Comment by CTDOCodebases
20 days ago
I wonder over the long term how programmers are going to maintain the proficiency to read and edit the code that the LLM produces.
20 days ago
I wonder over the long term how programmers are going to maintain the proficiency to read and edit the code that the LLM produces.
There were always many mediocre engineers around, some of them even with fancy titles like "Senior," "Principal", and CTO.
We have always survived it, so probably we can also survive mediocre coders not reading the code the LLM generates for them because they are unable to see the problems that they were never able to see in their handwritten code.
Honestly it’s not that hard. I already coded less and less as part of my job as I get more senior and just didn’t have time, but I was still easy to do code reviews and fix bugs, sit down and whip out a thousand lines in a power session. Once you learn it doesn’t take much practice to maintain it. A lot of traditional coding is very inefficient. With AI it’s like we’re moving from combustion cars to EVs, the energy efficiency is night and day, for doing the same thing.
That said, the next generation may struggle, but they’ll find their way.
Personally I planned to allocate weekly challenges to stay sharp.
It’s going to be extremely difficult if PR and code reviews do not prune unnecessary functions. From what I’m experiencing now, there’s a lot of additional code that gets generated.
"That's the neat part—you don't." Eventually the workflow will be to use the LLM to interpret the LLM-generated codebase.
I don’t read or edit the code my claude code agent produces. That’s its job now. My job is to organize the process and get things done.
In this case why can’t other agents just automate your job completely ? They are capable of that. What do you bring in the process of still doing manual organization ?
I still have to tell it what to do, and often how to do it. I manage its external memory and guidelines, and review implementation plans. I’m still heavily involved in software design and test coverage.
AI is not capable yet of automating my job completely – I anticipate this will happen within two years, maybe even this year (I’m an ML researcher).
13 replies →
[dead]
simonw alert!!!