Thanks for the shoutout, Akos. I had just finished writing the post I'm publishing tomorrow, and I had arrived at the same conclusion.
My comparison is more with smartphones, social media, and the ways of communicating we have right now. We thought that was a productivity boost, but it was a distraction. Instead of doing something meaningful, we created more noise with that communication.
With AI, I think we are going in the same direction. Instead of doing some meaningful code, if we prompt and stop thinking, we're just going to spend more time and deliver less work
If you work in a huge company where everyone uses AI, low performers will crank out brutal amounts of questionable code and then ask for a review. As much as AI coding can be a superpower in the right hands, ultimately, I think it’ll lead to the downfall of many big companies because:
* managers are now creating massive amounts of presentations, reports, and other agendas that developers now have to spend extra time reading through
* low performers can now look like they are working
* small companies with passionate engineers will move fast
I genuinely believe it can go two ways: you could offboard all your thinking to the LLM, or you could explore new things you’ve always wanted to try but never knew where to begin.
I’m trying to stay away from offloading the creative part of the work because I think with time, I could lose my ability to look at problems from a fresh perspective, and maybe offer a new approach – something that has had huge upsides during my career.
As for trying out stuff you never worked with, it’s a huge win. I mean… I just made an application for my Apple Watch without Swift knowledge 🤷♂️
I find AI tools incredibly useful for learning about unknown stuff. But ya, when it takes over the creative aspect of a certain thing, things stop feeling fun.
Thanks for the mention, Akos! Looking forward to your future posts.
Thanks, Saurabh! I'm still looking for the right blend of honing my programming skills, without making me too lazy. Next month, I'm going to give a shot at Codex or Claude code, hoping that the added friction between using a separate tool as a consultant will make me more mindful about how I'm using AI.
dittos. It doesn't help me grow as an engineer. I find the time diff between fixing AI code, and reading the docs to do it right my self, is close to equal. AI has helped for grunt work I know well but less so for new things
I want the AI to guide me in topics that are unknown to me, not to do all work (of questionable quality) instead of me.
Sure, once I find a pattern (refactorings are a great example of that), it's super helpful just to outsource the updates that are more complex than find & replace. But to generate solutions for existing codebases, from scratch... I'll spend just as much time reviewing that is it was someone elses PR.
I'm totally not ready to review 1000s lines of AI generated code.
Thanks for the shoutout, Akos. I had just finished writing the post I'm publishing tomorrow, and I had arrived at the same conclusion.
My comparison is more with smartphones, social media, and the ways of communicating we have right now. We thought that was a productivity boost, but it was a distraction. Instead of doing something meaningful, we created more noise with that communication.
With AI, I think we are going in the same direction. Instead of doing some meaningful code, if we prompt and stop thinking, we're just going to spend more time and deliver less work
Yes, and you know where this shows truly?
If you work in a huge company where everyone uses AI, low performers will crank out brutal amounts of questionable code and then ask for a review. As much as AI coding can be a superpower in the right hands, ultimately, I think it’ll lead to the downfall of many big companies because:
* managers are now creating massive amounts of presentations, reports, and other agendas that developers now have to spend extra time reading through
* low performers can now look like they are working
* small companies with passionate engineers will move fast
I genuinely believe it can go two ways: you could offboard all your thinking to the LLM, or you could explore new things you’ve always wanted to try but never knew where to begin.
I’m trying to stay away from offloading the creative part of the work because I think with time, I could lose my ability to look at problems from a fresh perspective, and maybe offer a new approach – something that has had huge upsides during my career.
As for trying out stuff you never worked with, it’s a huge win. I mean… I just made an application for my Apple Watch without Swift knowledge 🤷♂️
I find AI tools incredibly useful for learning about unknown stuff. But ya, when it takes over the creative aspect of a certain thing, things stop feeling fun.
Thanks for the mention, Akos! Looking forward to your future posts.
Thanks, Saurabh! I'm still looking for the right blend of honing my programming skills, without making me too lazy. Next month, I'm going to give a shot at Codex or Claude code, hoping that the added friction between using a separate tool as a consultant will make me more mindful about how I'm using AI.
dittos. It doesn't help me grow as an engineer. I find the time diff between fixing AI code, and reading the docs to do it right my self, is close to equal. AI has helped for grunt work I know well but less so for new things
I want the AI to guide me in topics that are unknown to me, not to do all work (of questionable quality) instead of me.
Sure, once I find a pattern (refactorings are a great example of that), it's super helpful just to outsource the updates that are more complex than find & replace. But to generate solutions for existing codebases, from scratch... I'll spend just as much time reviewing that is it was someone elses PR.
I'm totally not ready to review 1000s lines of AI generated code.