What’s actually happening is there’s a divide being created between engineers that know how to use it, and engineers that don’t or want to convince themselves that it’s useless or whatever.
Group 2 will not fare well in the coming months.
Deeper still is this: The group that openly relies on LLMs to code is likely the first group to be replaced by LLMs, since they've already identified the recipe that the organization needs to replace them.
More broadly, we live in an age where marketing is crucial to getting noticed, where good work alone is not sufficient and you have the perfect scenario for people to market themselves out of the workforce.
For new features, by all means code it by hand. Maybe that is best! But a codebase is much more than new features. Invaluable tool.
If it's a matter of months then latecomers will be up to speed in months as well, which isn't really that long a time.
But the internet took decades for people to get onboarded and figure out how to use. We're still figuring out how to use it in some ways. That's not to say the people should sleep on LLMs, but let's chill with the "Group 2 will not fare well in the coming months" nonsense. If anything group 1 is at a higher risk because they rely on others to set the trends they tell everyone else to catch up to.
