Individually, you can't. That's why collectively resisting the temptation to play to discriminatory tactics is so important.

But, if you're researching, it's trivial. You send out a bunch of CVs / resumés that are identical apart from age. You then compare the return rates.

Out of curiousity, have people done this study about tech workplaces, particularly in the Bay Area?

If it's incredibly trivial to prove, I'd be surprised if it hadn't. (I'd be equally surprised if it had been attempted and shown no age discrimination--it'd push against my priors, but I'd try to keep an open mind.)

In reality, I suspect it's much harder than you suggest to prove, and even harder to litigate on.

The biggest problem with age discrimination is that it's subtle-- not overt. It exists in the form of harsh age-grading for accomplishments, because the delta between the average and the maximum possible at a given age grows, so the average person's deficit compared to where they "should be" becomes more of an issue. No one would pass on a 55-year-old CS luminary who's still writing papers, but 40-year-olds with the careers expected of 35-year-olds are screwed. After a certain age, being merely above average isn't acceptable anymore.

Most companies would rather bet on a young 6 than an old 8, because they overestimate their ability to mold the 6 into a 9 (which is actually very hard to do).

That's why establishing age discrimination is impossible. You'd have to generate identically indicative/"qualified" resumes at very different ages, which is probably impossible to do.