so.

What about this devlog?

Heh, you virtually caught me up! I must admit that for some posts I ask ChatGPT to create some drafts. Let me be honest: I’m a tech geek, not a very talkative person in real life, nor a fan of writing. (And I’m not a native English speaker.)

But I never copy-paste GPT directly. All its output gets through my heavy editorial.

I really didn’t measure a fraction of AI-generated text, but I assure that I spend 1-2 hours on every post published here.

Hopefully I’ll find your support and understanding of my attitude to AI, my followers. I’ll do my best to make you happy with a high-quality product enjoyable to play and to mod.