I think Paul is treading ideas with some interesting implications for him, personally. He is the head of an organization that funds companies on the basis of a written & oral exam, whose principal selling point is the transmission of "credentials" upon the kids they pick. Part of it is the kids' merit, but part of it is YC's past success.
People go to YC for many of the same reasons people went to Stanford -- past seed stage, it's still largely a credentials game. So you have to ask whether the same potential for "corruption" is present. Are they failing to transmit credentials over generations because it's a revolutionary model, or because it's only been around less than one generation?
What if someone did a sweetheart deal for a YC startup as a personal favor/quid-pro-quo/etc? What does it mean when founders who got rich from a YC company start funding new YC companies because they are YC companies? Is that "corruption" in the same sense?
Some of what you say is true, but it's not true that the principal value of YC is the brand name. The YC brand has little weight to the customers of a startup, and without those it quickly dies. YC's principal value is the advice we (and the other YC startups) offer.
The part you're right about is that "admission" to YC is structurally a lot like college or grad school admissions. The difference, though, is that our choices are immediately tested. College admissions officers do not, as far as I know, follow up on the careers of the people they accept and see how they did. Whereas if we pick a bad group, that fact will immediately be thrust in our face. This causes us to work very hard to try to pick the best people.
It was being in a sense a colleague of college admissions officers that made me think about such topics. We're the next bottleneck downstream from them, and it quickly became clear to us what a bad job they were doing:
The YC brand has little weight to the customers of a startup, and without those a startup quickly dies.
It would be interesting to see numbers on that. Informally, I noticed that back when reddit was good, anyone who was asked had either heard about it because it was YC-funded, or heard about it from someone who heard about it that way. I'm not certain, but I would probably bet that the submitters of, say "Man sentenced to NINETY THREE days in jail for taking 6 seconds to sit down! Outrageous!" or "Warrant Issued For Man Paying Fine With Pennies" didn't find out about it the same way.
Reddit's a unique case because their initial user base (like News.YC's) consisted of other YC founders. And since they couldn't afford any form of promotion that cost money, news of the site spread mainly from those users.
I'm finding it difficult to square your assertion that Reddit is "unique" in this respect with the assertion that it's the web that's "making marketing and distribution free".
The only way that web marketing is "free" is if product awareness spreads between customers, or is pushed through the founders' personal contacts. Wouldn't this imply that we could expect this phenomenon - the initial user base being in some way close to the founders - to grow as the industry approaches the asymptotes you assert it is heading for?
And other followers of your essays, and people who got rejected from YC that first time. I checked out Reddit at first because I was curious what the folks who did get in were up to. And they happened to have built something cool, so I told a lot of my friends and family, and who knows who else they told.
I think it's a fairly good bet that Reddit did a lot better with their initial group of users being seeded from folks who'd heard of you or followed comp.lang.lisp, than if they tried to launch something among their UVa friends.
College admissions officers do not, as far as I know, follow up on the careers of the people they accept and see how they did.
I'm not sure that's true - in the UK at least, universities often issue press releases when one of their alumni does something notable (like becoming the Prime Minster of Thailand for example, which Oxford is crowing about at the moment).
And most universities make a big deal about any Nobel prizes.
And I've seen in Australia a follow-up 6-12 months later to see if you got a job / further study.
But I've never seen a substantial follow-up of the average population of a uni.
Possibly because the UK is obsessed with league tables for universities, secondary schools and even primary schools. The University of York send me an annual survey to find out what I'm up to - I'm pretty sure they gather this data on all (willing) alumni, and where the data shows them in a good light they certainly make no secret of it at open days, etc.
Same for the UK university I attended. The survey is badly design, pretty much designed with the assumption that a graduate would be working for a big company. There's no way of reporting freelancer/self employed for example.
Yes most universities believe graduate -> graduate training programme, percentage thereof is measure of institutional success. Maybe that was true the last time the admissions officers were in the job market, but it hasn't been for 20-30 years.
Nobel Prizes per undergraduate is an interesting measure. More than one in every 1000 Caltech undergrads wins a Nobel Prize, and that number rises substantially if one restricts just to science students, not engineering (which doesn't have a Nobel). This suggests that the admissions office there is doing something right.
Are you controlling for those not admitted that went to similar caliber schools, or have an example of how that statistic compares to those schools with a similar quality applicant pool?
You can show fairly quickly that the success of Caltech alumni in sciences, as measured by awards, is comparable only to the École Normale Supérieure (and possibly Phystech in Russia depending on how you count). Check out http://www.arwu.org/rank2008/ARWU2008_A(EN).htm
The column 'Alumni' compares the number of Nobel prizes or Fields Medals won by Alumni to the number won by Harvard (in percent). Caltech scores 52.8 on this, and ENS 53.4. But Caltech and ENS are by far the smallest of the universities on the top 100 list (unless I am mistaken). Both Caltech and ENS enroll about 200 undergraduates a year, with ENS enrolling only about 100 of those in science and engineering. Caltech has a population of another 1200 graduate students.
One thing to note about Caltech, ENS, and Phystech, they are extremely small, extremely selective, have no legacy admissions, limited or no affirmative action, and pay heavy attention to competitive exams, in the case of ENS and Phystech, administrating them.
Caltech admissions struck me as particularly clean. I think part of this is that they're very much into maths and sciences, fields in which SAT score do a great job of determining proficiency. You either know the answer or you don't. On top of that, they give back to their alumni in other ways; one couple that donated 200M to the university had their kid denied, but the university named a star system after them.
Even more successful would be Cambridge University. 83 Nobel Prize winners in total. (31 Nobels from Trinity College, Cambridge alone!) http://www.cam.ac.uk/univ/nobelprize.html
That said, Cambridge is a far bigger University than Caltech and others, but a large proportion of those degrees are awarded in non-science fields.
Ah, one issue from the SHJT list is that they weight more recent awards more heavily. Cambridge was a preeminent force in science and engineering in the earliest half of the century, but seems now to have diminished somewhat relative to others.
I know that Swarthmore, my alma mater, has had 5 Nobel laureates (if you count one Economics prize), and is comparably selective, and only slightly larger than Caltech. Note that it only has 20-30% as many science grads as Caltech [rough estimate].
I think it's actually Nobel Prizes per alumnus, i.e., it includes advanced degree-holders as well. (See http://www.admissions.caltech.edu/about/notable, which states that "Caltech has over 20,000 living alumni—of whom 17 are Nobel Prize-winning scientists", and goes on to list lots of notable Caltech Ph.D. alumni.) So the graduate admissions committees are apparently doing a good job, too (though clearly they screwed up in my case :-).
You can get a Nobel Prize science prizes for "physics, chemistry, physiology or medicine" but they often give them to 2 or 3 people so let's say 17 / (~8 per year over ~40 years) which gives you ~5%. That sounds great but I think most Nobel Ph.D. have 2 or 3 schools (if not more) in their background so it's really more like 2% which sounds about right for one of the top 25 science schools in the world.
I've heard two different figures for Caltech. The 17 figure I read on Steve Hsu's blog; presumably he has the same source as you, although I assumed he was referring to undergrad only. That assumption seems to have been a mistake. The reason I made the assumption is that while I was a postdoc, one of the oncampus papers ran an article that set the number at 27, if I recall correctly, roughly equally split between undergrad and grad. In either case they're a long way ahead of other US Schools.
Access to funding and marketing continue to matter a lot for most startups. Since web products are so easy to change and refine in response to feedback the initial quality of a product doesn't matter too much. As long as you can get users to provide feedback and funding to stay alive while you refine the idea you have a much higher chance of success. I'd argue that YC provides both: more attention and better access to funding, particularly because YC's track record at picking winners. The better you get at picking winners the more important the YC brand becomes.
College admissions officers do not, as far as I know, follow up on the careers of the people they accept and see how they did.
In Canada, one of the strongest selling points of universities is what fraction of their students got a job in a related field within X years of graduation -- so at least here, institutions do track this sort of data.
In addition, there's a lot of work done to validate admissions processes based on the success of students during their undergraduate years; universities have extensive "black books" which tell them how a student graduating from high school X compares to a student graduating from high school Y given the same nominal high school marks (here in Canada, admissions are done primarily based on high school marks -- we don't have any equivalent to SATs).
I've heard that U.S. universities have similar "black books." It's too bad they don't publish their data, because that would be an interesting way to compare the performance of high schools.
Companies also have unofficial "black books." We know, for example, what schools to focus our recruiting efforts on based on the quality of the applicants we see. We do look for correlations between the new-grad applicants we make job offers to (we're rather picky :-) and the school they graduated from.
The fraction of people getting a job in a related field isn't a good measure except for trade schools. Often, the most successful and interesting people do something unrelated to their undergraduate degree. Or they do something other than a "job".
France too, at least my school. Just had an email from them: "Please let you know if your employement situation changes in any way." Seems like they want to get some good data on the effects of the crisis.
>but it's not true that the principal value of YC is the brand name
Yes, not to the customers of the app, but the interest the YC brand generates in "insider" circles, which helps bootstrap popularity I think you are underestimating.
> YC's principal value is the advice we (and the other YC startups) offer.
Yes but as before, the "brand impact" in the blogosphere (I feel dirty writing that) is still a huge thing. If that comes under the umbrella of advice, then that makes sense.
"College admissions officers do not, as far as I know, follow up on the careers of the people they accept and see how they did."
I checked for an article from the Harvard Crimson about the degree to which the admission office follows up on admitted students after they enroll, although I didn't find the link just now. You've received other replies mentioning various ways, with various degrees of formality, that admission offices can and do check their results. Now that many admission offices are making very fine distinctions among applicants as admission rates plummet below 10 percent, there will probably be growing interest in validating the criteria on which admission decisions are made.
MIT admission office McGreggor Crowley, in one of his admission travel sessions in the last year, said his office has done studies of which students perform best in upper division courses at MIT. Taking really hard classes in high school is a better predictor of upper-division success at MIT than having a spotless G.P.A., for example. He asked a large audience of high school students how many had done baby-sitting, and very few raised their hands. He said if that is one thing that kept you busy in high school, be sure to mention that on your MIT application.
We're (TicketStumbler) a pretty good example of a startup that hasn't really been affected by being a YC company. [1]
I did YC because it seemed interesting and I needed a good excuse to quit my job. [2] Tom was bored and wanted to get out of Ohio.
When we launched on Techcrunch, we were met with widespread apathy. Techcrunch readers, by and large, weren't our target market and didn't give two shits about a ticket site. We expected this to happen, and it was still cool to be featured on Techcrunch anyways :). We weren't creating a new platform or becoming the next facebook - we were just solving a simple problem (finding tickets).
The investors we met with didn't care that we were with YC (many didn't even know), they cared that we were making money in an expanding market.
Also, our most targeted/best traffic has been from places that didn't know/didn't care who YC was (e.g. Ticketnews.com, Thrillist). Same situation for our users.
Finally, as Paul alluded to, the YC name can only do so much for you, if you don't have a good product none of it matters.
[1] - I'm only talking about the YC brand name/credentials, not the program itself, which certainly had a positive impact on TicketStumbler.
[2] - When much of your family lives in Detroit and has been recently or is, unemployed, it's difficult to say your ditching a cushie 60k per year job to do a startup. Having "backing" makes it much easier.
One could easily amend this to include "After Perceptions".
As a young geek, I was often passed over because I didn't look like I could do the job. It started with little league and continued on through high school and into the work place. I even quit one of my better jobs because a slicker looking but far less talented peer was promoted into what should have been my job.
No more. As a mild mannered adult geek, I still get the same pseudo-negative first impression reaction until I open my mouth and show them how much they'd benefit from me and technology.
The Performance Economy was a long time coming (and probably taken for granted by anyone under 25), but now that we're there, there's no going back.
On a complete tangent, this is one of the perks of living in a foreign country if you speak the language well - instant recognition that you have mastered their language, and that you must therefore not be a complete moron.
The downside of this in Europe, for me, has been moving about 4 hours away (absolutely nothing in the US, in terms of distance) and completely losing this advantage.
You'll have to start over. That's not such a bad thing. So, Italian before, what now?
It's good to still have a bit of an accent. I knew an American who had mastered Russian to the extent that he was indistinguishable from a native speaker. There were, however, glitches that would show up in his speech - an occasional wrong phrase or something. Because he spoke perfectly otherwise, Russians sometimes thought he was a dumb Russian. This made me realize that you want people to know you're a foreigner, just a foreigner who speaks amazing X.
I never spoke Russian as well as this fellow did (and most of it is now sitting freeze-dried in my brain somewhere), but there was one funny thing in my case. Much of what I learned came from reading, so sometimes I would speak Russian like a 19th century novel. This made my friends laugh, but they also liked it - Russians are big on literacy.
Edit: your core point is quite right, and is easy to verify by the impression it makes when you meet a non-native speaker who speaks excellent English.
Agree on the accent point. I knew a Dutch girl who'd lived in the US from the ages from 5 to 11, so she spoke English with an American accent ... but with the vocabulary of an 11-year-old. If she'd garbled the language just a wee bit she'd have been impressively good; as it was, she just sounded dumb.
But I'll agree with David that being an American that speaks the local language can be quite a novelty in Europe. I'm in the same boat. :-)
>On a complete tangent, this is one of the perks of living in a foreign country if you speak the language well - instant recognition that you have mastered their language, and that you must therefore not be a complete moron.
Absolutely. I'm a partner in a small educational venture in Taiwan, and my sales conversion rate is actually significantly better than that of our local staff. I don't believe for a second that it's because I'm that good. Having people shocked at my Chinese skills definitely gives me an extra oomph.
I'm not sure if Chinese people in the US would get the same credit.
> I'm not sure if Chinese people in the US would get the same credit.
I think the problem there is that you have no idea what background the person has. There are plenty of ethnically Chinese Americans who of course speak English perfectly, whereas if you're obviously "not from around there", people are going to connect the dots faster.
Seems like there are more efficient ways of proving your intelligence than moving to country with a language that you don't speak natively. But if you do it for other reasons, I guess its worth using to your advantage.
I think that in the United States a lot of the pressure on students to succeed on entrance exams are in terms of getting into more prestigious schools has more to do with desired prestige and college experience than future success. Few parents would unequivocally claim that going to Harvard over Big State University = more success for their kids. For parents and their kids I think its more of matter of wanting the prestige and college experience that's possible at Harvard.
I think it's effectively like buying an advertising and PR firm to work for you in the job market. Word of mouth and demonstrable talent substitute for the expense to the degree to which you stay in the (sadly) smaller circles who aren't significantly influenced by PR and advertising.
PG claims that a shift towards smaller companies results in people being judged on their performance rather than based on factors which do rather poor jobs of predicting said performance; my experience with tarsnap indicates otherwise.
I have a small number of customers -- somewhere between 10% and 20% -- who are using tarsnap based on a recommendation from another tarsnap user; but the vast majority of tarsnap users signed up based on some other factor which is at best predictive of the quality of tarsnap. Reasons I've heard include:
* I am the FreeBSD Security Officer,
* They find my blog to be interesting and well-written,
* I have a doctorate from Oxford,
* They like the fact that I made the tarsnap client source code available, and
* I beat them on a mathematics competition 10 years ago.
There are a great many instances of companies succeeding with inferior products as a result of having superior marketing; while I am fortunate that people are interested in using tarsnap for reasons which are quite unrelated to the quality of the tarsnap code or service, the fact remains that the market isn't (in PG's words) "[taking] every organization and [keeping] just the good ones" -- rather, the market takes every organization and keeps just the ones which rank best on a nebulous combination of quality, marketing, and luck.
Yes, credentials are an imperfect way of predicting quality; but are they any worse than the marketing and luck which the market introduces?
Initially the users of a new product will have random motivations. They can't all come from recommendations when you've only just launched. But a year from now the number of users you have will depend mainly on how good tarsnap is.
Certainly, once you get some momentum your startup will survive or die based on how good it is. But if I hear cperciva correctly, had it not been for his credentials lots of users might not have even tried tarsnap to begin with, and I assume that would also translate into far less users a year from now, even though the product would be the same.
I have to dispute the notion that the "initial seed" is irrelevant. Given a fixed "interest rate", the number of users at any time will be proportional to the initial number of users; and given that the initial number of users can easily vary by an order of magnitude based on external factors, I'd say this is pretty significant. (Of course, it's quite possible that the growth rate is not independent of the initial number of users.)
There's also a more dangerous factor to consider: The number of users of a startup is far more visible than its growth rate, so it's entirely possible that great startups end up shutting down due to a disappointing apparent lack of traction when -- if they had stayed around for longer -- their inherent advantages would have produced enough growth to make them wildly successful. Even if we suppose that the ultimate potential of a business is entirely dependent on the inherent quality of the product or service it produces -- which would be rather stretching credulity -- then the external factors of marketing and luck still have a significant impact on whether companies remain in existence long enough to reach their potential.
I didn't say the size of the initial user base is irrelevant, just that in practice the growth rate matters more.
And while growth rate is not in itself visible, what drives it is something that's usually even more visible than the size of the user base: the quality of the product.
while growth rate is not in itself visible, what drives it is something that's usually even more visible than the size of the user base: the quality of the product.
I'm not convinced. Most startup founders think that their products are the greatest thing since sliced bread, but they're often also aware that they are biassed. How many startup founders haven't at some point been disappointed by the size of their user base and said to themselves "gee, I think what I'm doing is really cool, but maybe the market for this isn't as big as I thought it was"?
I'm sympathetic to this argument, but the "long term" could easily be quite long on the timescale of a startup's solvency. FogBugz, for example, might well not have survived without the initial user base provided by the readers of Joel on Software. Having that promotion channel could easily have meant a factor of ten in the initial customer base compared to a "normal" startup. And that factor of ten represents more than three years of growth even if you double your users each year. Of course, with growth like that presumably you could raise money to increase your runway, but that adds the complexity of dealing with investors and adds more time for things to go wrong.
i'm having difficulty imagining a bug tracking? software having viral growth ... so i guess the 'interest rate' is steadily low; therefore, initial capital is important
Bug tracking software, like other tools for programmers, often travels as a virus with team members. You hire some guy, he says "why are you using this horrible crap we used fogbugz at my last job and it wuz not teh suck", and so you try it.
Maybe we can call this resume-viral: products that are passed around by virtue of become recognizable bullet items on resumes.
I think that in case of tarsnap it is relatively hard to make the decision if you don't have basic knowledge in cryptography, at least. That's why the information assymetry occurs and thus signalling has much more impact on it.
- if they survive despite all, why do we permit them?
If the market was regulated to set a cap on the size of a single company's workforce (or some other metric, perhaps market capitalisation), we'd also avoid the 'too big to fail' problem we have at the moment.
Not to mention quasi-monopolistic practices like large chains killing off small local competitor with an externally-funded price war (then raising prices once the competition disappears).
Is it that we need the largest companies for some large-scale activities (chip fabs, aircraft manafacture), or that they can be efficient at scale (fast food, wal-mart)?
Would the world/market be a better place if companies were forced to fission at a certain size?
I would really like to see more analysis of what structural differences point towards the lots of smaller firms that pg sees, and whether this is a short blip in time, or a long term trend.
Also, I hate to dredge up that ugly topic "politics", but currently, in the US, if you or your firm fail, that's great in terms of opening up space for others to try, but a little scarier in personal terms, with regards to things like health care. Long term, who knows how things like that will play out.
> If the market was regulated to set a cap on the size of a single company's workforce (or some other metric, perhaps market capitalisation), we'd also avoid the 'too big to fail' problem we have at the moment.
I don't think either of the metrics you mention would solve the "too big to fail" phenomenon. The fact is that "too big to fail" is a misnomer, which has actually been used to represent the concept "too connected to fail".
I.e. it is not Bear's/AIG's pure size that prevents us from allowing them to fail, it's the connectedness that they have in our financial system, which depends on a web of millions of transactions per day between every large bank. Then, if a big bank were allowed to fail (not that I'm endorsing this theory), that would cause a cascade effect where banks started charging higher rates for those millions of transactions, which would cause less of them to happen, which would cause even higher rates, etc. etc., ending in a completely broken financial system where each bank wanted too much money for each of the many transactions they depend on to exist in their current form.
So, I don't think that limiting the size of companies in the financial sector would prevent the "too big to fail" phenomenon by itself. Which doesn't mean it's not a good idea; just that I think that particular example doesn't suport the idea that well.
(edit: and read Coase, like davidw's comment suggests :)
this entire "lots of small firms are the future" meme is probably just wrong. paul graham knows a lot about websites but very little about the rest of the economy. show me the ten person company that can produce the equipment to generate and transmit electricity. or food production. where are all the ma and pa grocery stores? and banking...we have fewer banks now than ever and the number is still dropping. and PCs...making a PC is easy....so what happened to the 10,000 strip-mall PC makers?
show me real examples outside of the web industry that this theory is even remotely correct
I have to disagree. it is true that the only reason we see small companies in the tech sector is that it is a young market that is still changing. but this change will continue, all indications are that tech will only change faster in the future. this means that the tech sector will perpetually act as if it were a new market.
Any industry where the capital costs are low, or where a small group of niche specialists can add a significant multiplier to the value chain are candidates for this sort of effect. As more industrial tasks are farmed out to large companies that specialize in flexible production it is possible that you will see this show up in industries that are currently have strong vertical integration.
For example, ten people can't mass-produce a car. Ten people can provide spcialized bodywork and personalization of a mass-produced car. Ten people could probably design and oversee the outsourced manufacturing of discrete components in the car. Ten people could probably manage a company that deals with other small firms to create bespoke vehicles based upon mass-produced frames and engines. As you increase the specialization and make more parts interchangeable it is possible to decentralize and distribute a production chain to the point where a large collective of ten person companies can provide a higher-value good at a lower per-unit cost than a single large company.
Certain activities will always need large, high-capitalization organizations, but it seems that more and more of these companies and factories are being forced to become more flexible and nimble which is leading to a point where a larger and larger part of your world can be serviced by smaller teams than by large enterprises.
How do these things change with time, though? Do they always, invariably drop? Might computing conceivably go through another phase where things get consolidated into some kind of big servers that cost a ton of money? Doesn't sound likely to me, but I'm not very good at predicting the future. Actual research into these trends would be interesting to see.
no, that thins margins even more and drives to even faster consolidation. i can make PCs in my garage, yet there are only three real players in this market - dell, hp, apple. the lower the barrier to entry, inevitably you will erode margins to the point that only a large player can stay solvent.
Retyred is on to something here. Outside of web startups I think the trend may have actually gone the opposite way over the past 20 years or so. Companies are getting bigger, not smaller. Here are a few:
Retail: just went through a mass-consolidation. The more dynamic firms, the "places to be" if you wanted to do something interesting in retail, were the ones doing the consolidating. Think Wal-Mart, Home-Depot, Walgreens, Best-Buy etc... While it may not be interesting to the crowd here, innovation in things like supply-chain has been stunning.
Consumer Goods: P&G has been unstoppable, and that's where most people in that business want to go because that's where the interesting things were happening.
Consumer Electronics: Dell, Apple, Sony etc... have crushed everyone.
Banking: Ditto. Where did all the ambitious young financial engineers want to go? Easy: Goldman-Sachs. Not that innovation here was a good thing... but that's another story.
Agriculture: ADM, Monsanto and the agri-business giants have almost completely taken over. Small farms are near-extinct.
This list could go on for a while, but I'll stop. I think you see a similar winner-take-all pattern playing out in most industries. It'd be interesting to see the demographics of all this: the percentages of people employed by institutions of different sizes.
Perhaps pg is right that it's starting to happen though. Maybe the web world is at the vanguard of what will be a broad counter-trend. I suspect that won't be as strong a force as he thinks though - the advantages of being big are still too great in too many fields.
Then again, maybe the mass-consolidation was artificial. Maybe all the M&As owe as much to artificially cheap money as to natural advantages of scale. I guess we'll see.
I honestly don't think that 'cram schools' are particularly effective. I taught for the Princeton Review for a while ($20+/hr when I was 18 sounded pretty good). In my experience, the average student would only gain about an extra 50 points the SAT (relative to what they would have gained studying on their own). If memory serves, a standard deviation (on the old 1600 point test) was usually in the neighborhood of a 110 points.
Increasing your score by less than half a standard deviation was nice, but it rarely made or broke your college application.
Applications to top schools are extremely competitive; I could easily see 50 SAT points making or breaking an application. I was under the impression that these days your application has to be more or less perfect to even have a chance to get into an Ivy League school, unless you're a legacy or a URM.
Amherst isn't quite Ivy League, but I had a bunch of friends there with SATs in the 1400s, or even the high 1300s. I believe the median (from stats, not from friends) is about 1460.
It was fairly scary how many of my friends had 1590s or 1600s, though.
In India they seem to make a huge difference. Friends and relatives who went to "IIT Preparation Courses" (6-12 months, at a boarding school, for getting into one of the IITs, the top technical colleges in the country) usually did vastly superior to others (of similar ability) who did not.
Agreed with the point that on a dollar-per-SAT-point basis, there is little rationale for choosing commercial SAT preparation over simply self-studying with the released previous exams. The most decisive thing to do to raise SAT scores is to read widely and avidly in English.
Paul said: History suggests that, all other things being equal, a society prospers in proportion to its ability to prevent parents from influencing their children's success directly.
I am studying development policies and growth economics this year. His statement really doesn't count for the countries who have very little wealth to begin with. In fact, he needs to clarify that it is better for wealth to be passed through families, than for that wealth to be destroyed.
Paul seems to be hitting on the optimization by proxy issues seen in many areas, even in pagerank. Google looks not for quality of a page but for indicators of that quality, namely inbound links by quality websites. Similarly exams and credentials do the same. If you can write about these books, you know literature. If you got a degree there, you must be very capable. And in the first generation it works.
However, as people become aware of the ranking mechanism, they try to get ahead by overoptimizing for the narrow areas inspected. So pages do linkfarms and black-hat SEO, students goto cram schools, etc. etc. essentially producing the exact opposite of what was initially sought. I believe Joel spolsky had a good article on a simillar pattern in sales.
The question is if an algorithm that affects its input data can be made so that it adapts to the changes and gradually reaches some kind of equilibrium with the input data or stabilizes in some other way. As paul suggests, one way would be to make the system more transparent, or even 'social' such that the participants have a way to update it in an agreed way. This would keep the minority from cheating on the majority.
Doing away with such filters though might be a bit harder than we think, especially given the breadth of their applications. As aristus commented, even YCombinator has an examination system. In fact it wouldnt be too hard to imagine cram schools for YC given enough incentive. I fear this is a problem we have to either solve or live with.
Say you were given 19 yr olds at the end of Year 1 of Uni. You have 2-3 years part time. How would you get them in? Activities? Interview coaching? Resume building? Portfolio building? Specialised learning?
Intresting question, but the feasibility of it is not dependent on my ability to give a good answer. Regardless, I'll give it a shot for argument's sake. Of course everything you say is valid. you would need to do these things.
I would also mine all the successes, details of interview process and characteristics of successful applicants and their applications for commonalities and then see how many of these can be easily added to an application for 'bonus points'. Of course it depends on how black-hat you want to get about this.
The first thing that came to my mind though, having read pg's essays, would be 'teach them LISP' :)
My intention wasn't to corner you, just to start a discussion. For arguments sake, let's get about as black hat as a Chinese mother with a shot at Oxford.
I don't know much about YC or what influences them apart from fragments from pg's essays.
Insisting that LISP or some other language no one uses is absolutely the best tool seems like it might be a bonus point. But I think that is as a bonus. What would be the core? They say smartness & an attitude that they are going to do something big with or without the program. The latter can be gamed. The former, I don't know much about the criteria.
I'd probably build their social network: connect them to hundreds of influential sources of free marketing. If you're judging performance, the initial performance of a connected founder will greatly exceed an unconnected one. If you don't wait long enough to watch market effects take over before judging which startups to keep, you'll only end up with the networked ones.
If the leading criterion for admission was concrete evidence that the applicants collectively have the ability and motivation build things people want as measured in users and sales...
...then a cram school for YC would be a program much like YC in which younger teams built smaller, more easily guided technology start-ups.
If the leading criterion for admission was concrete evidence that the applicants collectively
Big if. I'm sure that any self respecting University will tell you that the best prep for their University is anything that makes the candidates smarter, more driven & more likely to succeed. Everyone uses indicators. Some are better then others. Maybe some can never be gamed (I doubt it) but as actually mentioned in this essay, the way around credentials is making entry less important.
But on that note, isn't the biggest criterion for university transfer admission is who admitted you in the first place and your consequential success at that school?
It stands to reason if one wanted to predict if a group would be successful creating a successful startup that a good predictor would be if that group had stated a successful startup in the past.
I'd assume there are other criteria, like: is what this group doing within our scope of expertise? Can we add value to this startup? Will admitting this group best add to the quality of this YC class? Will we make money on this investment?
That's a very interesting question. Above all I'd have them build things. We like anyone who's created an open-source project we've heard of, for example.
I'd also try to get them jobs at existing YC startups. Then they'd both learn how startups work, and (if they were good) get recommendations from people we know.
sadly i think college credentials in the US have lost their luster because its always a let-down. without a doubt there is only one school that i have never met a single retard from: caltech. cmu is a close second. every other school has let me down by graduating some goof who i have interviewed or worked with. never again will i get uptight or impressed by stanford, mit or the ivies.
my attitude is that i do not care where you went to school. in an interview, if you ace the questions, why do i care if you went to ball state? so what, maybe that is all you could afford. on the other hand, if you can't answer my questions, why do i care if you went to stanford?
furthermore its likely that most people here do not know what the best colleges in korea, japan, china and india really are. so best to just skip the education section of the resume entirely
"... every other school has let me down by graduating some goof who i have interviewed or worked with. never again will i get uptight or impressed by stanford, mit or the ivies. ..."
Don't confuse 'intelligence' and 'education'. There are plenty of smart people who have never been to University. Likewise there are plenty of people who go to University who graduate but can't think beyond what they have been taught.
any type of creative field runs into this problem a lot. a lot of people reach a certain level and realize that they don't actually have any talent for X, they were just good at following the directions.
I agree with this sentiment entirely. I went to a public magnet high school with SAT scores averaging around 1480 and at least half our graduating class every year ends up in an ivy / mit / stanford. I'd say more than half of those who made it to the top tier got in from legacy / extracurriculars / obsessive dedication to good grades and making teachers happy as opposed to genuine intelligence and love of learning. Also curiously, like you said, everybody from my graduating class who went to caltech, although not necessarily teh uberg3nius math wiz, was at least minimally competent in some technical field.
Although one could make the case that a lot of those schools, especially the ivies, are picking people based on their ability to lead and 'make a difference' in some general sense rather than pure raw ability. Therefore they may intentionally focus on apparently manifested soft qualifications like achievement in extracurriculars, personal discipline as exhibited by a flawless GPA, or well-established personal connections.
This is all strictly re: undergrads of course. I've never seen a single retard get admitted to a top-tier PhD program in science or engineering.
"This is all strictly re: undergrads of course. I've never seen a single retard get admitted to a top-tier PhD program in science or engineering."
That's not the case as I've witnessed it. I won't directly name names, but I am thinking about a school that is ranked number 4 for graduate engineering on U.S. News.
U.S. News has a fairly broken ranking system. I think they try and set things up so schools jump around with little rime or reason so people will buy this years ranking vs just reading last years. I mean if if my school will rise and fall 5-10 points while I am their why try and give a number just list those that make the cut.
the transmission of power between generations
is also happening within the startup environment. How? Successful entrepreneurs are funding their friends or friends' friends. Furthermore, VCs are known to fund people within their close network. Power Bloggers are known to cover their friends' startups or their friends friends'.
Luckily technology is fairly cheap, learning how to do stuff is free and only requires commitment and time so we can still turn our ideas into products and put it in front of millions of people without bloggers, angels, vcs.....
"History suggests that, all other things being equal, a society prospers in proportion to its ability to prevent parents from influencing their children's success directly."
Can we assume that pg is in favor of restoring the estate tax?
Actually I don't think inherited money is that dangerous. Unless it's coming from more than a generation back, you don't even get it till you're old yourself. And when people inherit money young, if anything it seems to screw them up rather than giving them an advantage.
I'm assuming that you mean national power. Plenty inherit the family business and/or money and wield local power.
In formulating my reply, I think I came around to the other point of view. Bad things can happen, but inherited money is probably a minor enabler in this, and just the opposite can happen as well (inherited money can do much good).
> Can we assume that pg is in favor of restoring the estate tax?
Can we see some evidence that the estate tax in the US ever had the characteristics cited by its proponents?
Consider, for example, the possibility that Caroline Kennedy will be appointed to the Senate by the NY Governor who just happens to come from a line of NY politicians. This isn't an isolated case - a huge fraction of current US senators are related to past senators and governors. (Biden's successor, one of his advisors, is seat-warming until his son is ready to run.) The House of Representatives isn't much different. The same thing happens at the state level - IL is mainstream in this.
The estate tax also doesn't hit the rich. The wealth of Gates et al will not be subject to the estate tax. It will, however, work to the benefit of their kids. (The Fords, Rockefellers, and GM heirs worked the same game.)
The only folks who pay estate tax are medium size family-owned biz. Many of them have to sell to pay the tax bill.
Interestingly enough, Buffet makes a lot of money buying such biz. He also makes a lot of money selling insurance to their founders to pay estate taxes.
PG, I enjoy your essays a great deal. Often while reading the content I click on your numerically referenced notes. Clicking on the numbered reference takes you to the details of that note. Please link the number in the notes section back to the location in the essay. This way I can click the note in the text and read the note and then click the note reference to return to the essay content and continue reading. This will prevent me from having to scroll back up to place where I left off. I would appreciate this user interface improvement. Thank you again for your thought provoking work.
That second to last paragraph sticks out like a sore thumb and should be removed, imho.
'left' and 'right' are pretty meaningless terms, their origins being the antithesis of American ideals. That paragraph says more about pg's political self-identification than anything else.
I think the idea is that leftist policies retard the shift from credentialing and large organizations to small, nimble organizations by being anti-business. Therefore, by default leftist policies stick us with credentialing. This is not a conscious goal of the left, it's just what happens.
I'm sure the left would disagree with this, but that's the right-wing point of view.
I can give an immediate example. The left wholeheartedly supports Teacher Unions who are fundamentally opposed to merit pay, prefering seniority and tenure.
And in Missouri at least, teachers' unions have made it so that any teacher, no matter how bad, can get a raise just by getting a master's degree or a PhD.
MiT and D.Ed. degrees are almost completely worthless (even worse than MBAs) largely as a result of such programs.
I had several teachers in the DC public schools who were considerably worse after getting graduate degrees. They weren't remotely clever to begin with, and having faddish pedagogy drilled into them didn't help at all.
I said that the left has to fall back on credentials (the wording of which implies they don't like them).
And while the right may not like estate taxes, I don't think anyone on the right (unless there are any ancien regime fans left) favors the use of that wealth to buy position. At most they'd probably concede the possibility is an unfortunate consequence of not taxing people twice.
I don't think that's true. There are certain ideologies that travel together, whose fortunes tend to wax and wane together. If you told me an ideology (say, environmentalism or unionism), then I could easily tell you whether it was on the right or left (in this case, both left).
There are exceptions in practice, but to say the labels are meaningless is silly.
Meh, I disagree. There are groups that claim to be socially liberal and fiscally conservative. There's no catch phrase for someone who tends toward slight conservatism, but is generally open to government intervention, though I imagine such people exist. That's at least two dimensions to consider, it seems to me we should at least have a plane.
Actually there were a lot of surprises. That credentials are seals between generations; the distinction between direct and indirect methods of helping your kids; that sealing off direct methods promotes indirect ones; the emptiness of the explanation schools give of legacy admissions (which now seems to me as damning evidence as the practice itself); the precise connection between credentials, measurement, and organization size; the fact that falsely impressing colleagues in a small organization wouldn't have a lasting effect, because the organization wouldn't last; the reason the word yuppie arose, and the connection between startup founders and yuppies; that the spread of measurement would bring credential granters into line as well; and that the left, because their attachment to equality (of result, not opportunity) makes them dislike markets, have to fall back on a broken method of preventing transmission of power between generations.
I think the reason I didn't sound excited was the way this
got written. I started it this summer, didn't work on it
for months, and then started again in November. I rewrote it so many times that all that's left of the first version is the beginning, up to "Chinese example," and one paragraph in the middle that used to be in the beginning.
I agreed with the point of the piece beforehand. And there was plenty of newness in it. I wanted to gauge how it was written. And the analysis is helpful.
A lecturely, authoritative voice seems inevitable as you move away from the excitement phase. But excitement may be more convincing.
And that's one advantage that blogs have over essays, that I hadn't considered. Being published so much quicker, it's easier to retain excitement. But I'm a bigger fan of essays. The medium still has steam left, even if you include it's other anachronisms.
I agree with much in the essay, but it's interesting that PG used law as an example of an organization that broke from the deferred compensation principle.
Most "elite" law firms actually do pay by seniority. An associate is generally paid according to the number of years he or she has been with the firm. And while the salaries for young lawyers is very high, they are actually (often) working for the potential of becoming partner - a reward for work done earlier in life. This is the essence of deferred compensation.
Law firms also tend to be more impressed with credentials than almost any other segment of the economy. The pecking order of law schools would surprise many people from the engineering world. Of course, some schools do have stronger reputations (for good reason), but in law, Harvard > Columbia > Cornell. Students accepted to a school higher in the chain will prefer that school by a wide margin. Whereas engineering students, especially at the grad level don't tend to have an MIT > Stanford > Berkeley kind of mentality... they'd be more interested in how their research interests match up with faculty in that area.
I think Law, Medicine, and other professions will be the holdouts against the trend away from school-based credentials. It's probably no surprise - these are the few sectors of the economy that can actually ban someone (with the full weight of government behind them) from practicing based on degrees and exams.
The question is, when will the tide overtake them (if ever)? Will the opportunities in a free market eventually become so appealing that only second-rate people will seek out the credentialed industries?
I'm not joking here. Law school is three years, but we all know that people learn at different rates. I suspect that some people could learn the material in a few months, whereas others would need six years, and some would never be able to learn it at all. The requirement of a 3 year degree most definitely sends some of the brightest folks into other fields where they can leverage their exceptional ability to learn quickly (software seems to be a favorite field for these people).
> I think Law, Medicine, and other professions will be the holdouts against the trend away from school-based credentials. It's probably no surprise - these are the few sectors of the economy that can actually ban someone (with the full weight of government behind them) from practicing based on degrees and exams.
These are also fields where the actions of a single bad/incompetent actor can have very, very serious consequences for their customers. Death and long-term incarceration are pretty heavy negative externalities and most of us do not want to wait for "reputation" to sort out the bad apples. These are also fields with large bodies of knowledge that an applicant in the field needs to master, with relatively easy mechanisms for testing this knowledge.
There is a reason your doctor has copies of his diploma's on the wall, and a reason why your eye searches them out when you are discussing your problem. Sometimes knowing that the other party has passed through several well-regarded and reputable gatekeepers is a good thing. Will the "wisdom of the crowd" ever replace the bar exam, medical boards, and graduation from an elite university? No.
I definitely don't want to give the impression that I am opposed to credentials in all instances. Actually, Medicine isn't that great an example, because you don't become licensed as a physician or surgeon until you've completed a residency - ie., you've proved you can "do it". Whereas in Law, you only need to obtain a degree and pass an exam. So in a way, you really are relying on reputation in law - which may be why "big firms" like to broadcast their connections with Ivies and so forth.
Anyway, the licensing process for MDs is so different from JDs that I shouldn't have mentioned them in the same sentence.
Here's my worry, though. If a truly bright person can learn much more quickly than the average, that person may eventually be discouraged from pursuing law (or even medicine) because they are essentially throttled by the system. Also, credentials may not matter as much to this person, because they seek out areas where they can prove themselves without worrying about the particular degree.
If this happens (and I'll admit it is unlikely in the near term), you could have a situation where professions that make heavy use of credentials are most appealing to those who are good at getting credentials, but not much good at anything else. For those folks, the payoff of a credential is highest. Whereas for someone who could do well with or without a credential, it's just a path with many irritating roadblocks.
The interesting thing will be to see if the professions adapt. Perhaps there are some creative ways to retain some of the valuable aspects of credentialing without forcing everyone to learn at the same pace? Perhaps a heavier focus on exams, rather than than hours or years logged in a program? Another approach might be to limit the most restrictive licensing to more specialized cases - ie, require a 3 year degree and full membership in the bar (and, I'd hope, even more than that) to try a death penalty case, but require a shorter and easier exam for more transactional work where a minor amount of money is on the line. Even in medicine, I suspect that nurses can handle many of the things that only MDs are currently allowed to do (in some states, nurse practitioners can do this - a two year MS followed by a one year apprenticeship and exam is sufficient to independently treat some cases, rather than a 4 year degree followed by a three year residency).
Hard to say if it can happen, or if it is even desirable. But I'd guess there will be pressure, and that's not necessarily a bad thing, depending on how it all goes down...
One important factor Paul left out was the ability to terminate bad employees. If an organization can't get rid of poor performers at reasonable cost, due to restrictive labor contracts or tenure for example, the organization will have to put much more effort into screening candidates.
This is true. I didn't go into detail about the reasons large organizations can't cheaply measure the performance of recruits, but this is the sort of thing I meant.
The usefulness of measuring depends on how actionable the measurements are. I think this is one of the big reasons that the startup culture in Europe has never really developed. In most European countries (and US government agencies) it's almost impossible to fire someone after a short probationary period. This distorts labor markets so badly that even left-leaning labor economists (e.g., Brad DeLong) have sharply criticized European labor laws.
This brings up an interesting sorting issue. If it's easy for people to be terminated and/or quit (because they know that there are other jobs, available in part because other employees have been fired or quit), there should, over time, be an improvement in the matching between employees and jobs. This should lead to higher productivity as well, not to mention greater job satisfaction.
Paul argues testing will be pushed aside by measurement of actual performance. He may be right but existing methods of testing are amazingly unsophisticated and open to improvement. For instance:
Most work in the real world is done in teams, but schools tend to grade on average team output, which is vulnerable to free riders. Why not rotate individuals through teams and measure independent contributions with some simple statistics? That's what coaches do for sports like rowing.
Knowledge in the real world is rarely all or nothing. It matters how confident you are in what you know. Will you risk your company's future on an algorithm? Why not adjust test results by student's own ratings of their confidence in their answers? Then we could see who can really be trusted with their knowledge.
i believe, without having done any research, that credentialism worldwide has GROWN massively as the middle classes try to protect their positions for the selfish reasons PG states. less biased, more thoughtful research would make for a more reasoned debate about an important topic, to Y combinator, and to wider society.
Ask someone who works at Google if there was ever a time when in the company's history when a Stanford or Berkeley degree wasn't a major advantage in the hiring process. Even at a small startup, you're probably going to have more faith in an unknown from Stanford than an unknown from Western Kentucky U.
Paul's essay is interesting, but I don't think any of recent academic / sociological studies would confirm his conclusions. In general, the best indicator of one's success is (a) your parent's income and (b) your education. Academic connections open so many more doors and opportunities, even to those who don't have the intelligence and skills to match. An idiot who graduates from an unknown college will probably go nowhere, but one from Stanford may end up as CEO. (Personally, I've seen the latter occur.)
I've definitely seen evidence that the school one goes to correlates with success, but are you sure that the causality flows in the direction: academic credential --> success?
It seems plausible at least that if the parent is wealthy, seeing to it that their kid is well-educated is a good way of signaling that their kid is decent material, it's a costly signal so only the wealthiest of parents will probably be able to afford it. And since the kids of wealthier parents tend to do better, the education signal, by some accident ends up being used a great deal to gauge future success.
I actually think that it used to just be a straightforward signal of wealth, but these days we don't want to admit that that's the case, we instead want to justify it as producing some useful economic product.
"Ask someone who works at Google if there was ever a time when in the company's history when a Stanford or Berkeley degree wasn't a major advantage in the hiring process."
All my info is based on what they said (and hence somewhat unreliable), but the factors cited in my case were my startup experience, the interview feedback, and the fact that my skillset happened to fit what they're looking for at this point in time. I also have an Amherst degree, so I can't rule it out as a factor, but nobody once mentioned it. And my GPA sucked, so it certainly couldn't have been all academic.
"... It was not so easy 25 years ago for an ambitious person to choose to be judged directly by the market. You had to go through bosses, and they were influenced by where you'd been to college ..."
What was the hack Apple, MS and other tech companies used? Aligning themselves to rapidly improving technology where greater performance was really noticeable?
I still see the barriers described here occurring particularly in market bottle-necks (interfaces where technology collides with its ancestry) where old-world business has influence. The biggest one I can think of is access to capital. It seems the closer you get to capital, the greater the effect credentials can come into play balancing out performance.
"...
What cram schools are, in effect, is leaks in a seal. The use of credentials was an attempt to seal off the direct transmission of power between generations, and cram schools represent that power finding holes in the seal. Cram schools turn wealth in one generation into credentials in the next ..."
Is YC acting as an anti-cram school hack? Using YC credentials to gain leverage to gain capital to the next generation Startups to be measured by performance?
Paul mentions small companies interacting directly with markets as the catalyst to shifting from credentials to performance.
Another potential catalyst is the growth of freelance/contract labour. Essentially, what is being created is a less secure labour environment where there is more incentive & ability to directly measure performance. The assumption is that you are fired at the end of the job. If you can get further work or a referral, it's a bonus.
There is something obscuring the effects of this though. The growth of freelance/contacting partly coincides with & partly is driven by the internationalisation of labour markets. That will obscure some aspects of it. A copywriter in a cushy job in a high wage country won't leave to pursue a $25/hr career as a elence copywriter. They stay away from competition with international labour markets.
I don't think that will go away any time soon. What would be interesting to study would be these kinds of markets in low wage countries where hiring freelancers on international labour markets is not a cost advantage to hiring local employees
In any case, if we are moving to a higher performance/credentials ratio, I think this is a big part of the equation, bigger then startups.
208 comments
[ 2.8 ms ] story [ 120 ms ] threadhttp://is.gd/c0a2
(The wikipedia link had a set of paren that killed the URL)
I'm sure that I can speak for many of us here when I say, thank God that's not the case here anymore.
People go to YC for many of the same reasons people went to Stanford -- past seed stage, it's still largely a credentials game. So you have to ask whether the same potential for "corruption" is present. Are they failing to transmit credentials over generations because it's a revolutionary model, or because it's only been around less than one generation?
What if someone did a sweetheart deal for a YC startup as a personal favor/quid-pro-quo/etc? What does it mean when founders who got rich from a YC company start funding new YC companies because they are YC companies? Is that "corruption" in the same sense?
The part you're right about is that "admission" to YC is structurally a lot like college or grad school admissions. The difference, though, is that our choices are immediately tested. College admissions officers do not, as far as I know, follow up on the careers of the people they accept and see how they did. Whereas if we pick a bad group, that fact will immediately be thrust in our face. This causes us to work very hard to try to pick the best people.
It was being in a sense a colleague of college admissions officers that made me think about such topics. We're the next bottleneck downstream from them, and it quickly became clear to us what a bad job they were doing:
http://paulgraham.com/colleges.html
It would be interesting to see numbers on that. Informally, I noticed that back when reddit was good, anyone who was asked had either heard about it because it was YC-funded, or heard about it from someone who heard about it that way. I'm not certain, but I would probably bet that the submitters of, say "Man sentenced to NINETY THREE days in jail for taking 6 seconds to sit down! Outrageous!" or "Warrant Issued For Man Paying Fine With Pennies" didn't find out about it the same way.
The only way that web marketing is "free" is if product awareness spreads between customers, or is pushed through the founders' personal contacts. Wouldn't this imply that we could expect this phenomenon - the initial user base being in some way close to the founders - to grow as the industry approaches the asymptotes you assert it is heading for?
Awareness of Reddit did spread between users.
I think it's a fairly good bet that Reddit did a lot better with their initial group of users being seeded from folks who'd heard of you or followed comp.lang.lisp, than if they tried to launch something among their UVa friends.
I'm not sure that's true - in the UK at least, universities often issue press releases when one of their alumni does something notable (like becoming the Prime Minster of Thailand for example, which Oxford is crowing about at the moment).
And I've seen in Australia a follow-up 6-12 months later to see if you got a job / further study. But I've never seen a substantial follow-up of the average population of a uni.
The column 'Alumni' compares the number of Nobel prizes or Fields Medals won by Alumni to the number won by Harvard (in percent). Caltech scores 52.8 on this, and ENS 53.4. But Caltech and ENS are by far the smallest of the universities on the top 100 list (unless I am mistaken). Both Caltech and ENS enroll about 200 undergraduates a year, with ENS enrolling only about 100 of those in science and engineering. Caltech has a population of another 1200 graduate students.
One thing to note about Caltech, ENS, and Phystech, they are extremely small, extremely selective, have no legacy admissions, limited or no affirmative action, and pay heavy attention to competitive exams, in the case of ENS and Phystech, administrating them.
That said, Cambridge is a far bigger University than Caltech and others, but a large proportion of those degrees are awarded in non-science fields.
http://nobelprize.org/nobel_prizes/lists/universities.html
Wikipedia has another list based on what universities claim themselves:
http://en.wikipedia.org/wiki/List_of_Nobel_Laureates_by_univ...
So while Caltech is doing very well indeed, there are 15 universities with more affiliated awardees.
The Wiki is the one.
In Canada, one of the strongest selling points of universities is what fraction of their students got a job in a related field within X years of graduation -- so at least here, institutions do track this sort of data.
In addition, there's a lot of work done to validate admissions processes based on the success of students during their undergraduate years; universities have extensive "black books" which tell them how a student graduating from high school X compares to a student graduating from high school Y given the same nominal high school marks (here in Canada, admissions are done primarily based on high school marks -- we don't have any equivalent to SATs).
Yes, not to the customers of the app, but the interest the YC brand generates in "insider" circles, which helps bootstrap popularity I think you are underestimating.
> YC's principal value is the advice we (and the other YC startups) offer.
Yes but as before, the "brand impact" in the blogosphere (I feel dirty writing that) is still a huge thing. If that comes under the umbrella of advice, then that makes sense.
I checked for an article from the Harvard Crimson about the degree to which the admission office follows up on admitted students after they enroll, although I didn't find the link just now. You've received other replies mentioning various ways, with various degrees of formality, that admission offices can and do check their results. Now that many admission offices are making very fine distinctions among applicants as admission rates plummet below 10 percent, there will probably be growing interest in validating the criteria on which admission decisions are made.
MIT admission office McGreggor Crowley, in one of his admission travel sessions in the last year, said his office has done studies of which students perform best in upper division courses at MIT. Taking really hard classes in high school is a better predictor of upper-division success at MIT than having a spotless G.P.A., for example. He asked a large audience of high school students how many had done baby-sitting, and very few raised their hands. He said if that is one thing that kept you busy in high school, be sure to mention that on your MIT application.
despite my Cs in english, pol sci, and other humanities courses, i got in ... to make it more dramatic, i also got graduate research assistantship :D
I did YC because it seemed interesting and I needed a good excuse to quit my job. [2] Tom was bored and wanted to get out of Ohio.
When we launched on Techcrunch, we were met with widespread apathy. Techcrunch readers, by and large, weren't our target market and didn't give two shits about a ticket site. We expected this to happen, and it was still cool to be featured on Techcrunch anyways :). We weren't creating a new platform or becoming the next facebook - we were just solving a simple problem (finding tickets).
The investors we met with didn't care that we were with YC (many didn't even know), they cared that we were making money in an expanding market.
Also, our most targeted/best traffic has been from places that didn't know/didn't care who YC was (e.g. Ticketnews.com, Thrillist). Same situation for our users.
Finally, as Paul alluded to, the YC name can only do so much for you, if you don't have a good product none of it matters.
[1] - I'm only talking about the YC brand name/credentials, not the program itself, which certainly had a positive impact on TicketStumbler.
[2] - When much of your family lives in Detroit and has been recently or is, unemployed, it's difficult to say your ditching a cushie 60k per year job to do a startup. Having "backing" makes it much easier.
As a young geek, I was often passed over because I didn't look like I could do the job. It started with little league and continued on through high school and into the work place. I even quit one of my better jobs because a slicker looking but far less talented peer was promoted into what should have been my job.
No more. As a mild mannered adult geek, I still get the same pseudo-negative first impression reaction until I open my mouth and show them how much they'd benefit from me and technology.
The Performance Economy was a long time coming (and probably taken for granted by anyone under 25), but now that we're there, there's no going back.
On a complete tangent, this is one of the perks of living in a foreign country if you speak the language well - instant recognition that you have mastered their language, and that you must therefore not be a complete moron.
The downside of this in Europe, for me, has been moving about 4 hours away (absolutely nothing in the US, in terms of distance) and completely losing this advantage.
It's good to still have a bit of an accent. I knew an American who had mastered Russian to the extent that he was indistinguishable from a native speaker. There were, however, glitches that would show up in his speech - an occasional wrong phrase or something. Because he spoke perfectly otherwise, Russians sometimes thought he was a dumb Russian. This made me realize that you want people to know you're a foreigner, just a foreigner who speaks amazing X.
I never spoke Russian as well as this fellow did (and most of it is now sitting freeze-dried in my brain somewhere), but there was one funny thing in my case. Much of what I learned came from reading, so sometimes I would speak Russian like a 19th century novel. This made my friends laugh, but they also liked it - Russians are big on literacy.
Edit: your core point is quite right, and is easy to verify by the impression it makes when you meet a non-native speaker who speaks excellent English.
But I'll agree with David that being an American that speaks the local language can be quite a novelty in Europe. I'm in the same boat. :-)
Absolutely. I'm a partner in a small educational venture in Taiwan, and my sales conversion rate is actually significantly better than that of our local staff. I don't believe for a second that it's because I'm that good. Having people shocked at my Chinese skills definitely gives me an extra oomph.
I'm not sure if Chinese people in the US would get the same credit.
I think the problem there is that you have no idea what background the person has. There are plenty of ethnically Chinese Americans who of course speak English perfectly, whereas if you're obviously "not from around there", people are going to connect the dots faster.
I have a small number of customers -- somewhere between 10% and 20% -- who are using tarsnap based on a recommendation from another tarsnap user; but the vast majority of tarsnap users signed up based on some other factor which is at best predictive of the quality of tarsnap. Reasons I've heard include:
There are a great many instances of companies succeeding with inferior products as a result of having superior marketing; while I am fortunate that people are interested in using tarsnap for reasons which are quite unrelated to the quality of the tarsnap code or service, the fact remains that the market isn't (in PG's words) "[taking] every organization and [keeping] just the good ones" -- rather, the market takes every organization and keeps just the ones which rank best on a nebulous combination of quality, marketing, and luck.Yes, credentials are an imperfect way of predicting quality; but are they any worse than the marketing and luck which the market introduces?
That's the mistake. The number of users you get long term depends more on the "interest rate" than the size of the initial seed. E.g. Google.
There's also a more dangerous factor to consider: The number of users of a startup is far more visible than its growth rate, so it's entirely possible that great startups end up shutting down due to a disappointing apparent lack of traction when -- if they had stayed around for longer -- their inherent advantages would have produced enough growth to make them wildly successful. Even if we suppose that the ultimate potential of a business is entirely dependent on the inherent quality of the product or service it produces -- which would be rather stretching credulity -- then the external factors of marketing and luck still have a significant impact on whether companies remain in existence long enough to reach their potential.
And while growth rate is not in itself visible, what drives it is something that's usually even more visible than the size of the user base: the quality of the product.
I'm not convinced. Most startup founders think that their products are the greatest thing since sliced bread, but they're often also aware that they are biassed. How many startup founders haven't at some point been disappointed by the size of their user base and said to themselves "gee, I think what I'm doing is really cool, but maybe the market for this isn't as big as I thought it was"?
Maybe we can call this resume-viral: products that are passed around by virtue of become recognizable bullet items on resumes.
- why do they survive?
- if they survive despite all, why do we permit them?
If the market was regulated to set a cap on the size of a single company's workforce (or some other metric, perhaps market capitalisation), we'd also avoid the 'too big to fail' problem we have at the moment.
Not to mention quasi-monopolistic practices like large chains killing off small local competitor with an externally-funded price war (then raising prices once the competition disappears).
Is it that we need the largest companies for some large-scale activities (chip fabs, aircraft manafacture), or that they can be efficient at scale (fast food, wal-mart)?
Would the world/market be a better place if companies were forced to fission at a certain size?
http://en.wikipedia.org/wiki/The_Nature_of_the_Firm
I would really like to see more analysis of what structural differences point towards the lots of smaller firms that pg sees, and whether this is a short blip in time, or a long term trend.
Also, I hate to dredge up that ugly topic "politics", but currently, in the US, if you or your firm fail, that's great in terms of opening up space for others to try, but a little scarier in personal terms, with regards to things like health care. Long term, who knows how things like that will play out.
I don't think either of the metrics you mention would solve the "too big to fail" phenomenon. The fact is that "too big to fail" is a misnomer, which has actually been used to represent the concept "too connected to fail".
I.e. it is not Bear's/AIG's pure size that prevents us from allowing them to fail, it's the connectedness that they have in our financial system, which depends on a web of millions of transactions per day between every large bank. Then, if a big bank were allowed to fail (not that I'm endorsing this theory), that would cause a cascade effect where banks started charging higher rates for those millions of transactions, which would cause less of them to happen, which would cause even higher rates, etc. etc., ending in a completely broken financial system where each bank wanted too much money for each of the many transactions they depend on to exist in their current form.
So, I don't think that limiting the size of companies in the financial sector would prevent the "too big to fail" phenomenon by itself. Which doesn't mean it's not a good idea; just that I think that particular example doesn't suport the idea that well.
(edit: and read Coase, like davidw's comment suggests :)
show me real examples outside of the web industry that this theory is even remotely correct
Any industry where the capital costs are low, or where a small group of niche specialists can add a significant multiplier to the value chain are candidates for this sort of effect. As more industrial tasks are farmed out to large companies that specialize in flexible production it is possible that you will see this show up in industries that are currently have strong vertical integration.
For example, ten people can't mass-produce a car. Ten people can provide spcialized bodywork and personalization of a mass-produced car. Ten people could probably design and oversee the outsourced manufacturing of discrete components in the car. Ten people could probably manage a company that deals with other small firms to create bespoke vehicles based upon mass-produced frames and engines. As you increase the specialization and make more parts interchangeable it is possible to decentralize and distribute a production chain to the point where a large collective of ten person companies can provide a higher-value good at a lower per-unit cost than a single large company.
Certain activities will always need large, high-capitalization organizations, but it seems that more and more of these companies and factories are being forced to become more flexible and nimble which is leading to a point where a larger and larger part of your world can be serviced by smaller teams than by large enterprises.
How do these things change with time, though? Do they always, invariably drop? Might computing conceivably go through another phase where things get consolidated into some kind of big servers that cost a ton of money? Doesn't sound likely to me, but I'm not very good at predicting the future. Actual research into these trends would be interesting to see.
no, that thins margins even more and drives to even faster consolidation. i can make PCs in my garage, yet there are only three real players in this market - dell, hp, apple. the lower the barrier to entry, inevitably you will erode margins to the point that only a large player can stay solvent.
Retail: just went through a mass-consolidation. The more dynamic firms, the "places to be" if you wanted to do something interesting in retail, were the ones doing the consolidating. Think Wal-Mart, Home-Depot, Walgreens, Best-Buy etc... While it may not be interesting to the crowd here, innovation in things like supply-chain has been stunning.
Consumer Goods: P&G has been unstoppable, and that's where most people in that business want to go because that's where the interesting things were happening.
Consumer Electronics: Dell, Apple, Sony etc... have crushed everyone.
Banking: Ditto. Where did all the ambitious young financial engineers want to go? Easy: Goldman-Sachs. Not that innovation here was a good thing... but that's another story.
Agriculture: ADM, Monsanto and the agri-business giants have almost completely taken over. Small farms are near-extinct.
This list could go on for a while, but I'll stop. I think you see a similar winner-take-all pattern playing out in most industries. It'd be interesting to see the demographics of all this: the percentages of people employed by institutions of different sizes.
Perhaps pg is right that it's starting to happen though. Maybe the web world is at the vanguard of what will be a broad counter-trend. I suspect that won't be as strong a force as he thinks though - the advantages of being big are still too great in too many fields.
Then again, maybe the mass-consolidation was artificial. Maybe all the M&As owe as much to artificially cheap money as to natural advantages of scale. I guess we'll see.
Increasing your score by less than half a standard deviation was nice, but it rarely made or broke your college application.
It was fairly scary how many of my friends had 1590s or 1600s, though.
http://talk.collegeconfidential.com/college-admissions/41382...
but on the other hand, any one applicant with perfect SAT scores can still be rejected by plenty of colleges.
http://talk.collegeconfidential.com/parents-forum/377882-how...
Agreed with the point that on a dollar-per-SAT-point basis, there is little rationale for choosing commercial SAT preparation over simply self-studying with the released previous exams. The most decisive thing to do to raise SAT scores is to read widely and avidly in English.
I am studying development policies and growth economics this year. His statement really doesn't count for the countries who have very little wealth to begin with. In fact, he needs to clarify that it is better for wealth to be passed through families, than for that wealth to be destroyed.
However, as people become aware of the ranking mechanism, they try to get ahead by overoptimizing for the narrow areas inspected. So pages do linkfarms and black-hat SEO, students goto cram schools, etc. etc. essentially producing the exact opposite of what was initially sought. I believe Joel spolsky had a good article on a simillar pattern in sales.
The question is if an algorithm that affects its input data can be made so that it adapts to the changes and gradually reaches some kind of equilibrium with the input data or stabilizes in some other way. As paul suggests, one way would be to make the system more transparent, or even 'social' such that the participants have a way to update it in an agreed way. This would keep the minority from cheating on the majority.
Doing away with such filters though might be a bit harder than we think, especially given the breadth of their applications. As aristus commented, even YCombinator has an examination system. In fact it wouldnt be too hard to imagine cram schools for YC given enough incentive. I fear this is a problem we have to either solve or live with.
Aren't they all? :-)
Say you were given 19 yr olds at the end of Year 1 of Uni. You have 2-3 years part time. How would you get them in? Activities? Interview coaching? Resume building? Portfolio building? Specialised learning?
I would also mine all the successes, details of interview process and characteristics of successful applicants and their applications for commonalities and then see how many of these can be easily added to an application for 'bonus points'. Of course it depends on how black-hat you want to get about this.
The first thing that came to my mind though, having read pg's essays, would be 'teach them LISP' :)
I don't know much about YC or what influences them apart from fragments from pg's essays.
Insisting that LISP or some other language no one uses is absolutely the best tool seems like it might be a bonus point. But I think that is as a bonus. What would be the core? They say smartness & an attitude that they are going to do something big with or without the program. The latter can be gamed. The former, I don't know much about the criteria.
...then a cram school for YC would be a program much like YC in which younger teams built smaller, more easily guided technology start-ups.
Big if. I'm sure that any self respecting University will tell you that the best prep for their University is anything that makes the candidates smarter, more driven & more likely to succeed. Everyone uses indicators. Some are better then others. Maybe some can never be gamed (I doubt it) but as actually mentioned in this essay, the way around credentials is making entry less important.
But on that note, isn't the biggest criterion for university transfer admission is who admitted you in the first place and your consequential success at that school?
It stands to reason if one wanted to predict if a group would be successful creating a successful startup that a good predictor would be if that group had stated a successful startup in the past.
I'd assume there are other criteria, like: is what this group doing within our scope of expertise? Can we add value to this startup? Will admitting this group best add to the quality of this YC class? Will we make money on this investment?
I'd also try to get them jobs at existing YC startups. Then they'd both learn how startups work, and (if they were good) get recommendations from people we know.
my attitude is that i do not care where you went to school. in an interview, if you ace the questions, why do i care if you went to ball state? so what, maybe that is all you could afford. on the other hand, if you can't answer my questions, why do i care if you went to stanford?
furthermore its likely that most people here do not know what the best colleges in korea, japan, china and india really are. so best to just skip the education section of the resume entirely
Don't confuse 'intelligence' and 'education'. There are plenty of smart people who have never been to University. Likewise there are plenty of people who go to University who graduate but can't think beyond what they have been taught.
Although one could make the case that a lot of those schools, especially the ivies, are picking people based on their ability to lead and 'make a difference' in some general sense rather than pure raw ability. Therefore they may intentionally focus on apparently manifested soft qualifications like achievement in extracurriculars, personal discipline as exhibited by a flawless GPA, or well-established personal connections.
This is all strictly re: undergrads of course. I've never seen a single retard get admitted to a top-tier PhD program in science or engineering.
That's not the case as I've witnessed it. I won't directly name names, but I am thinking about a school that is ranked number 4 for graduate engineering on U.S. News.
http://online.wsj.com/article/SB122644964013219173.html
Luckily technology is fairly cheap, learning how to do stuff is free and only requires commitment and time so we can still turn our ideas into products and put it in front of millions of people without bloggers, angels, vcs.....
Can we assume that pg is in favor of restoring the estate tax?
That's why it's dangerous :)
Buffet has it right: leave them enough so that they can do anything, but not enough so that they can do nothing.
I.e. Paris Hilton has the money that she uses for hedonistic purposes, but she won't be put into a position of power.
In formulating my reply, I think I came around to the other point of view. Bad things can happen, but inherited money is probably a minor enabler in this, and just the opposite can happen as well (inherited money can do much good).
Can we see some evidence that the estate tax in the US ever had the characteristics cited by its proponents?
Consider, for example, the possibility that Caroline Kennedy will be appointed to the Senate by the NY Governor who just happens to come from a line of NY politicians. This isn't an isolated case - a huge fraction of current US senators are related to past senators and governors. (Biden's successor, one of his advisors, is seat-warming until his son is ready to run.) The House of Representatives isn't much different. The same thing happens at the state level - IL is mainstream in this.
The estate tax also doesn't hit the rich. The wealth of Gates et al will not be subject to the estate tax. It will, however, work to the benefit of their kids. (The Fords, Rockefellers, and GM heirs worked the same game.)
The only folks who pay estate tax are medium size family-owned biz. Many of them have to sell to pay the tax bill.
Interestingly enough, Buffet makes a lot of money buying such biz. He also makes a lot of money selling insurance to their founders to pay estate taxes.
Any company that respects a traditional resume is a company I don't want to work for.
These aren't any left or right I recognize. Who is PG thinking of?
'left' and 'right' are pretty meaningless terms, their origins being the antithesis of American ideals. That paragraph says more about pg's political self-identification than anything else.
I'm sure the left would disagree with this, but that's the right-wing point of view.
I had several teachers in the DC public schools who were considerably worse after getting graduate degrees. They weren't remotely clever to begin with, and having faddish pedagogy drilled into them didn't help at all.
I said that the left has to fall back on credentials (the wording of which implies they don't like them).
And while the right may not like estate taxes, I don't think anyone on the right (unless there are any ancien regime fans left) favors the use of that wealth to buy position. At most they'd probably concede the possibility is an unfortunate consequence of not taxing people twice.
There are exceptions in practice, but to say the labels are meaningless is silly.
Here's one, Net Neutrality, what is that?
It felt more like an attempt to convince us, rather than a revelation to yourself. Not that that's bad, but you didn't seem excited.
What were your diffs?
I think the reason I didn't sound excited was the way this got written. I started it this summer, didn't work on it for months, and then started again in November. I rewrote it so many times that all that's left of the first version is the beginning, up to "Chinese example," and one paragraph in the middle that used to be in the beginning.
A lecturely, authoritative voice seems inevitable as you move away from the excitement phase. But excitement may be more convincing.
And that's one advantage that blogs have over essays, that I hadn't considered. Being published so much quicker, it's easier to retain excitement. But I'm a bigger fan of essays. The medium still has steam left, even if you include it's other anachronisms.
Most "elite" law firms actually do pay by seniority. An associate is generally paid according to the number of years he or she has been with the firm. And while the salaries for young lawyers is very high, they are actually (often) working for the potential of becoming partner - a reward for work done earlier in life. This is the essence of deferred compensation.
Law firms also tend to be more impressed with credentials than almost any other segment of the economy. The pecking order of law schools would surprise many people from the engineering world. Of course, some schools do have stronger reputations (for good reason), but in law, Harvard > Columbia > Cornell. Students accepted to a school higher in the chain will prefer that school by a wide margin. Whereas engineering students, especially at the grad level don't tend to have an MIT > Stanford > Berkeley kind of mentality... they'd be more interested in how their research interests match up with faculty in that area.
I think Law, Medicine, and other professions will be the holdouts against the trend away from school-based credentials. It's probably no surprise - these are the few sectors of the economy that can actually ban someone (with the full weight of government behind them) from practicing based on degrees and exams.
The question is, when will the tide overtake them (if ever)? Will the opportunities in a free market eventually become so appealing that only second-rate people will seek out the credentialed industries?
I'm not joking here. Law school is three years, but we all know that people learn at different rates. I suspect that some people could learn the material in a few months, whereas others would need six years, and some would never be able to learn it at all. The requirement of a 3 year degree most definitely sends some of the brightest folks into other fields where they can leverage their exceptional ability to learn quickly (software seems to be a favorite field for these people).
These are also fields where the actions of a single bad/incompetent actor can have very, very serious consequences for their customers. Death and long-term incarceration are pretty heavy negative externalities and most of us do not want to wait for "reputation" to sort out the bad apples. These are also fields with large bodies of knowledge that an applicant in the field needs to master, with relatively easy mechanisms for testing this knowledge.
There is a reason your doctor has copies of his diploma's on the wall, and a reason why your eye searches them out when you are discussing your problem. Sometimes knowing that the other party has passed through several well-regarded and reputable gatekeepers is a good thing. Will the "wisdom of the crowd" ever replace the bar exam, medical boards, and graduation from an elite university? No.
Anyway, the licensing process for MDs is so different from JDs that I shouldn't have mentioned them in the same sentence.
Here's my worry, though. If a truly bright person can learn much more quickly than the average, that person may eventually be discouraged from pursuing law (or even medicine) because they are essentially throttled by the system. Also, credentials may not matter as much to this person, because they seek out areas where they can prove themselves without worrying about the particular degree.
If this happens (and I'll admit it is unlikely in the near term), you could have a situation where professions that make heavy use of credentials are most appealing to those who are good at getting credentials, but not much good at anything else. For those folks, the payoff of a credential is highest. Whereas for someone who could do well with or without a credential, it's just a path with many irritating roadblocks.
The interesting thing will be to see if the professions adapt. Perhaps there are some creative ways to retain some of the valuable aspects of credentialing without forcing everyone to learn at the same pace? Perhaps a heavier focus on exams, rather than than hours or years logged in a program? Another approach might be to limit the most restrictive licensing to more specialized cases - ie, require a 3 year degree and full membership in the bar (and, I'd hope, even more than that) to try a death penalty case, but require a shorter and easier exam for more transactional work where a minor amount of money is on the line. Even in medicine, I suspect that nurses can handle many of the things that only MDs are currently allowed to do (in some states, nurse practitioners can do this - a two year MS followed by a one year apprenticeship and exam is sufficient to independently treat some cases, rather than a 4 year degree followed by a three year residency).
Hard to say if it can happen, or if it is even desirable. But I'd guess there will be pressure, and that's not necessarily a bad thing, depending on how it all goes down...
This brings up an interesting sorting issue. If it's easy for people to be terminated and/or quit (because they know that there are other jobs, available in part because other employees have been fired or quit), there should, over time, be an improvement in the matching between employees and jobs. This should lead to higher productivity as well, not to mention greater job satisfaction.
Most work in the real world is done in teams, but schools tend to grade on average team output, which is vulnerable to free riders. Why not rotate individuals through teams and measure independent contributions with some simple statistics? That's what coaches do for sports like rowing.
Knowledge in the real world is rarely all or nothing. It matters how confident you are in what you know. Will you risk your company's future on an algorithm? Why not adjust test results by student's own ratings of their confidence in their answers? Then we could see who can really be trusted with their knowledge.
Paul's essay is interesting, but I don't think any of recent academic / sociological studies would confirm his conclusions. In general, the best indicator of one's success is (a) your parent's income and (b) your education. Academic connections open so many more doors and opportunities, even to those who don't have the intelligence and skills to match. An idiot who graduates from an unknown college will probably go nowhere, but one from Stanford may end up as CEO. (Personally, I've seen the latter occur.)
I actually think that it used to just be a straightforward signal of wealth, but these days we don't want to admit that that's the case, we instead want to justify it as producing some useful economic product.
All my info is based on what they said (and hence somewhat unreliable), but the factors cited in my case were my startup experience, the interview feedback, and the fact that my skillset happened to fit what they're looking for at this point in time. I also have an Amherst degree, so I can't rule it out as a factor, but nobody once mentioned it. And my GPA sucked, so it certainly couldn't have been all academic.
What was the hack Apple, MS and other tech companies used? Aligning themselves to rapidly improving technology where greater performance was really noticeable?
I still see the barriers described here occurring particularly in market bottle-necks (interfaces where technology collides with its ancestry) where old-world business has influence. The biggest one I can think of is access to capital. It seems the closer you get to capital, the greater the effect credentials can come into play balancing out performance.
"... What cram schools are, in effect, is leaks in a seal. The use of credentials was an attempt to seal off the direct transmission of power between generations, and cram schools represent that power finding holes in the seal. Cram schools turn wealth in one generation into credentials in the next ..."
Is YC acting as an anti-cram school hack? Using YC credentials to gain leverage to gain capital to the next generation Startups to be measured by performance?
Isn't a startup itself an "economic cram?"
"Cramming" isn't necessarily "bad" when the most effective preparation is achieving thing itself.
Another potential catalyst is the growth of freelance/contract labour. Essentially, what is being created is a less secure labour environment where there is more incentive & ability to directly measure performance. The assumption is that you are fired at the end of the job. If you can get further work or a referral, it's a bonus.
There is something obscuring the effects of this though. The growth of freelance/contacting partly coincides with & partly is driven by the internationalisation of labour markets. That will obscure some aspects of it. A copywriter in a cushy job in a high wage country won't leave to pursue a $25/hr career as a elence copywriter. They stay away from competition with international labour markets.
I don't think that will go away any time soon. What would be interesting to study would be these kinds of markets in low wage countries where hiring freelancers on international labour markets is not a cost advantage to hiring local employees
In any case, if we are moving to a higher performance/credentials ratio, I think this is a big part of the equation, bigger then startups.