The Automation Paradox | Hacker News Reader

The Automation Paradox

theatlantic.com

31 pointsergest64 comments

Are you kidding me? The article stopped making sense right at the beginning:

> But, perhaps surprisingly, electronic discovery software has not thrown paralegals and lawyers into unemployment lines. In fact, employment for paralegals and lawyers has grown robustly.

WHAT?! The statistics they chose from the US Department of Labor were cherry picked and out-of-context. Lawyers are seriously hurting right now:

> The national unemployment rate for law graduates has grown for the sixth year in a row to a whopping 15.5 percent, according to a report by the National Association for Law Placement. This unemployment exists despite soaring law school tuition.

Taken from:

http://www.huffingtonpost.com/natalie-gregg/mamas-dont-let-y...

http://qz.com/206705/the-us-lawyer-bubble-has-conclusively-p...

Another good read (don't have to watch the video to get the gist):

http://abovethelaw.com/2013/10/this-pretty-much-sums-up-the-...

You are talking about newly minted lawyers. If they have never a held paid position as someone's representative, or in some other way practiced law in exchange for money, they would not be able to get into any unemployment line as a lawyer.

Automation will be used as much as possible, up to the point where a human must be involved, either to exercise judgment or because that human's job hasn't been automated away yet.

Consider the farming industry. In a couple of centuries, it has gone from 90%+ of the population to a couple of percent, while the number of people being fed has increased several times over. That's automation at work, pretty fast in human terms. Did it cause a massive dislocation and unemployment when 90% of the human population was more or less replaced by machines?

Yes, it did cause a massive dislocation. And it isn't over yet, people are still moving to cities.

Yep. I didn't phrase that well - it caused dislocation, but not permanent unemployment. Society adapted, and new careers came into being that couldn't have existed when we were devoting all our human capital to manual labor on the farm.

> Did it cause a massive dislocation and unemployment when 90% of the human population was more or less replaced by machines?

It most certainly did. I'm not that familiar with what it looked like among laborers and displaced farmers (the Luddites come to mind, but they're too early); but this had a disruptive effect all the way to the top. Look at the long English tradition of hostility to the new world of manufacturing (William Blake, H.G. Wells, and J.R.R. Tolkien immediately come to mind); or read Fritz Stern's _Gold and Iron: Bismark, Bleichroder, and the Building of the German Empire_, which discusses at some length how the new, low-margin agriculture ruined the German aristocracy.

So the old class of exploiters who depended on an agrarian economic structure got replaced by a new class of exploiters who depended on an industrial economic structure. I cry no tears for them. And now the industrialist overlords are getting creamed by the information overlords. Notice a common thread here?

The industrialists won by pushing people out of agriculture to manufacturing, and then out of manufacturing and into services. Now, the information overlords are cleaning whatever remnants were there in production and pushing the rest out of service jobs. Where do you think those people will go?

It's not about the top 1%. They can stay filthy rich for all I care. It's about the rest of us, about our families and friends. With career in software, I could probably outlast most people. By then though, I'll have to work to provide for 20 people, who have no marketable skills whatsoever. That's the core issue. Automation is replacing classes of jobs, and besides the already automated labour-intensive areas and currently automated brainpower-intensive areas, I don't see any class left for people. At least not big enough to hold everyone who needs a job.

So maybe restructure society so people don't need "jobs"?

Yes, that's the point. It won't happen by itself though, and pretending that it's not an urgent issue isn't helping.

I think part of it for me is that there's a huge difference between "don't have a job" and "don't work". There's a perception that if people don't have jobs, they'll just sit around eating government cheese and watching tv all day. But I think hard work is in our bones, at least the bones of the interesting people. Heck, even if a majority sit around doing nothing all day, if we give those creative, hardworking people the opportunity and resources to act, they'll do amazing things.

And personally, I think the need for a "job" is the #1 impediment to creative, productive work by the people engaged enough to do just that.

At least "the old class of exploiters" offered legally enforceable employment for life (a paid servant could sue if dismissed for getting old, and a tenant was guaranteed the tenure of his farm at a fixed money rent not indexed to inflation; serfdom, meanwhile, had been abolished in England back in the 1500s). The industrialists didn't do that.

Here's US employment by category: [1] Only 14.1% of the workforce makes "stuff" - that's agriculture, mining, construction, and manufacturing combined. Those are shrinking slowly. Most of that is already automated. A century ago, that was 90% of the workforce. That's the "hollowing out of America" people talk about. The US manufactures more stuff than ever, but with far fewer people.

The big sectors remaining are retail, professional/business services, health care, and state and local government. (The Federal government is tiny - 1.8%.) Health care is growing. State and local government (which is mostly teachers and cops) is down a bit. Retail is declining slightly. There's a little growth in professional/business services.

Retail is struggling. America is littered with dead malls. (See "deadmalls.com") Online ordering hasn't crushed brick-and-mortar retail yet; it's still only 7.4% of total retail sales.[2] It's gaining another 1% or so each year. There's still a future in retail, but the jobs are low end.

"Professional/business services" is the category IT automates. It's where most of the good middle class jobs are. 12.7% of the workforce now. (This doesn't include finance, at 5.3%, down a bit, and also highly automatable.) This is the sector likely to clobbered next, as machine learning gets going.

That's what's going on. Take a good look at those numbers and evaluate your future.

[1] http://www.bls.gov/emp/ep_table_201.htm [2] https://ycharts.com/indicators/ecommerce_sales_as_percent_re...

According to the article, when computers start doing the work of people, the need for people often increases. I simply cannot take anymore the doom and gloom of the supposed AI revolution and how it will make us all jobless.

You should visit some manufacturing plants. Ask them how many people used to work there and how many do now.

Companies have a social responsibility to the communities they profit from.

If the entire plant is automated, how is that profiting from the community?

More likely the community is profiting from the company in terms of tax revenue and business fees.

That's laughable. Companies will usually do whatever it takes to pay as little as possible in taxes, evade regulations from environmental to labor, and attempt to get away with it by putting on a big show as "job creators" - meaning that if other people, little people, are brought here to spend money and pay their taxes, then the town should be downright grateful that the company is there to give people work.

I know it sounds hyperbolic but I have talked to local business leaders and they do indeed see themselves this way.

It's only a little bit hyperbolic.

If a facility isn't providing jobs to local residents, there's no leverage to get sweetheart deals on local taxes.

This is at its core a problem of elasticity of demand. Most factories are only tooled to produce a certain output, and business inputs are constrained more by their contract terms and raw material prices than they are by labor bottlenecks. As a result, the company is effectively moving from one bottleneck (labor) to another (resource constraints or market demand).

Compare this to labor-intensive industries like law software engineering. Increasing the marginal productivity of labor while there is sufficient demand for increased output generates that required output, until it reaches its new equilibrium. In these cases where there is substantially more work to be done than there is time to do it, and there's sufficient market demand to handle additional output (thinking especially with software engineering), then increases to productivity will generate increased demand without appreciably changing per-hour market-clearing wages.

I was at the bmw plant in munich and they said the plant was 98% automated. The guy said they did need someone to go around and change the welding tips every 8 hours.

Because fewer people are involved, they can generally make the products to a more predictable standard ("quality") and cheaper. Lets say the item the plant made with all those people used to cost $10 for consumers to buy, but the efficiency improvement from the manufacturing plant now means that it costs $5 for consumers to buy.

All over the town/country/world, people who used to part with $10 now only part with $5. They now have an additional $5 to spend in their community. The manufacturing plant has enabled many others to have their money go further - how is that a bad thing?

Which fool drops their prices like that :-)

Consumers should now pay $9.99 and we keep $4.99 profit in our community of billionaires

:-)

> Which fool drops their prices like that :-)

Anyone who has competition, be they direct or reasonable alternatives.

That's not how it works :). If the productivity doubles, the product will still cost $10, but the company will happily pocket that $5 improvement - and no, they won't raise the salaries of their workers, because why would they? There's zero reason to. Also note that "efficiency improvement from the manufacturing plant" usually means a) automating a part of the production process, and b) firing the workers who are no longer needed.

So now you have a product that's still $10, but some of the people in the community don't have a job anymore, so they can't buy it.

Of course, then a competitor shows up, driving the product price down to $4.50. The original company has to adjust, so it fires some more people and cuts down the paychecks for the rest.

The end result is: the product is cheaper, but the community is poorer, and some of its members don't have a job anymore.

But hey, isn't this the whole point of optimizing business through competition? Isn't the goal to make more with less resources and less people?

That is exactly how it works, unless there is a monopoly and no alternatives. You've forgotten all the people and communities benefiting from the lower prices, who (normally) vastly outnumber the factory workers. And the same factory workers benefit from the lower pricing in what they buy made by others too.

Yes the goal really is to improve productivity. If you want to do as much as possible locally with many people, then try subsistence farming. It has only been ~200 years since ~98% of people in the US worked in agriculture - now down to ~2%. Yet vastly more food is made and it is cheaper.

Trade and specialisation/productivity are what has driven prosperity and quality of life. It is very tempting to see all this as a zero sum game (ie the level stays the same and gains match losses). But in reality it is a rising tide lifting all boats (ie gains outpace losses).

There are temporary blips while change happens. To an individual life, that can be serious, and a humane caring society will address it through various means of support and opportunity. By far the stupidest way of providing support is requiring commercial enterprises to provide jobs and work to do that they don't require, increases their prices and helps their competitors, with no societal benefit. Note that this approach has been tried numerous times - it results in a stagnant economy, slow or negative prosperity, and doesn't really help.

Hang on, why would you drop your prices when your margin just increased? Why let "the community" keep their extra $5 and not pocket it yourself instead?

Financially speaking: what's the incentive?

Unless you are a monopoly, one of your competitors will drop their prices since they too will be looking for improvements. As Bezos says "your margin is my opportunity".

Say you have an idea of what the demand curve looks like. For the sake of argument, let it be (100-p), so 100 people would want it for free, and none would want it at $100.

If your marginal cost is $50, to maximise profit you'd set the price equal to $75 and make $625 net.

If your marginal cost is $40, you'd set the price at $70 to maximise profit and make $900 net.

You should visit some farms. Back in the 1800's agriculture was 80%+ of jobs. With the end of slavery and the evolution of machinery that number dropped to <2%.

Where was the economic implosion when automation took over agriculture?

Some believe that reduction of farm labor was the underlying cause of the great depression.

Whether or not it did, you don't see a lot of people complaining they're no longer working on a farm. And we've replaced all of those jobs with arguably better and safer ones.

You shouldn't have been downvoted. I was in a fully automated "lights out" factory producing PCs. Big trucks came up to the doors where the components (produced by robots in other factories) were unloaded by robots and passed from robot to robot until the entire machine was assembled. Once fully debugged, the entire factory was moved offshore to be owned by a group of the executives who would be able to sell to the company and hide all the income from taxation (the group included executives from "competing" companies to avoid any legal issues of "self-dealing".) So the billionaires made more billions by selling at high prices to our society while hiding their income.

I should add that they didn't automate boxing the computers, so that they could have a few people work each day doing boxing. We used low-cost humans (with degrees!) to box just to double check to make sure we were running the right products as scheduled for the factory. We could have automated that, too, but people are cheap. (Contrary to the beliefs of many, we more often stopped automation projects due to workers being too cheap to bother replacing, not because it was too hard to automate the job.)

I think it's different that a) we (you?) can build a robot factory to make stuff and b) we let them use it as a tax dodge

Fixing the tax system is a different problem, but no less complex - but cool factory.

Good point. From my perspective, the major problem with free trade is that it lets the billionaires move production offshore and avoid all taxes while hiding their obscene profits. (And they then import the products while laying-off US employees.)(And the guy who owned the factory, let's call him "Steve", made many billions offshore for which there is no US accounting at all.)

And it was a very cool factory - I had done automation projects, and some buddies of mine at the company were working on that project. As someone who had done robotic automation, it was truly amazing to watch and was probably the most state of the art at the time. We literally turned out the lights just to stand on the catwalk and watch the robots in the glow of the hundreds of LEDs on the electronics.

That's not free trades problem - that's this problem

Apple sells a million iPhones in Europe and now has a billion Euros in bank accounts in 23 countries. Which of the following statements is true:

Apple is a US company so USA should get the tax on that billion

Apple has the money in a German bank account so Germany gets to tax it first

Apple used sales people in Ireland to take the order on the phone, so Irish gov gets to tax it

Ten thousand iPhones were sold to Dutch citizens and shipped from France to NL. By a UK shipper. We all want a cut

It's a total fucking mess is what it is. In fact it reminds me of the BBC podcast Infinite Monkey Cage on race just now where the biologists just went "can we not use the word race, it is totally meaningless"

I suspect if you find a tax lawyer they have a similar view in "nation state" "soverignty" and "tax revenues"

I think the problem can be simplified. Apple and Steve hid the bulk of their wealth from all taxation (I have personal knowledge of this being true.) So figuring out what they are hiding and how they are hiding it is the first issue to solve. Once the numbers are clear, then it becomes possible to decide policy. The policy might be that if a company wants to sell in the US then it must employ in the US and pay taxes in the US or face sanctions (there are lots of ways to restrict sales and marketing of products. There are many channels and product lines.)

Datacenters are another example. Once they're built and turned on, my understanding is that they take maybe 35 permanent people to operate at the datacenter itself.

In the dozens of datacenters I've visited while working on projects for various clients, head counts are often far lower than 35.

In cases where the datacenter is being rented out to customers for colocation, the entire staff - which might be a dozen or less - is often working as glorified remote hands for an ops team located elsewhere.

I've also worked with a number of Fortune 1000s that have a sizable on-site datacenter footprint, but zero FTE resources dedicated to datacenter operations itself. Everything is divvied up among existing groups (facilities, IT ops, security, etc.), and occasional projects (structured cabling, rack and stack, etc.) are outsourced.

There's an old probably apocryphal story about Henry Ford III taking the head of the UAW on a tour of his new automated factory and asking "Think you'll be able to get these robots to pay union dues?" to which the UAW head replied "Think you'll be able to get them to buy cars?"

This hits on one of the most basic of economic principles which is often ignored. If the community becomes unemployed due to automation and is not reimbursed properly for the work done by the machine, then the community will not have the purchasing power to buy the goods produced by the machine. Thus rendering the machine useless.

True, but the real problem lies on the road from here to there. I.e. you could probably make half of the population totally unemployable while selling the output of robots to the other half. But will the first half just lie down and starve to death, or will they try to fight back? And, given the incentive structure of market economy, will we be able to get off this road altogether before reaching that point?

I work at a Manufacturing Plant - a Steelworks:

In the early 1980's there were 7000+ people directly employed here

Early 2000's about 2000

2011 (after a major restructure) about 900 people

That number is falling.

Our plant was largest employer in my town, now the University + tourism sector are the largest employers. At times I very much feel like I am working in a dead industry.

The drops in the 1980's were largely caused by automation and technological efficiency. That loss has already been felt. The losses nowadays are caused by shutting down/moth-balling successively more and more of our plant.

Competition from cheaper imports - particularly China is absolutely killing us. Unless you work in the industry it is difficult to comprehend the shear volume China produces. The market is hugely over-saturated.

All the advanced technology in the world can't compete with that. In 20-30+ years China will probably undergo a transformation of it's own as all the manufacturing moves to whatever the next big developing market is (Africa maybe??)

It's nice to say that companies have a social responsibility but at the end of the day they answer to their board and the shareholders. In fact they have a duty to act in the best interests of investors even if it goes against what would be best for their employees. That's the sad fact of capitalism.

"According to the article, when computers start doing the work of people, the need for people often increases."

For the Atlantic (or any given periodical), compare how many they employ now versus 10 years ago versus 20...

Number of jobs is only half of the equation though- are the wages of all of those new jobs the same? If automation replaces expensive skilled labor with inexpensive labor than there is still plenty of room for concern.

Most people are grossly simplifying this discussion. The worst for me still is those who yell "Luddite Fallacy" whenever this discussion comes up.

What most people are missing is that it's not so much automation but rather digitalization which allows for a completely differen kind of automation thats not just mechanical but instead has some kind of intelligence embedded in.

The point is that technology is simulating higher and higher levels of abstraction and thus are replacing not just human physical abilities but also our intellectual ones.

And if you want to understand what that means. Then ask yourself what profession did the horses enter once cars made them obsolete in transportation.

It's also interesting to notice that the last 100 years only new kind of job have emerged and that's computer scientists. Most other jobs today are really just modernized versions of older ones.

The luddite fallacy is in itself a fallacy and everyone who subscribe to that view really should explain exactly what new kind of areas humans will be going into. Surely they should already exist today.

And if you want to claim that new jobs are being replaced when technology takes some away then you really need to explain this graph.

https://plot.ly/~BethS/8/job-growth-by-decade-in-the-united-...

> It's also interesting to notice that the last 100 years only new kind of job have emerged and that's computer scientists.

I think you're missing a couple there, the number of professions that hold a sizable number of people has absolutely exploded since the 1950's.

Ranging from all the jobs that were created by the electronics revolution, the transportation revolution, all the new media and the IT revolution is much more than just 'computer scientists'.

How do I know this? Because at some point I was responsible for maintaining a table that mirrored the Dutch bureau of statistics annual table with the names of all the professions. The number of new professions that were recognized each year was absolutely staggering, but if you went back over time you could see an inflection point somewhere just after world-war II.

Which jobs are you talking about? I am talking about fundamentally new types of jobs.

If you want to be pedantic, even computer scientists were around 100 years ago (Ada Lovelace, for one).

Really, if you want to discount all the fields I've named on principle because to you those jobs are not 'fundamentally new types' then computer programmer also doesn't qualify as a fundamentally new type of job.

Assembly line worker wasn't a profession in 1912 either, it only became one in 1913 when the assembly line was invented.

Classes of professions have not been invented since about 500 years ago, typically the classes of professions that are recognized haven't changed much at all, and yet, you'd be hard pressed to find 'tour-operator', 'social media strategist', 'geneticist', 'nuclear reactor operator' and so on in a list of jobs from the early 1900's. And by that reasoning 'computer scientist' is just a sub-class of scientist, actually one of the oldest professions we have.

'Fundamentally' is a cheat, there are no new 'types' of jobs, there are only jobs that did not exist before and there are jobs that nobody holds anymore.

I don't see how that is pedantic at all.

Bringing up Lovelace however is quite pedantic and I think you know that very well.

Assembly line worker did not require any new skills. Instead it required a subset af the previous skills people had whether welders, blacksmiths, carpenters, painters etc.

So you haven't really pointed to any new jobs and no fundamentally new type of jobs isn't a cheat.

Furthermore I am not really sure what you are trying to prove. Even if I had missed a few (which I haven't) it wouldn't change the facts.

I'll supply astronaut instead and see you argue around how that really is just another form of 'traveler'.

I'll bite. An astronaut is basically a highly trained horse carriage passenger. Most of spaceship operations are already controlled by machines anyway, and they do it better than humans. Hell, when you see discussions about manned Mars missions, the recurring question is, "why bother with people when machines can already do it better?".

If we're going to be strict then there are two basic forms of work: thinking and moving stuff around. Everything else is a variation on those two.

Thats not even true.

Crafts are neither just thinking or just moving things around.

This is getting silly and if this is all you got lets just stop here.

No one is just an astronaut. Everyone who even get to get into space have some sort of profession behind them which is why they are hired to begin with.

I think the point is that indeed, there are no new "types" of jobs - only variations of the few basic ones. And automation is a problem because it eliminates types of jobs. For each labour-intensive profession you now have a corresponding machine. For each precision-related occupation you now have an industrial robot design. And nowadays we're coding up programs addressing brainpower-related jobs. What else is there? What other type of jobs we have left, and is there anything preventing them from falling prey to technological progress?

The example is a bit cherry-picked. You could argue that the process went like this:

* 1980s - "takes a lot of time and money to churn through old legal docs, let's hire a paralegal"

* 2010s - "it's easy to churn through old legal docs, making it much easier to manage multiple legal cases in parallel. Let's hire a paralegal to help me juggle all of these cases."

...

* 2020s - I've hired HireASuperParalegal.com to do all of my paralegal sleuthing. They help me manage the workload of four paralegals for the price of one.

Meanwhile, how about telemarketers? Have their numbers increased in droves like--apparently--paralegals? I suspect not.

Anyway, @ergest mentioned that the doom-and-gloom is tiresome, and it certainly is. Personally though, I just find it a fun thought-exercise to think through how we'll matter in the long-run. In the meantime there'll be a "new" class of workers whose jobs move from paper-pushing to being specialized software-users and from there... what? Certainly "computer-or-software fix-it guy" jobs will be universally useful from now into the foreseeable future, but will they pay well? :-)

I'm still a bit mystified by the people who believe that automation can, or will, destroy jobs. It takes a very weird view of the economy to see it as fundamentally static, fixed in a given form that can't be changed other than who or what does the work that's already been done. In nearly every corner of the economy there is an infinite amount of work that needs to be done, but of course not all of it can be done so only some of the available work actually gets done. But if you automate some of that work it's not like the undone work goes away, it's not like people won't pay good money to do that undone work.

> I'm still a bit mystified by the people who believe that automation can, or will, destroy jobs.

There are certainly huge local issues that happen when you shut, say, a steel plant or coal mine. We know that this happens. What would happen if we were to do the same thing to a whole tranche of jobs all at once?

Perhaps you're talking on much longer scales, but you shouldn't ignore the shorter term (which is still long in human scales, it's not like these things sort themselves out after a couple of weeks).

> But if you automate some of that work it's not like the undone work goes away, it's not like people won't pay good money to do that undone work.

If they'd pay more for that undone work than the work that's being done, why don't people shift jobs?

This is a great insight: "In nearly every corner of the economy there is an infinite amount of work that needs to be done"

I agree with you but i can understand how people can think differently. They are basing their ideas on what they see in front of them. And they don't see how it worked out before and they don't see all the additional work that can be done. I think it takes an imagination to understand that new jobs will form.

Yes, it is easier to find the dead-end roads when speaking of future societal changes. It is easy to panic about massive global unemployment in the face of automation. The fact that a conclusion is easy to come to doesn't make it any less likely. The patterns we see now are based on recent, past trends. What we can't see, or extrapolate from what we've experienced already, are the effects of accelerating exponential change in the tech sector. What I get from this article is the beginnings of the political rationalizations for denying basic guaranteed income. When your spreadsheets say there isn't a problem, when your economists tell you that work for humans is actually increasing, it is much easier to ignore the needs of the people on the ground.

I have been wracking my brain about this problem for a few years now, and I still can't seem to find the essential leverage point human labor would have against automated, robotic labor. I once thought it was our ability to synthesize and generate novel solutions to problems that would keep us in a dominant position within the economy, but that is quickly turning out to be a naive delusion. I don't fully understand the direction neural networks and machine learning are moving, but it looks like they will match us in those "uniquely human" areas of cognition very soon.

that is the thing though, why should we care about human labor having a leverage point? If all labor was automated, it would be a great net positive for humankind as a whole AS LONG AS it is not a requirement to be a laborer to receive a livable wage.

The problem is that as time goes on it seems less and less labor is available that can be done by the majority of the population, while the requirement to labor in order to earn a wage is not changing.

There is an infinite amount of work that can be done by humans that cannot be done by machine, but that kind of work usually does not command livable wages: learning to play a musical instrument, and becoming moderately competent at it, learning to build furniture with hand tools, and becoming moderately skilled at it, learning to be a good spouse, and being there when your partner needs your help.

It is easy to see how a large amount of occupations still available today can be automated, it is harder to see how the large number of people displaced by this automation will be able to manage.

Once automobiles became widespread, were the horses able to "retrain" into "the new jobs that became available"? Improving the cart is not the same as making the horse replaceable.

"If all labor was automated, it would be a great net positive for humankind as a whole AS LONG AS it is not a requirement to be a laborer to receive a livable wage."

This is where I am pessimistic. The political will is not there, not too mention the glacier speeds at which governments move in comparison to technological innovation.

While electronic discovery software has become a billion-dollar business since the late 1990s, jobs for paralegals and legal-support workers actually grew faster than the labor force as a whole, adding over 50,000 jobs since 2000, according to data from the U.S. Census Bureau. The number of lawyers increased by a quarter of a million.

I'm not convinced by this. Maybe paralegal jobs grew more than other jobs, but what does that say? Weren't other jobs also automated? Why then, if automation was what fueled the paralegal jobs' growth did it not also fuel other jobs to grow at the same rate? Might there not have been another reason why paralegal jobs in particular grew more?

For instance- did that growth in paralegal jobs have nothing to do with the growth in lawyers' jobs? Could it just be that a quarter million more lawyers needed at least 50,000 more paralegals? Is it possible 50,000 paralegal jobs are actually too few for a quarter million lawyers?

Also- just growth over a period on its own doesn't mean anything. Was growth more than what was expected? What was the trend in the decades before and after the "late 1990s"?

Or more generally: is that amount of growth really surprising? And why is it surprising?

"Commoditize your complements"

Reply on news.ycombinator.com