The collective awakening to agentic programming
I think the past couple of weeks have probably been the most intensive in terms of AI time contraction. So much happens in so little time, and the collective opinion seems to have shifted. From “yeah, it works and can write some code for me”, to “wow, I can run multiple agents in parallel to run a bunch of things”.
It’s a sort of collective awakening. What I’ve been experiencing myself in the last couple of months seems to have now become the new normal. A year ago, if I told you that you’ll mostly only review code those days, and not write a single line yourself, you would have looked at me weirdly.
Or you would have probably told me that the code the LLM produces just isn't good, and that you'd lose more by having AI assist than by writing the code yourself.
Truth is, now it’s better code than the average Software Engineer, and if you know what you’re doing (driving AI + programming), you can produce an ungodly amount of good-looking code in a fraction of the time it would have taken you before.
In those last few weeks, I’ve seen even the most fervent defenders of “manual coding” or “antique programming” as we know it slowly but surely drift toward agentic programming.
As you may (or may not) know, I’m on Twitter a lot, and here the AI community pretty much agrees that programming and software engineering are already pretty much solved and automated. It’s already our…