EDIT: Somehow, the parent comment this was made as a reply to (notacoward's above-thread) has moved, and this comment has become top-level (and a non-sequitur). Is this a bug?
The reality is that years of experience in the general area include a lot of transferable skills to actually deliver in a landscape of new versions and languages. That landscape won't suddenly stop changing to make whoever learned only what was hot in the last 3 years a stable senior engineer, but pretending it will is a great way to cut out any older engineers.
Domain knowledge and experience is criminally underrated nowadays. A company I've worked for is currently going through one fire after another constantly because the executives haven't valued domain knowledge and has forced or outsourced just about everyone who knows anything about the projects that they still have paying clients for and need to support to keep making revenue, and things are just barely not falling completely apart (looks like things are going to get worse here pretty soon though).
And this isn't even ageism, really, even the young devs that have only been there a couple of years are being forced out.
Yes they could but seeing that most companies don’t, why not just accept reality for what it is and study and keep yourself marketable? I’m in my mid 40s and I make it a point to stay qualified for the well paying developer/architect jobs that come in my email. When I see that the company I am working for isn’t going to keep me marketable, it’s time to leave.
Hiring people based on skills is perfectly fine. Hiring people (or firing them, which is what this is about) and just saying it's about skills often is not. That's why I put "skills" in scare quotes. It's like when companies use "culture fit" to mean whichever races, genders, and fads the founders feel comfortable with. It's not that culture fit isn't a real thing, but much more often it's a shibboleth (or "dogwhistle" in current parlance). Similarly, "skills" is clearly the messaging IBM has chosen. Now it's up to them to prove that they were truly making efforts to optimize skills and not something else.
Disparate impact analysis will require them to prove something closer to that they actually we're optimizing skills, not just making efforts to, at least presuming that the unequal impact prong is proven.