Everyone talks about how fast AI is improving, but never mention how fast humans can improve

AI is in the news everyday. Everyone is always talking about the insane rates of improvement it's making, and how quickly it's improving. There is always an air of fatalism; as if it's predetermined that one day, and one day soon, AI will outpace you.

I'm here to tell you that you can outpace AI.

I've been building Phrasing now for over three years. At the moment, the entire app is built by hand. No part of it was vibe coded.

However, there was a prototype before the current version that was sort of vibe coded. At the time, the term didn't exist. The best model (and more or less only available model) was gpt-3.5.

I didn't know how to set up a database. I didn't know how to set up microservices or background tasks or data processing pipelines. I had some idea, as I'd done these before when I was learning to program, but in the same way you could kind of reconstruct a recipe you haven't cooked for 10 years.

At the time, this felt like magic. Like it was already so far ahead of me and had somehow lapped me. But then the cracks started to show; the in cohesion of the app, how utterly non-maintainable it was, how expensive and impractical it was to run. 

It was unsalvageable. And so I rewrote the entire app myself, without AI.

Over the past 3 years, AI has gone from the initial launch of chatGPT, to whatever it is now.

Every month there have been "insane breakthroughs" and showing "no signs of slowing down".

However, over the last 3 years, my abilities have far outpaced that of AI. In the past few months, I have:

• Architected, setup, managed and maintained multiple new data pipelines

• Created and implemented a performant sync engine to support offline mode and instant loads

• Created my own GPU accelerated charting library to show data from hundreds of thousands of reviews at once

This might sound like gibberish to the non-technical user, as it would have sounded like gibberish to me three years ago. The fact of the matter is these are hard problems, that AI's could not do today.

I've asked LLMs several times throughout each process. Of course I wish there was an easy button. I've asked it to help me brainstorm architecture. I've asked it to help me solve bugs. I've asked to help me write tests. 

Every single time, it's responses are garbage. Had I followed any of its advice, I would not have been able to achieve the performance I have now with Phrasing.

I went from looking up to the generated code of gpt-3.5, to looking down at the code and decisions made by even Claude Mythos/Fable.

I outpaced AI.

And here's the thing: AI is showing massive signs of slowing down. The internet is clamoring over GPT-5 being worse than GPT-4.5. Opus 4.7 being worse than 4.6. Most people these days couldn't tell the difference between any two versions. Every new version in incrementally less impressive than it's predecessor.

But the human spirit does not slow down in the same way.

We are all capable of far more than we imagine. If you're willing to put in the time, put in the work, and actually struggle, there is no upper limit to what you can learn. There are not diminishing returns to your knowledge. There are not rate limits to your abilities.

Yes AI will be able to list more facts than you, sure. It will be able to write faster than you and read more in a minute than you can read in a lifetime.

But none of that holds a candle to what one person can do when they put their mind to something.

You can outpace AI. 

Don't let anyone convince you otherwise.