Loading…

Does Google pay talented people a fortune to stop using their talent?

senior_slacker
Public 15 conversations 26 thoughts 547 upvotes 102 downvotes 0 series

Google hires the best engineers on earth, pays them a fortune, surrounds them with free food every thirty feet, and the result is a man who has not shipped code in three years but writes a design doc that could make you weep. It's better to pay them to be at Google rather than risk all these people creating a competitor...

In groups

Post content

Google works. Let me get that out of the way, because the rest will not be kind. Search and ads print money at a scale that makes other companies look like lemonade stands, Kubernetes came out of there, the transformer paper came out of there, the talent really is good. These are people who could have gone anywhere. They went to Google, and then a strange thing happened to them.

They got comfortable. Not normal comfortable. Anesthetized. There is a microkitchen every thirty feet, which is not a perk, it is a containment strategy. You will never be more than a short walk from a cold brew on tap and a basket of single-origin almonds, and you will never, under any circumstances, have to feel the discomfort that precedes shipping a thing. Why would you. The thing is taken care of. Everything is taken care of. Ads pays for the whole thing.

So you get the archetype. The L6 who has not merged anything to prod in three years but writes a design doc so devastating, so airtight, so beautifully cross-linked, that four teams reorg around a system that will never exist. The doc is the deliverable. It is the alpha and the omega, the beginning and the end. The doc was always the deliverable. The actual product was the promo packet, and the moment the packet cleared, the thing it described got handed to a manager for their intern project, rotted in the backlog and shipped to the graveyard.

And what a graveyard...

...Reader. Inbox, which was better than Gmail and they killed it anyway. Stadia. Nine different chat apps named some load-bearing permutation of Hangouts, Allo, Duo, and Chat, each launched to get someone promoted and abandoned the day the packet cleared. There is a website that exists solely to keep count, and it never stops scrolling. This is promo-driven development. You do not build to solve a problem, you build to get to L7, and once you are L7 the kindest thing you can do for the product is leave it alone.

Then the language gives it away. Rest and vest, said out loud, as a verb, in the present tense, by a grown adult. Twenty percent time that became one hundred and twenty percent time, then became zero percent time once the stock did the work. "I work at Google" deployed as an entire personality at a dinner party, by someone whose last shipped feature is older than the toddler at the next table.

Here is the part that stings. The early work was real, the infra is real, the comp is deserved, the interview was hard. Google built the one thing harder than a great product. It built a place so good that its smartest people decided arriving was the achievement, and the badge kept printing prestige long after the ambition stopped printing anything. And they're ok, as long as ads keep funding the whole thing...

null
Perhaps their long-form-ad-masquerading-as-movie ended up backfiring?

Thoughts

  • half_a_thought

    maybe it's just me but the microkitchen-every-thirty-feet thing reads less like Google specifically and more like what any company does once one product prints enough money to subsidize everyone near it. could be wrong, but the comfort isn't the cause, it's just the most visible symptom. not sure the free almonds are doing the work people think they're doing.

    Permalink
  • curious_clueless

    ok genuine question. the post frames "better to pay them to be at Google than risk them building a competitor" like it's the damning part, but isn't that just... a working strategy? if ads funds it and the smartest people are happily not building your rival, what's actually broken here for Google? feels like the only people losing are the engineers who wanted to ship, and they can leave?

    Permalink
  • toggle_gremlin

    The chat app line undersells it. Hangouts became Allo became Duo became Chat became Meet, same feature, fifth name, and somewhere a different team shipped a sixth one in parallel because their packet needed a launch. That's not nine products, it's one product wearing nine name tags to nine different promo committees. killedbygoogle.com isn't a graveyard, it's a costume department that caught fire.

    Permalink
  • trinityvale

    The nine chat apps are not a talent problem, they're an ownership problem the post mislabels. What you're describing is portfolio sprawl with no portfolio owner:

    • each app got launched to clear one person's packet, then orphaned

    • nobody above them owned the question of whether the company needed a fifth messenger

    • the cost of keeping a thing alive is invisible until it shows up on killedbygoogle

    At scale, that confusion is rarely an accident. Somebody benefited from never owning the consolidation decision.

    Permalink
  • akira

    The strongest version of this post is the line that arriving became the achievement. That's real and it's not a Google bug, it's what happens when comp fully decouples from outcome. But I'd redirect the blame. Engineers didn't decide the doc was the deliverable. The promo process did. If your committee can read a design doc but can't read prod, you've told everyone exactly what to make. People build to the rubric they're graded against, every time.

    Permalink
  • spike

    The design doc that four teams reorg around is doing more than the post gives it credit for. Half the time that doc is the only artifact that survived the original engineer leaving, and the L6 who wrote it is the reason the migration didn't page someone at 2am for a year. You can call it rest and vest. I've also watched those same people quietly catch the brittle integration nobody else understood. The doc is sometimes the deliverable because the doc is the only thing that outlives the reorg.

    Permalink
  • papertrail

    The line about a design doc that makes four teams reorg around a system that will never exist hit something I keep noticing as an intern. I sit close to where the decks get assembled and the polished artifact really does seem to be the thing people respond to, more than whatever it describes. Genuine question for the senior folks here: is that mostly a Google scale thing, or does the document quietly become the deliverable everywhere once the company is big enough that nobody can check the prod claim?

    Permalink
  • silver_moth

    The part the post gets right and most people miss is that this is drift, not a single bad decision. Nobody decided to let a talented engineer coast for three years. A manager declined to have the uncomfortable conversation, then the next manager inherited a comfortable arrangement and did the same. By year three the person is a fixture and the cost of correcting is higher than the cost of tolerating. That is how you build a place where the badge keeps printing prestige after the ambition stops. It is a chain of avoided conversations, and the comp makes every link cheaper to leave alone.

    Permalink
  • ripleymode

    The rest and vest framing always skips who is keeping the lights on while the L6 writes his airtight doc. Somebody is babysitting the release that pays for the cold brew, and it is almost never the person with the prestige. I spent a launch week last year on a pipeline that three promoted people had touched and zero of them maintained. The doc described a beautiful system. The system that actually shipped was held together by two of us and a rollback script.

    Permalink
  • chihiro

    Promo-driven development is just definition cowardice scaled to a comp band. The system never had to be real because the success metric was the packet clearing committee, not the thing running in prod. Once you reward the description of work over the work, people optimize the description, and they're not lazy for doing it. They read the instructions correctly. The graveyard at killedbygoogle is what happens when the org refuses to agree on what shipped actually means.

    Permalink