$50M is tiny compared to the opportunity cost of having 100 of your best engineers jump ship to a competitor or found a new one. Google will pay $50M for a startup with 10 proven engineers and then shut down their product, just so that they aren't on the open market where they can challenge the $50B+ golden goose.

Silicon Valley basically works on the principle of "Never let your employees do to you what you did to your past employer or competitors." There have been multiple cases where a small group of 10 or so engineers has quit en masse and then gone on to found or strengthen a billion-dollar competitor, eg. Shockley => Fairchild => Intel => AMD, HP => Apple => [General Magic, Claris, Radius, NeXT] => Apple, Cisco's revolving door of startups, [Western Digital, PARC, Bell Labs] => Google => Facebook. The cost to the company of this is easily in the hundreds of millions, oftentimes billions, and sometimes existential. It makes sense to blow $50M or so to keep a proven team of tech talent in-house.

> $50M is tiny compared to the opportunity cost of having 100 of your best engineers jump ship to a competitor or found a new one.

See my comment to nnq. You can't just take "100 of your best engineers", and stick them on some bogus project that regularly gets raided for its best talent by hiring managers.

If they'd rather work on a bogus project than on a real product, they're not very useful engineers.

The project will quickly be seen as a joke, and not just by one unnamed "person who has spoken to Fuchsia staff", but by everyone in the company. No serious engineer would stay on it.

There's literally no evidence for this far-fetched theory, and no verified examples of this happening in the industry at all. If an engineer isn't producing something worthwhile, he's not a good engineer, certainly not "one of my best engineers".

The whole theory is pretty crazy and it's funny it ended up so high in the comment thread that we're discussing it seriously.

> Silicon Valley basically works on the principle of "Never let your employees do to you what you did to your past employer or competitors." [...]

This is a lot of unsubstantiated hand-waving with no real data to back it up. The data I've seen is that involuntary turnover (aka termination) in FAANG is 5-8% annually, which is very reasonable for healthy companies.

Have you worked for a FAANG? People are regularly put on notice and fired for poor performance. Even a very successful company can't afford to keep poor performers indefinitely. If they do, they would not remain successful for long.

You're describing a public company with fiduciary accountability to its shareholders as if it can be run like some corrupt retirement home for aging diva former-engineers. You realize that if Google was throwing away $50-100m of investor money per year, an exec would eventually have to publicly explain to the shareholders why their money is wasted this way, and if Google was trying to hide that it is a bogus project, then they would be lying to investors, which is a criminal offense?

This theory is frankly nonsense.

> Have you worked for a FAANG?

I was at Google for 5.5 years and was the beneficiary of an engineer-retention project for a year of that. I'm now working on a startup.

Think of it from the company's perspective: they have an engineer who has already been "paid off" in the sense of having generated more revenue for the company than they are likely to get in salary during a lifetime, and is likely independently wealthy through stock options. That engineer comes to them and says "I'm unhappy here and thinking of leaving so I can work on personal projects." The company does not want to see them on the open market, where they can potentially compete with Google or work for a competitor. So their manager asks them "Well, what is it you really want to work on? Maybe we can make it happen here and you don't have to leave all the perks of Google behind." The engineer says "I've always wanted to do OS development, and there's a bunch of neat research developments like capabilities that haven't yet been incorporated in a production OS." The manager says, "Hmm, I know of some guys over in this other department that have similar ideas. Why don't you connect with them and see if your interests align?"

Google wins because it keeps these talented engineers off the market, and if the project succeeds they can win in a really big way. The employee wins because they don't have to deal with all the risk and bullshit of being unemployed or starting their own thing. Competitors lose but screw them. Consumers probably lose, but it's not a sure thing because if their project succeeds, consumers can win in a big way, so Google doesn't get into antitrust or anticompetitive trouble for it.

This is not unlimited, BTW. The unofficial rule-of-thumb when I was at Google is that "you get one freebie" - if you were instrumental to the success of a project that made Google much more than you're ever likely to make in salary, you get one chance at a self-directed project. If that fails, you either go back to being an ordinary rank-and-file engineer and have to prove yourself, or you get marginalized and usually leave. Public examples of such projects include Google Scholar (Anurag basically ran the crawl single-handedly for many years, and so he gets to indulge his passion project because Google wouldn't exist without him), Golang (Ken Thompson is Ken Thompson and Rob Pike basically built the log handling infrastructure for Google), Google Wave (the Rasmussen brothers basically founded Google Maps), Hello.com (Orkut Buyokokkten's passion project, eventually spun out), and Andy Rubin's robot collection.

There's more: Cos and the open source Vanadium, John Hanke and Ingress/FieldTrip, Niniane and Lively, Hartmut and the quantum lab, Craig N-M and google.org, etc.

But you left out the largest retention side project: Jeff Dean and Brain.

Google Brain has been pretty awesome for academia: https://ai.google/research/pubs/?team=brain

also sebastian thrun and google[x] / self driving cars

craig got a whole new york office. he did well.

> So their manager asks them "Well, what is it you really want to work on? Maybe we can make it happen here and you don't have to leave all the perks of Google behind." The engineer says "I've always wanted to do OS development, and there's a bunch of neat research developments like capabilities that haven't yet been incorporated in a production OS." The manager says, "Hmm, I know of some guys over in this other department that have similar ideas. Why don't you connect with them and see if your interests align?"

What you describe is simply allowing a well-performing engineer to switch to new project, which of course happens all the time.

It's very different from "a bogus project for the sole purpose of retaining some engineers" that OP theorized.

> Google wins because it keeps these talented engineers off the market, and if the project succeeds they can win in a really big way.

So more or less like any ambitious new project? :)

> This is not unlimited, BTW.

Right, and again, goes against what OP was implying, that Google is going to run a huge project indefinitely just to retain a bunch of diva engineers who wouldn't work on anything else.

As I wrote above: "a rich employer may retain an unproductive employee for a limited amount of time if they expect it to ultimately pay off".

This is exactly what you are describing.

> if you were instrumental to the success of a project that made Google much more than you're ever likely to make in salary, you get one chance at a self-directed project.

Right, and Fuchsia is not a "self-directed project" of this kind. It's a big project with over 100 engineers, the vast majority of whom did not have the impact of Rob Pike or the others you mentioned.

Your comment is good and informative, but ultimately does not lend any support to the notion that Fuchsia really is some bogus project with no business case or future, solely for retention purposes, which is what OP was implying and what I am disputing.

Wave and Hello.com each had over 50 engineers, and Andy Rubin spent several hundred million (perhaps billions) on robotics acquisitions, all for projects that ended up getting canceled.

At some point the difference between "engineer retention project" and "risky venture that probably won't pan out but could win big" ceases to matter. Larry Page is exceptionally good at "Heads I win, tails I win even more" gambits - historically, these included Google Toolbar (initially conceived as a defensive play against Microsoft, but ended up getting Google Search in front of hundreds of millions of new users - this, BTW, was Sundar Pichai's first big win on the way to CEO); Google Chrome (ditto, with the side effect of becoming the dominant browser); GMail (initially an internal tool to improve productivity, then became a major consumer product); and Google Fiber (worst case, force Internet speeds up through competition, best case, new ISP). It's likely that Sundar & Larry's thinking of Fuchsia is that worst-case, it keeps 100 talented engineers at the company, and best-case, it's a dominant new OS that drives computing forwards. Either one is worth the $50M or so, so they do it.

> Wave and Hello.com each had over 50 engineers, and Andy Rubin spent several hundred million (perhaps billions) on robotics acquisitions, all for projects that ended up getting canceled.

And they all got canceled once their business case became void.

None of these were bogus projects. They all initially had a business case, and were run by engineers who proved themselves before. When that business case fail to pan out - they were canceled.

> At some point the difference between "engineer retention project" and "risky venture that probably won't pan out but could win big" ceases to matter.

Maybe to you as an employee. Not to Google as a business, nor its shareholders.

If I'm a Google shareholder, and you're wasting hundreds of millions of dollars of my money with nothing to show for it but some vague fears that "my engineers will go to competitors", then I'm going to be unhappy. Beyond a certain point, you're opening yourself up to lawsuits and even criminal charges (Google is a public company).

All of your examples are projects that were interesting and had some passionate senior engineers pushing them forward, but also had a business case. When the business case went bust, so did the project.

To support OP's assertion that Fuchsia is just a retention project with no future, you'd have to show a project of its size (or any size, really) run for the sole purpose of retention with no plausible business case.

I don't think such a thing exists, and it's not in any of your examples.

I have some personal experience with Wave. It had a plausible business case and Google was promoting it heavily. It wasn't bogus. I'm sure the same is true for the others.

The final risk factor listed on Google's 10-K [1] is:

"We rely on highly skilled personnel and, if we are unable to retain or motivate key personnel, hire qualified personnel, or maintain our corporate culture, we may not be able to grow effectively."

Acquiring and retaining key people is in the interests of shareholders, and Google invests significantly in this activity. I started this subthread with a comment about $50M acquihires where they shut down the product immediately afterwards. That's just one example - others include funding STEM education for children, the Google Summer of Code, multi-million-$ retention bonuses, and constant recruiting efforts.

This is all disclosed on the prospectus and annual reports. If you don't understand why this is important for a tech company, you should probably not be investing in tech, but I will be happy to buy your shares from you.

[1] https://www.sec.gov/Archives/edgar/data/1288776/000165204416...

> If I'm a Google shareholder, ... then I'm going to be unhappy.

Are you a voting shareholder (class A or B)? If you have class C (no vote), they don't have to care. If you have class A (voting public) shares, which get 1/10 the vote of the class B (management) shares, your influence is incidental at best.

I would not characterize fiber as having any notable impact on the country at large's internet speeds/caps.

It worked for Kansas City, Provo, and Austin, and basically failed everywhere else.

I think Fiber was a bit of a miscalculation in that they misjudged where the bottleneck would be. In the early days of Fiber (I worked on it, briefly), the assumption was that key risk factors would be that incumbent ISPs would raise their speeds to match, or that marketing would fail to sign up enough customers to make the build-out economical. These turned out to not be major problems, but the incumbent ISPs blocked expansion to new municipalities through permitting & right-of-way restrictions. The assumption when I was working on it was that if Kansas City could demonstrate a successful deployment, other cities would be lining up to clear roadblocks and bring Google Fiber to their citizens; this didn't happen.

Would you say that most major cable providers now having gigabit offerings now wasn't at least partially pushed by the threat of Google Fiber? I mean, in an area Fiber never launched in now, two of the three ISPs here provide gigabit via coax now (Comcast and WOW), and I'm hard pressed to believe they would've jumped from ~150 Mbps service to gigabit service at about the same price if someone else wasn't threatening to do it first.

Sure, it took a while after Google Fiber launched, but there was a lot of infrastructure to deploy. I'd argue Google had a hand in this.

> It's very different from "a bogus project for the sole purpose of retaining some engineers" that OP theorized.

I mean, it's not like Google is inventing bogus projects in a top-down sense, with some middle-manager somewhere coming up with "private works projects" and staffing them with engineers Google wants to keep. Nobody is claiming that.

No, what the GP was proposing is that the projects are bogus in a bottom-up sense: they're projects that are entirely useless to Google's shareholders in both a short-term and long-term sense, but which the engineers really want to work on, and which at least aren't actively harming Google (or perhaps they are, but not as much as Google would be harmed if the same project were developed in the wild.) This is very different from "basic research" (Google X et al), which is useless in a short-term sense, but may yield large positive returns later on. These projects are bogus because the best-case scenario for them is no net effect on Google as a business. (They might have quite an effect on the the world, but if there's no way for Google to convert that effect into shareholder value: bogus.)

Fuchsia is a perfect case: at best, it replaces Android, for no net benefit to Google (since they already dominate the mobile ecosystem with Android.) But if the engineers working on Fuchsia couldn't work on it at Google, they'd have started a similar project outside Google... and then Google would be in the position of a non-Google-owned OS displacing Android's share of the mobile-OS market.

Much better that these engineers do something useless [from a business perspective] that merely keeps stable Google's share of a market, than that they do something that actively lowers Google's share of a market.

Or more that they're likely useless from a shareholder perspective, but have a small chance of paying off in a large way, and worst-case keep the engineers occupied & employed at the company.

Google doesn't generally allow projects that everyone knows have zero chance of being useful in a bottom-line sense; you can't, for example, build a D&D campaign index on company time even if you're T9. (There was a big debate over this near the tail end of the Eric Schmidt years, when you actually could, and this was one reason for Larry's "More wood behind fewer arrows" campaign.)

But for something with a small chance of having a large impact, like a new OS or programming language or attempt to speed up scientific progress? Google can totally get behind that, because worst-case, you keep the engineers employed and available for future use, while best-case, you've got a computer science breakthrough. Fuchsia fits right into this case, as does Dart and Go and Unladen Swallow and several other projects.

Your comments have been enlightening. Thanks for confirming what I always suspected [1].

[1] https://news.ycombinator.com/item?id=12410662

Funny how you only took into account the post that confirms your preset bias.

This makes sense to me. I wonder how much the secret sauce comes into play. For example blowing a good/great engineer is one thing. But worse if they can apply some of the seemingly unique techniques google used to do engineering.

I’m thinking of things like the build system, mono repo & the borg etc

> If an engineer isn't producing something worthwhile, he's not a good engineer, certainly not "one of my best engineers".

Not my experience. I've seen amazing engineers happily convince themselves whatever they are working on is important regardless of what it is. In fact its pretty common for really good engineers to simply fall in love with the code - or be convinced if they only finished this refactor it would be beautiful. I almost never hear them talk of business value.

The ones that do have this insight end up going on to be great managers - as they can bridge the divide.

For a functional employer, the definition of "a good engineer" is "someone who produces valuable engineering solutions for me".

An engineer who isn't doing that is either mismanaged, misplaced, or is no good. In the first two cases, I can possibly make use of them. But if the engineer is a diva who would only work on some made-up fancy project, then he is not a good engineer for me.

Maybe he will be able to do good work elsewhere, but that's not really relevant for me, my business, or my shareholders. Not anymore than the fact he may be a great painter or musician. Good for him, but sadly irrelevant for my business.

Generally its the managers responsibility to ensure they are working on what is important. That's literally their entire job. An employee capable of convincing themselves whatever they work on is important is a net benefit to the manager - they are more malleable and can be convinced to work on the important but "boring" parts.

An employee refusing to work on something because in their opinion its not worthwhile is a tough place for both the employee and the manager.

Who said Fuchsia was some bogus project and a waste of talent and resources? My original comment was that Google was willing to let its enginners work on duplicate project so they remain at Google instead of leaving to work elsewhere. If Fuchsia turns out great, that will be a win for Google. It's not like these enginneers are just being paid to kill time.

Google bought Android, then also created Chrome OS and is now making a tthird OS named Fuchsia.

They had Gchat that they scrapped for Hangouts, which they pivoted into a business chat (Meet), but the Hangouts client is still available for the general public. They also made Allo, which eventually went nowhere, so now they're pushing for the adoption of RCS (Android Messages).

Gchat, Hangouts, Meet, Allo, RCS. That's 5 messaging apps right there.

RCS isn't an app (and neither it's something that Google built).

But they acquired Jibe, which they now offer as a service to telecom companies, so with Android Messages, they're invested in RCS.

You talk tough but I don't think your understanding of the problem space warrants that tone. Doing research & development is not something that involves lying to investors. Investors love new products. That is how they make money by investing.

Consider the project mentioned in the article. Sure, it will cost millions of dollars to develop. But OSes are in high demand, existing OSes have quite a few sore points, and plenty of people have made money selling operating systems or devices that need a custom operating system (MS/DOS, Windows, iOS, etc.) So it is not categorically insane from a business or engineering perspective to say "hey, we're going to invest in the development of a new operating system that will change the world."

On the other hand, maybe all the OS world needs are some patches on top of Linux to smooth over the rough spots.

People that see the project the first way will say they're working on a "moonshot". People that see the project the second way will say they're working on a "senior engineer retention project."

Any big company will likely want to hedge its bets, and indeed... apparently Google has people working on a new OS, and they obviously have people patching the Linux kernel (Android). No matter which approach ends up being correct, Google makes money. I don't think the investors are going to be filing many criminal charges against the board.

Back to the retention aspect, remember what the downside is in this scenario. The people that want to build their own OS to change the world or whatever just leave and go do it. They have money and can probably rope in some investors. If they end up being right about the approach... there is no more Android and that means no more Google ads on mobile phones, no more location history, no more searches... etc. That is quite a downside for Google. So even though the project is "this seems crazy, but we don't want these people to quit", it's 100% the right financial decision.

I think the shareholders can appreciate that.

That said, as an employee... this was all a little too much for me when I was at Google. No matter what you worked on, there was always a competing project with twice as much headcount and 1/10th the users. You would have a meeting with them along the lines of "hey, this thing you're assembling a 30 person team for... my team already made and it's in production and working." They would be like "nah, our use case is slightly different and rather than editing 1 line of code and getting no promotion, I'd like to build a HUGE team and put an EZ promotion packet together." Not for me.

> $50M is tiny compared to the opportunity cost of having 100 of your best engineers jump ship to a competitor or found a new one.

One of the following must be true:

- Fuchsia has perceived product value, and it's an investment

- Google doesn't have enough interesting work right now, the amount of interesting fluctuates up & down, and retaining expensive employees by having them work on "Fun Fuchsia" during the current down cycle is more efficient than hiring expensive employee attrition+recruitment

- Fuchsia has no product value, and Google has too many expensive engineers

> the cost of keeping them must be at least 50 million per year.

There's a high likelihood that at least some of their work will be of value to the company—or, if everything pans out, they get a fancy new platform. The real motivation is probably somewhere in the middle (retention, product development).

This is a solid bet to make, especially when you are effectively not worried about the cost (thanks, ads!).

A lot of Google is heavily research focused like this: let engineers (mostly) freely explore complicated problem spaces. If it's a product—awesome!—see if there's market fit, and then grow it. If not, no problem; see where it, or parts thereof, can be applied elsewhere within Google.

Source: I worked there.

Sure, that's more reasonable. You're describing a match between business needs (the potential Android successor - quite an important business target) and a critical mass of senior engineers eager to work on the product.

I'm sure Google is aware that more interesting and ambitious projects do tend to capture and hold the interest of the most talented engineers, and that figures into their business decisions.

What I strongly dispute is this notion that such a project would be subsidized solely for retention. Everyone knows about big Google projects that were shut down when their business case lost its appeal. I've also seen much smaller projects shut down when their value proposition was no longer deemed sufficient, even when their members were talented, passionate, and at real risk of leaving once their project shut down.

> It just doesn't make any sense to retain someone indefinitely if they're not going to produce anything of value

They likely will. Whenever a new projects starts, engineers can be recruited for it from the pre-vetted internal pool. Hiring and the ramp-up time for new hires would cost more. Yeah, if you're an ossified compnay rarely starting new projects it doesn't make sense. But if your Google this is not the case...

This sounds good in theory, but in practice what you're describing will not be stable, sustainable, or at all practical.

You can't really run this sort of revolving-door "talent pool" project for long, certainly not a massive project of 100+ engineers. If every time a new project opens up, its hiring managers get to unreservedly raid the "talent pool" project for star performers, then pretty soon this project will be known internally as a joke, that is at priority zero, and any but its most dysfunctional engineers will be rushing to other projects, either internally within Google or (more likely) outside it.

You'll end up with all the best people quickly leaving for real projects where they can do real work on real products for real pay (bonuses for revenue-yielding projects are much, much higher than for anything else). The project itself will be a dysfunctional joke since you can't run it with the best talent routinely recruited away. The only people who will work on it will be dead weight who don't care to work on real products or even make any progress on a bogus one, which is the definition of the sort of people you should immediately fire.

In fact, the whole initial "pool" of candidates sounds toxic, if the only reason they were given this project is that any other _real_ work would cause them to leave. As a hiring manager, I'd be very skeptical of anyone who seemed to be this sort of indifferent diva, and would not be looking to hire them.

To sum up, it's a fun little theory, but extremely unlikely to work in practice, and I've never seen anything like it. The simple explanation that Fuchsia is an Android successor is much more likely, and there's zero evidence for the far-fetched "talent pool" explanation other than an unsupported comment by some anonymous "one person who has spoken to Fuchsia staff". It's ridiculous.

> That's a lot of money to pay just for retention of people who are not producing anything of value, consuming many other resources (office space, network, hardware, perks), and deeply committed to a project that supposedly will never benefit their employer.

I think Google figure it out its real bussiness like MacDonalds did: Not burgers, but real state bussiness. I think Google understood they are not in the make "valuable useful tech" bussiness as one is inclined to think, they are in the "tech talent retention bussiness".

Even if they dont use the talent, if they are able to lock the talent inside Google as a kid in a Circus, their competitors wont have access to the talent pool they need, and this will make a big monopoly on Google.

Of course, this is only possible if you have the ammount of cash and credentials Google have to attract tech people like Disney do to kids.

I think this is their real primary mission; attract human talent, retain and ultimately lock it from the competition.

Do you ever wonder why the startup scene is so dry lately, even with so many enginners and tech focused people no short of ideas? The titans are just sucking all the oxigene in the room making everyone else choke.

> If these are all "senior engineers" that "Google is trying to retain", the cost of keeping them must be at least 50 million per year.

Wait uh... are you saying seniors at Google make an average of $500k?

That seems like a low estimate of the cost of a senior SWE at Google. They probably earn at least 400k in cash + bonus + stock, and there are employer costs beyond that (office space, benefits, payroll taxes).

Glassdoor has avg total comp for senior swes at google at 193k...

I am very surprised if that's the case.

For what it's worth, I just started at Google as a Senior SWE this summer. My base is in the 180s. With an average annual bonus, my moving/starting bonus, and stock grant (ignoring stock price movements), my first four years should be 400ish...very high 300s really.

BTW, https://www.levels.fyi/ is another source of data which might prove useful

190 total comp is more in line with what I expect for an early career person. Welcome to 2018 tech company salaries.

That's a pretty great package... brb applying to work at Google.

That is definitely not correct.

Senior engineer can either mean "an engineer with a significant amount of experience" or "a title for a specific position at a specific company".

Either way based on my experience as a manager at Google, the answer to your question is yes.

Yes -- https://danluu.com/startup-tradeoffs/ cites $250k as a "conservative" (low) estimate for a senior SWE there, and was written in 2015.

Salary is usually only 2/3 of the cost of an employee. Benefits and overhead are a big part.

Must be the first law of HN: Every thread devolves into a discussion of $1mn+ engineer salaries ;)

Some do. If you add bonuses and stock grands it can get to at least half a million per year

> they often merge to a single deliverable that is better than any of the projects could produce individually.

Do you have an example of this?