The system you talk about is mostly dead.

Google promos everywhere up to a certain level are in-org (IE not unknown engineering peers). They are done alongside rating calibrations. Google promos everywhere up to an even higher level are basically in-PA.

For cloud, they are in-org to even a higher level. So the stuff talked about in this tweet would have been evaluated not just by cloud engineers, but almost certainly fairly local cloud engineers.

My experiencing having been on promo committees for google for 15 years is that for every case in my org i see of google incentivizing the wrong thing (or disincentivizing the right thing), i see probably 5-6 cases of it doing the right thing.

I see a lot more cases that are like:

1. Person builds new shiny thing for ... questionable reasons.

2. They make a mess of the world by not making it easy to migrate people to it and get rid of the old thing

3. Things are in a bad half-state, where both things have to be supported and the new thing does not provide obvious higher value.

4. Google doesn't promote them, they get pissed off about it and post on twitter about how Google didn't care about product excellence or something.

Than cases like:

1. Person toils away on the right thing forever

2. Person tries to get promoted for doing the right thing and having impact

3. Google turns them down

4. They get pissed off and post on twitter about how Google didn't care about about product excellence or something.

Externally, of course, people can't distinguish these two cases because they look the same (angry people on twitter).

I have been promoted from SWE III to Senior Director at Google, working only on open source until i was an L7.

Agreed. I work in the PP's PA and was promoted on my first try for doing unsexy but worthwhile maintenance work on an important tool. It can happen. In my estimation the best thing to do if you want to get promoted is 0) have a positive relationship with your manager 1) read the SWE ladder for the next level 2) give your manager a document that describes your work using the language of the ladder as closely as possible. Said ladder does not say you need to deliver a shiny new thing, despite what is commonly assumed in this forum.

It's good to know the system I'm talking about is mostly dead; the committee of unknown engineers always felt bonkers to me. That being said, a ratio of 1:5 or 1:6 of bad incentives at promo committees still sounds pretty suboptimal to me — when I was on calibration committees at FB, it was super rare for me to come away with the feeling that the wrong thing was being incentivized or rewarded at an engineering level (oh boy was product management a different story, but that's a different post). There were a couple times where I disagreed with an outcome, but that was like 1:150 or better.