SV AI Startups Are Embracing China's Controversial '996' Work Schedule | Hacker News Reader

SV AI Startups Are Embracing China's Controversial '996' Work Schedule

wired.com

48 pointsAndrewDucker67 comments

The Chinese 996 work schedule is indeed hard work but it doesn't mean working 72 hours according to how westerners calculate working hours. For one, tech workers arrive at the office at 9am, not that they start working at 9am: once they arrive they simply head to the office cafeteria to have breakfast between 9am and 10am. And it's really common to spend one hour each for breakfast, lunch, and dinner. Unlike westerners who'd be happy eating some cold sandwiches for lunch, the Chinese are rather picky about their food being hot and served with soup. Second, the Chinese have a weird culture of napping at the office after lunch. Not just some nap pods here and there, but institutionalized naps with day beds for all. So right off the bat, you deduct up to four hours of non-working time from twelve hours of in-office time. Finally, the Chinese also have a culture of not leaving before their bosses leave the office. This means even if a worker has finished all assigned tasks, the worker simply kills time and waits for their boss to leave. It might be reading internal documentation, or it might be something less productive.

996 refers to time in office, not time worked. The actual hours worked is closer to 48 hours per week than 72 hours.

Great. Does 996 require the schedule that it implies? If so, respectfully, those excuses are moot.

The distinction between hours in the office and hours working is interesting to an employer. It is not interesting to the employee. The fact of the matter is that 996, assuming a 30 minute commute, means people are spending 2/3 of their lives working for someone else. And then when the exit comes, the founders and the investors can screw you in a heartbeat and still get theirs (see Windsurf).

It's actually not weird to nap after lunch. It's weird not to (when considering a global perspective).

Since there are no nap pods in the West, the closest I found myself doing is relaxation exercises in a phone booth for 30 minutes, and this significantly boosted my afternoon productivity.

IMO that really is still the same thing as working because you can't spend that time with your family. The fact that it's nice doesn't change that you're working, spending time away from those you love.

Very true, but that's 12 hours spent at your workspace. Having dinner at home is different from at your job, that's obvious, right? I bet that most people would rather have 30 minute meals and go home one hour early, with everything else the same, because that's one extra hour spent however the way they want.

Everything we know about systems resilience shows that 60% to 80% utilization is the sweet spot. We treat our silicon topology better than our human topology.

This article is arguing for working hours that equate to about 65% of waking hours. So your range seems too high.

I think from context a more likely interpretation of the parent comment is that it is saying "60% to 80% utilisation [of working hours]".

Perhaps. But what defines working hours?

I have worked hours like that, for years. Over time, i learned that it's mainly good if you don't know what you're doing, and you want to waste lots of time, and not be at your sharpest. While also often feeling like crap.

The next time I'm leading a team, I'm leaning towards 40 solid hours/wk being the official way to go.

Up to around 60 hours/wk of mixed-solidity work is also sustainable long-term for most people, without ill health effects iff some of that will be much lower-productivity time and the people don't have daily family obligations and the overall stress isn't too high.

But if some people wanted to try 50-60 hours and a mix of pace themselves, I'd probably have them structure it so that they appeared to 40-hour colleagues as if they were also doing 40 hours. Which means no slacking off in front of them, nobody getting disturbed on off hours with messages or pull request reviews, etc.

>The next time I'm leading a team, I'm leaning towards 40 solid hours/wk being the official way to go.

We are sick as a civilization and a species.

@happytoexplain, please explain?

Does anyone else feel more productive doing about 4 hours per day? I really have found it to be a sweet spot and since it’s so sustainable (endurance pace), I enjoy getting a couple hours in during weekend early mornings.

Definitely. I'd much rather have people put in a smart 4 hours, then more hours. The older I get, the smarter I work, and the fewer hours it takes to get more done.

There will be no liberation with AI. Hasn’t been with any other tech. Output always fills the space/time created

Depends on whether your company / business is on the winning or losing side of the shift

Productivity goes up while wages stay the same. Whatever side the business is on, an hour still costs the same, whether it’s more productive or not.

How are they going to recruit for that without paying megabucks

They pay megabucks.

Most startups are poor though, so I cant see this becoming widespread

Sometimes literally with that SI prefix.

By leveraging an oversaturated labor market.

This will upset people, but I love programming enough that I would do 996 at 40k-50k usd, with half going to rent and potentially even paying out of pocket for medical insurance.

If you're willing to take a loss like that, you should spend your time programming for yourself, not willingly making yourself a victim of exploitation.

Why? At that salary just be a bus driver or something and code for fun.

Same reason why "starving artists" (eg. actors, musicians, and game developers) don't become bus drivers and do their art over the weekend. They like the job so much they'd rather take the pay cut and poor working conditions just to work in the job. It's a well-known concept in economics: https://en.wikipedia.org/wiki/Compensating_differential

But would you be OK with the fact that your boss is making millions off your passion?

As an ex-academic, I know where you're coming from. In academia it's easier to justify working all thr time, since at least no one is making money. It's just for the love of pursuing knowledge. I cannot imagine breaking my back smiling just to make someone else rich.

I think OP's point isn't that he'd accept taking a job at [insert tech company here] and have them pay him $50k, when the going rate is $300k or whatever, it's that in a hypothetical scenario where programming isn't prestigious and the pay was only $50k, he'd still take that job (as opposed to finding another line of work like finance or whatever).

Why wouldn't you do freelance or speculative work if you're putting that many hours in?

That’s so dumb, you don’t have to do that to spend all your time programming. Honestly

By bullshitting to employees that what they're building is the next big thing. AI has created a very bad culture and these startups have become very good at grabbing new grads or junior engineers that have a lot of energy and time and by saying they're going to be doing amazing things by working these 996 schedules...

The reality is the vast majority of these companies will be hit hard by a burst of the AI bubble. It won't be a surprise the so called millions they're looking for end up being a 0.

with this and the recent Windsurf “acquisition”, what’s even the draw to startups anymore? 25% base pay increase for 80% more hours (not even considering overtime) with equity that is tantamount to useless (with or without Windsurf being the new norm)? super hard pass and I hope everyone with half a brain does the same

AI is being used as an excuse to strip away every labor norm and law in the name of “not being left behind”, with a product that has yet to truly be earth shattering on the scale it’s been predicted to be. hard not to see the current slate of VCs and founders as leechs

It's a race to the bottom, and it hurts all of us.

This is true. Everyone I have visited in SV is working from wake to sleep, at some Airbnb, where you are expected to just exist there. It’s slavery.

“The 996 work culture is, in fact, a symptom of corporate incompetence.”

Only when management is unable to deliver real results to investors do they fall back on saying, “Look, we’re all working passionately, we’re doing 996!” The implicit message is: “Even though I’m incapable, we’re working hard—please keep investing in us.” It’s a methodology of working for performance theater, not actual productivity.

I’ve worked as a professional executive in China for over 20 years. From 2000 to 2008, I grew in foreign enterprises. After 2010, I joined Chinese internet companies and worked with many firms, mainly focusing on IT governance to meet IPO-level audit requirements.

One principle has remained consistent throughout my career:

Overtime is a sign of managerial failure. And when a company turns overtime into culture, it is collective incompetence—from the organization and its leadership. That should be a source of shame.

When I began working in Chinese internet companies around 2010, leadership both appreciated my capabilities and deeply resented my foreign-corporate mindset. They felt I always touched their most painful nerve: that 996 is a mark of failure. They found it hard to work with me because I always structured everything with clarity, and what they hated most was clarity. I spoke in logic and market truths; they hated that, always insisting they were “crossing the river by feeling the stones.”

So when I see Silicon Valley now embracing 996, I know this is the final straw for the AI bubble:

When people start working for the performance, not for the purpose, the collapse is already on its way.

Brilliant, 996 yourself out of a job.

The only boss I'd work 996 for is myself.

Every time I am stressed by work, I think of this obnoxious work "ethic", and I feel better.

In other news, what happened to AI? Isn't it supposed to make life easier? Work for one weekend day, no health insurance, etc - we back to preindustrial oligarchy, fam?

I interviewed at one AI startup that basically was doing this. From what I could tell, they were feeling pressured to get something out fast since there is so much money and competition.

What was odd was that they were already doing this for about 6 months and needed people who could build things, the founder was very frustrated. Technically it looked like there were severely over complicating their code base (I never saw it, this was just based on discussions) and needed people to fix their perf issues/etc, which AI is very bad at.

It really feels like the purpose of AI is realizing the capitalist's dream of the perpetual value machine which would solve the problem of human labor.

Hard to feel optimistic for such a future.

We need more class traitors.

Capitalists really want slavery again.

Why again? It was never gone, just in foreign lands. There is a reason why China became our workbench.

Or we need to end capitalism. Once scarcity is not an issue allocation of capital becomes more apt to planning than competition.

We need less class traitors. Imagine the power of the working class if they weren't fooled into worrying about illegals, poors, and ethnics.

[flagged]

I suspect the GP meant the working class are being tricked into thinking the "illegals, poors and ethnics" are the evils that are holding the working class back by Republicans.

Nevermind the fact that casual bigotry against working class people (blue collars and uneducated are too dumb to know what is good for them) is just as repugnant as what you thought they meant.

I meant more that the elite corpo classes of society need to become class traitors and help workers. Nick Hanauer is the only person in living memory that I would consider to be an actual class traitor and champion of the people.

His involvement in the "Fight for $15" movements have literally lifted people out of poverty through substantial lasting democratic change.

I can't think of another person like him, if you know one please share their name because their stories need to be shared.

This should be illegal, outright.

Monetary systems and their economies are innately dystopian, and that can't be changed, but given that I'm a reluctant capitalist. Simply because the other option is much worse and incomparably more dishonest. But if private industry wants to remain relatively unrestricted it has to operate in good faith. Which means not pushing the envelope toward de facto slavery conditions, as is its nature to do. It needs self-control. If industry can't control itself, it's making the case for being hyper-regulated. And then its down-hill from there.

Industry and labor are in an inescapable social contract, the balance of which determines wider society's quality of life. Both factions have to commit to the middle, relatively speaking.

In part, that means that the job market standard doesn't evolve to six days of double shifts. Hire a second shift, and retreat from the type of psychopathy that advanced society wisely left behind.

It's self correcting. Corporations aren't needed for the software industry, they were just the best way to structure things for a while. If collaborating that way continues to become intractable forward looking, creative people (the kind you need to write good software) will find other ways to collaborate.

We need unions in this industry so badly

Do you work at an AI startup? What would the union do to help you?

A lot of people working these jobs are being paid millions per year, it's really not the right group to feel bad for. Also nearly everyone I know who does this loves it

There's a handful of people being paid millions. And they will expect everyone underneath them to have the same work ethic, for far less pay

A typical engineer at a top AI startup makes over 1M/year, mostly equity. You may think that the equity is worthless, but they don't and that's part of why they are willing to work hard

- What background do you assume a typical engineer has? (I.E., title, years of experience, internal level)

- Which firms qualify as top AI startups?

- What percentage of the $1M is paid as salary?

- How widely do equity packages vary for people with this background?

0-10 years experience (wide range)

Let's say top 30 by valuation

Depending on the startup, 10-25%

Very widely

I claim the median is around 1M at these top companies. Not the median offer, but the median current employee

You're claiming contradictory things

> A typical engineer at a top AI startup makes over 1M/year, mostly equity. [0]

>> - What percentage of the $1M is paid as salary?

> Depending on the startup, 75-90%

Which one is it? I've worked for multiple AI startups and have many friends in the space and I don't know that many are making $750-900k cash, but calling $100-250k cash + dreams and wishes a million doesn't add up either.

0. https://news.ycombinator.com/item?id=44698470

I read it backwards, 75-90 is equity. I fixed.

But I should make it clear that ordinarily I'd agree with the haters, but AI is actually different and so much wealth will be created that many of these unicorns will end up exiting for plenty. The people who got left out of the windsurf deal will still make millions (just not tens of millions)

> startup [...] mostly equity

So, a purported 1M/yr but it's in funny money.

Sure, but some of these companies are valued at multiples lower than public companies doing AI. It's reasonable to discount that but foolish to ignore it

If you're going to burn yourself out you better be paid millions because you'll need to afford the crash that comes after.

There is no crash if you like what you're doing and are surrounded by people who work hard and feel like they're on the same team

Work for a public company and get paid in real equity or start your own company if you want to burn yourself out.

Anything less is a monumental waste of time and energy.

Looks like you’re a founder based on your profile so you’ve also figured this out.

Burnout is: exhaustion, inefficacy and depersonalization. The crash happens when inefficacy kicks because that’s very market-dependent and not really something anyone can control. The easiest of those to control is exhaustion but it’s also nigh-impossible to measure. So I think our best bet is to keep a close eye on depersonalization and shift gears immediately if that ever happens.

Frankly this sounds like a cult mentality that I would cautiously endure for a lucrative salary.

Reply on news.ycombinator.com