Software has never been about the code. It’s about what software can do for us.

AI has widespread adoption because even though it’s imperfect, it achieves the results we are looking for more often than not.

When I open the Krispy Kreme app, if I see the “hot now” neon sign lit, and I drive to the store, more often than not hot fresh donuts will be waiting for me. That’s what software is. It doesn’t mean I haven’t been misled from time to time, driving all the way to Burbank only to have my dreams crushed, again!

And so, some AI-written software will work perfectly until it gets hacked. Legacy systems got hacked every day, before AI.

The best coders out there can surely write better code than AI today. But it’s not like all human coders have always written perfectly formatted, immaculate, completely legible, maintainable, secure, and bug-free code since the dawn of computers, right?

The scoreboard/statistics will start to matter more over time. Will more AI-coded apps get hacked than human apps? Will AI apps have more or fewer data breaches on average than those written without AI? Will self-driving cars have more or fewer accidents than human-driven cars?

At a certain point, AI-generated code (that few humans understand) might have fewer security incidents than code written by the top security experts. This doesn’t mean that AI-generated code won’t get hacked or cause massive security incidents.

Like any tool, AI coding agents will be wielded by geniuses and idiots alike, each getting wildly different results from the same products.