For the past week and a bit my feeds have been one story, the internal Google memo that leaked, the firing that followed, and the enormous, exhausting argument either side of it. If you have somehow missed it: an engineer circulated a document arguing, amongst other things, that the gender gap in software is partly down to innate differences rather than bias or culture, it leaked publicly in early August, and Google let him go. Then the entire industry had opinions, loudly, for days.
I am not going to relitigate the document line by line. Plenty of people with more patience than me have done that, and the science it leaned on has been picked apart by people actually qualified to do so. What I want to write about is the thing underneath, the bit that's quieter and more uncomfortable, because it's the bit I actually have experience of: how engineering teams end up looking the way they do, and why "we just hire the best people" is the most self-flattering lie we tell ourselves.
the pipeline excuse
The most common defence I heard this week was the pipeline. We can't hire people who don't apply, the argument goes, and the applicants skew a particular way, so the team skews that way, and there's nothing to be done short of fixing schools. It sounds reasonable. It even has some truth in it. And it's also a very convenient place to stop thinking, because it locates the problem entirely outside the building.
I've sat on enough interview loops to know it isn't the whole story. The pipeline is real, but so is everything that happens after the CV lands. Who gets the benefit of the doubt on a shaky whiteboard answer. Whose "culture fit" gets a tick and whose gets a frown. Which candidates we chase and which we let go quiet. None of that is in the pipeline. All of it is in us.
how monocultures actually form
Here is the mechanism, as best I can describe it from the inside, and it isn't malice. It's comfort.
You hire someone. They're good. They get on with the team. The next time you're hiring, you reach, half-consciously, for more of the same, because the last one worked and pattern-matching is what brains do under time pressure. Interviews are time pressure distilled. So you reward the candidate who reminds you of the people already there, who uses the same shorthand, who laughs at the same references, who interviews in the dialect you happen to speak.
Do that for a few years and you have a team that's homogeneous in ways nobody decided and everybody contributed to. Then someone writes a memo explaining that the homogeneity is natural, and a portion of the team nods, because the memo flatters the exact process that selected them. That's the part I find genuinely sad. It's a closed loop that congratulates itself.
what I've actually seen work
I'm wary of turning this into a listicle of solutions, because the people who write those usually haven't tried them. But a couple of small, dull things have measurably changed loops I've been part of, and dull is a compliment here.
The first is a scorecard written before the interview, not after. Decide what the role needs, write down what you'll assess, and grade against that. It feels bureaucratic and it is, and it also makes "I just had a good feeling" much harder to launder into a hire/no-hire. The good feeling is exactly the bias. The whole point is to slow it down.
The second is structuring the actual questions so every candidate gets the same ones in the same order. Unstructured interviews are, by a wide margin, the worst predictor of on-the-job performance we have, and they happen to be the best vehicle for "is this person like me". You give up some of the easy rapport. You gain a process that isn't quietly sorting for sameness.
Neither of these is radical. Neither requires believing anything controversial about biology or psychology. They just require accepting that a competent engineer is also a biased one, because all humans are, and building a hiring process that assumes that rather than one that pretends otherwise.
the cost of the loud version
There's a second thing this week has done, and it's quieter than the memo itself. It has made the conversation harder to have at all. When the public version of a debate becomes this hot, this tribal, the version you can have at work goes cold. Nobody wants to be the person who raises diversity in a standup the week the internet is on fire about it, because whatever you say will be read as a flag you're planting rather than a point you're making.
That's a loss, and it's a subtle one. The day-to-day work of making a team less homogeneous is dull and procedural, the scorecards and the structured loops and the noticing. It isn't a battle of ideas, it's a set of small habits. But small habits need calm to install, and a fortnight like this one drains the calm out of the room. The people who were quietly improving their hiring get tarred with the loud argument, and the people who never wanted to change anything get handed a convenient reason to keep not changing it: look how toxic this has all become, best to leave it.
I don't think the answer is to avoid the subject. I think the answer is to keep doing the boring version regardless of what's trending, and to resist the pull to turn a process problem into a referendum on anyone's worldview. The scorecard doesn't care what you believe about biology. It just makes you write down what good looks like before you meet the person, which is the whole trick.
the part that nags
What nags at me about this week is that the memo's author wasn't a villain in a film. By most accounts he was a real engineer who genuinely believed he was being rigorous, and who wrote down a thing a lot of people quietly think and don't say. The firing made him a martyr to one crowd and a punchline to another, and both reactions skipped neatly past the actual question, which is: if your team's makeup is the natural result of merit, why does the merit keep arriving in such a narrow shape?
I don't have a tidy answer. I have a scorecard, a structured loop, and a habit of noticing when I'm about to hire someone because they remind me of me. That's not a revolution. It's just the small, boring discipline of not trusting my own first impression, which is the same discipline I apply to a green test suite or a passing build that looks too good. The feeling of certainty is the warning sign, not the all-clear.
Anyway. The feeds will move on to the next thing by the weekend. The loop in the building won't, unless someone in the building decides to change it.