Normally this isn't a problem as screening criteria like experience in language X is a bona fide job requirement, so even if such criterion filtered out minorities disproportionately, it's not a liability. Tacking on filters like 'must know programming language not used at this company' is going to be a huge stretch, and if your African American candidate pool is filtered out more regularly, that's a lawsuit waiting to happen.
So yes, crafting heuristics is hard, but manager's paychecks are larger for a reason.
That said, the African-American candidate pool in America is so vanishingly small that subtle tech bias at hiring time isn't the issue - the heuristics to filter them out started before they ever finished high school. (This has actually been a point of interest for me for a long time. I mentioned elsewhere a friend who has terrible spelling but a very successful career. He also happens to be black, which has been something else to overcome. We've talked for years about ways to encourage more black students to go into IT, which we both believe is a route to middle-class success with substantially less racial animosity than they'll face in other fields.)
Dan's point is largely that for each new & shiny technology you add to your stack, and then add to your hiring filter, you shrink your hiring pool substantially. You can do new & shiny, or you can hire day 1 productive staff experienced in your tech stack. Many firms look at their budgets and choose the latter, but if you don't have budget to train staff, you probably don't have budget to hire experienced staff either. So while the recruiting slogan is 'we only the best', in practice you get 'we only hire the tiny fraction of new grads who've been researching our exact tech stack in their spare time'.