I have been playing with AI in one way or another for a while now. I have read a ton about it and thought even more about its implications — what it will do to this world, what it will do to my career, what it will do for our future.

What does it actually mead to use AI?

Comming soon

AI deleted my data and I am the one to blame

I have read plenty of posts from people describing how an AI tool deleted their data. It always made me wonder the circumstances that lead to the failure — until it happened to me.

Now that AI is opening all sorts of doors to building software to your exact specifications, I have been using it more and more to satiate my curiosity. In this particular case, I was writing something for my own personal gratification — a side project I had been slowly building on over the past few weeks. A new feature here, some code cleanup there. Nothing major, until I added a filtering function to the UI. That decision came back to bite me.

What ended up happening was that the AI treated the filtered dataset as the entire dataset when writing to disk, which overwrote all of the data that had been hidden behind the filter. Needless to say, I was supremely annoyed to find that data gone on the backend — all because the AI made the wrong assumption about which dataset to keep.

I say all of that to say this: no matter what the AI got wrong in the code, ultimately I am the one to blame. I was the one who decided early on to write data to a simple file instead of something more robust. I was the one with a poor backup strategy. I was the one who blindly trusted every piece of code being generated. I was the one who failed to put the right protections in place to prevent the AI from touching the filesystem carelessly. It was me. I was the root cause, and AI was the tool doing my bidding.

This is not the first time I have done something shortsighted by not thinking through all of the potential consequences of my actions. But like every time before, it was a valuable learning opportunity — one I will carry with me for the rest of my career.

What I am doing differently:

• Backup everything. A simple backup, no matter how small the code or data, before each save would have made this a five-minute recovery instead of a permanent loss.

• Better AI guardrails. Ensuring that any operation available to the AI is appropriately isolated and guarded from dangerous paths, preventing it from doing errant things even if by accident.

But the deeper lesson is about trust. AI-generated code should be treated with caution: It looks right. It runs. But you're not 100% sure it is correct. And because you didn't write it line by line, you can't know implicitly it will do what you expect it to. That gap between "it looks fine" and "I understand exactly what this does" is where problems arise. When you are moving fast, adding features, and trusting the AI to handle the details, that gap quietly grows wider. The code that deleted my data wasn't obviously broken. It was a reasonable interpretation of the instruction provided — and I never questioned, or frankly cared, whether the code was fully working or not.

That will change.

You have nothing to fear

Comming Soon

Why the future is bright

Comming Soon