Hacker News new | past | comments | ask | show | jobs | submit login

If you're looking for The Single Greatest Predictor of Future Stock Market Returns[1], here it is: http://www.philosophicaleconomics.com/2013/12/the-single-gre...

This is a long read, but it's worth it. The metric can be calculated in FRED[2], and as a predictor of future returns, it outperforms all of the most common stock market valuation metrics, including cyclically-adjusted price-earnings (CAPE) ratio[3]. (Basically, the average investor portfolio allocation to equities versus bonds and cash is inversely correlated with future returns over the long-term. This works better than pure valuation models because it accounts for supply and demand dynamics.)

[1]: http://www.philosophicaleconomics.com/2013/12/the-single-gre...

[2]: http://research.stlouisfed.org/fred2/graph/?g=qis

[3]: http://www.multpl.com/shiller-pe/




Wow, really interesting read. Thanks for the link.

Only trouble is that once people find patterns like this, they have a habit of disappearing. Hopefully this one is based on solid enough fundamental market forces that it persists after its publication. It was published in 2013 so we won't know for sure until after 2023.


I actually made an automatically updating chart for this using FRED data: http://financial-charts.effingapp.com

TLDR: The correlation did go down a bit since publishing but still seems alright.


This is awesome. Would it be possible for you to extend the spx 10 year return line with a greyed out or dotted line from 2007 onwards that shows the running annualized return so far? So for example for 2012 it would show the spx annualized return from 2012 to 2017.


Thanks for this. Can you also add a way to change the window for S&P returns from 10 years to other time windows? It'd be interesting to see how the plot changes with adjustments to the window.


I remember looking at other time windows, anything 5 years or below wasn't great.


this is great! you should add shaded areas for official recession periods, it's pretty easy to find:

https://fred.stlouisfed.org/series/JHDUSRGDPBR


I saw lot's of suggestions above, how about making it on github so anyone can hack it?


Just found its repository here: https://github.com/effinggames/financial-charts


> Only trouble is that once people find patterns like this, they have a habit of disappearing.

It took me a few seconds to figure out that you meant the patterns and not the people as the targets for disappearing.


> Hopefully this one is based on solid enough fundamental market forces that it persists after its publication.

If you can find any way to predict the future price of things, you can make money by performing arbitrage across time (instead of space, which is how people usually think of arbitrage). This (nominally) describes all types of model-informed time-based investment.

The thing with arbitrage is that there's a finite amount of money you can make. If there's a price difference on some cross-listed stock between HKEx and SZSE, you can make money by buying on the cheap one and selling on the pricier one. But, as a side effect of doing so, the price goes up on the cheap one and goes down on the pricier one. There is only a finite amount of stock you can exchange before the price difference approaches zero and you can't make money. The price information has been communicated, and the market has served its function.

The same is true for arbitrage across time. There's only so much money you can spend before the "prices" (more complicated than spacial price differences, because you have to worry about some more complex utility theory to find the expectation value of something in the future) equilibrate and market information has been propagated "through time".

This is why these things "have a habit of disappearing"; every single way of making money via models corresponds to a market inefficiency. Model-based traders are eliminating market inefficiencies and making a cut off the benefit, just like any other good business. The inefficiencies are just more abstract than usual.

So if a model is good, there's no way for it to persist. If it actually predicts something we didn't expect, that corresponds to some inneficiency in the market that some trader can eliminate in exchange for a nice fee.


Which is all just another way of saying (albeit more interestingly and in more detail), that when new information becomes available, the market adjusts to better reflect the true value of things.


This is true in a theoretical free market where there is price discovery. In the "markets" that actually exist, though, we don't have true price discovery. We have a market with centrally controlled interest rates, central bank purchasing (in a variety of ways) and other institution-based interference that create inefficiencies that are not organic, and are therefore not self-correcting.


I don't think that changes the previous statement. It just changes the value of truth. Maybe not self correcting, but certainly self regulating towards the new normal as set out by regulatory constraints.


None of the stuff you mentioned invalidates any of the mechanisms I described, although regulations may decrease their effectiveness (as with any other price-discovery mechanism) by introducing friction or breaking efficient instruments.


aka "policy". Also self correcting, but with a longer time scale.


Depends on what your modeling. A risk vs inflation vs ROI curve should be maintained by the market rather than destroyed by it.


This should be baked into your calculation of future expectation values. This factors in risk, uncertainty, time preference, inflation, etc.


You should read the author's next post where he pokes a bit of fun at his metric:

http://www.philosophicaleconomics.com/2013/12/valuation-and-...


self destructing prophecy :)


The 10-year window is kind of arbitrary, and knowledge of this pricing model won't erase the pattern, but may change the window over which it is effective to something much shorter as timescale compresses.


> "... once people find patterns like this, they have a habit of disappearing."

Hah! Grammatically, "they" refers to the people (not the pattern).

IMHO, sometimes these details matter, as in:

"Let's eat grandma!" vs "Let's eat, grandma!"

or

"know your shit" vs "know you're shit"

:)


what makes his sentence not an amphiboly? What's the grammatical rule that fixes the "they" to apply to the people not the pattern?


Thanks jxramos, good questions. FTR I had to look up "amphiboly":

  > "Linguistically, an amphiboly is an ambiguity which results from ambiguous grammar, as opposed to one that results from the ambiguity of words or phrases—that is, Equivocation."
Your reply - and a handful of rare / unexpected downvotes led me to re-read my comment, and in retrospect I realize I did a poor job communicating. I intended simply to convey my amusement at the ambiguity. I think for all intents and purposes his sentence was an amphiboly.

"Fixing" it would probably involve rephrasing to reverse the order of "people" and "pattern" in the sentence or just put all the emphasis on the pattern:

"... these patterns, when discovered, have a habit of disappearing"

Strict correctness regarding subject:verb agreement can, like other grammatical constructs, lead to non-idiomatic, awkward phrasing. Formal "rules" about prepositions are the first related example I can think of. ("... of which I can think" is awful, right?). My point was just to laugh at the ambiguity. (shrug)


PS The sinister "people are disappearing" interpretation was the 1st way I read it.


One thing to keep in mind is that predicting a downturn in the short-term is different than predicting long-run performance over 10, 20 or 30 years. I think the website is trying to do the former and your article is talking about the latter.


That "predictor" or monetary stock and flow theory that blog post describes is well over 100 years old now and very well known (pre-Keynesian and going back to at least Irving Fisher). It has also been studied in detail in academia for decades if anyone wants to look up proper research studies about it. The blog author maybe should have mentioned that this is already a well known and tested predictor, and if the author hadn't heard of before writing that blog post I would question that entire site.

There are also several companies that specialize in providing this exact cash/equity data research to financial firms to help them manage their trading strategies and offering more practical details in their data than FRED data offers. For example TrimTabs [1] (no affiliation) has been around since 1990 according to their website [1] trimtabs.com


After being receptive to this article at the outset, I've become much more skeptical after thinking about it some more, to wit:

1. Why 10 years as the forward return period -- would be nice to see the situation for other return periods, even close to 10 years. Granted, it is a round number and doesn't appear to be cherry-picked, but still.

2. Why 1952 as the starting point -- presumably because the data from FRED starts there, but can this not be extended further back?

3. Why an implausibly linear relationship between the equity allocation percentage, which is bounded (0%-100%) vs. the returns, which is unbounded on either side, over such a large range of values? What is the mechanism to explain this? In other words, what is the proposed model that transmits allocation percentage into a return, even if the directionality of correlation is at least likely? What happens around 0% and 100% allocation -- do the extreme cases make sense?

I'd like to run my own data before I believe this.


The idea definitely makes sense, but what is the use of this graph? It only shows past returns based on information we already know. Far more interesting is the realization, which is not news for people who observe these things, that prices cannot go down until there is a crunch on the money supply. After all, the allocation of stocks usually remains constant or goes up when prices are going up, so only money supply can change the dynamic of price escalation. The same happens on the way down, because prices are decreasing, the allocation naturally reduces, even if investors want to keep it constant. Only new money supply can reverse the trend of equity price declines.


In the same blog as your reference [1] you can find one of the more thought provoking essays I have read in the last little while. It's a discussion about how easier stock market diversification should lead to permanently higher multiples on earnings.

http://www.philosophicaleconomics.com/2017/04/diversificatio...

Highly recommended.


"Unfortunately, when we flip the point around, and say that the universe of risk assets should grow more expensive in response to improvements, people get concerned, even though the exact same thing is being said."

Discredited the article to me..

Improvements of a market reduce its friction and make it more efficient: transaction costs disappear for instance. Cost and price aren't the same thing. The latter includes the former.

Price can be seen as formed by a supply-demand process. In this process investors and capital seekers are BOTH subject to the cost of an inefficient market. They're not two sides of a coin.

Maybe not so recommended.


Nothing he says is contradictory to the point you are making here. As markets grow more efficient it become easier for buyers to enter the market which increases demand which causes asset prices to rise.


You are right in saying he doesn't contradict me as I am giving precision and I am trying to contradict him. :-)

Basically you say that the decrease in friction in a market drives the demand up: as markets grow efficient it become easier for buyers to enter. However I see that it become easier for seller to enter as well! So it is not necessarily a driver of demand..

This argument can be correct in an asymmetrical market like the ones for bonds, equities, real estate, commodities but it doesn't hold in FX markets for instance (perhaps the money market as well) or other derivative markets where there is not so much bias in being a buyer rather than a seller.

Although, I might say this argument can be available in a world were investors are in majority of the 'buy and hold' type. Which is the world we live in actually! So I can agree with him that new market opportunities and improvements in market efficiency can be an explanation for the flow of cash, and thus inflated prices... even bubble genesis! The venture capital markets or the crypto-currency ones are perfect examples.

To conclude you made me think twice about that and I kind of agree with him but I still don't think it is a universal argument.


Thanks, it would be interesting to see how this has trended throughout this year. The FRED chart is only current to January 1 2017, despite the terms being updated quarterly.


I personally don't think the past 100 years are going to look anything like the next 100 as far as the stock market is concerned. But we will see.


Pure astrology


Not nearly as bad as some of the "Technical Analysis" I've seen. From what I gather, "Technical Analysis" just amounts to this is what the stock has done in the past, maybe it will do that in the future?


Yes and no. It depends what type of technical analysis you are doing... If your technical analysis has no other inputs except the price(over any period(s) of time), then you may as well go buy a lottery ticket.

If you are wondering which category you are in, if at any point during your "analysis" you find yourself looking for things with names like "evening star", "bullish engulfing", "head and shoulders", etc.... save yourself some time and money, and go buy that lottery ticket. Day trading is not for you.

If your technical analysis is looking at volume/price/depth-changes, then congratulations, you are on the right track. Next you have to overcome the largest and most devastating hurdle, which is your own psychology.


Could you suggest books that go deeper in the second type of technical analysis? Most of what I find online fits well in the first type you described.


The best resource I know is John Grady's "No BS Day Trading" book. It's part of his basic course at http://www.nobsdaytrading.com/courses/basic-course/

Reading that and applying it was a huge turning point for me in my trading.


Couldn't this strategy be easily automated, and therefore, probably has been competed away by now?


The more this strategy is automated, the more effective it becomes. This is a simple explanation for "trends", as people see liquidity consumption and enter on the side to further consume it - more people repeat the pattern. The net result is a "trend", which is a blunt way of saying "demand exceeded supply in <x> direction".

There is no sane way to automate this, since everyone's description of "enough" volume delta is different. Perhaps I open positions at a CD imbalance of 15%, bank of america waits on 19%, and chase at 25%. My actions at 15% further the delta, causing bank of america's 19% threshold to fire, which in turn increases the delta to 25% causing chase's threshold to fire. Until every day trader "gets on the same page" so to speak, this will not be automated. And...if they ever do get on the same page, then it only takes 1 person with decent equity to take the other automated strategies to the cleaners.

tldr; This will be automated away when greed no longer exists, i.e. never.


This is most adaptive algorithms work, though. As your scenario gets played out over and over, the high threshold people will notice the decline in profitability and look for ways of improving the algorithm. They'll look at the various parameters and notice that the lower the volume delta the higher the return, so the volume delta parameter will come down. As everyone's volume delta comes down, the strategy will work less and less until it roughly equals the discount rate.

Anything that can be completely automated will lead to an elimination of excess returns. Having a single parameter that differs between market participants is not enough to stabilize a long-term imbalance.


I appreciate the comment, but it just doesn't work out that way. In theory yes, what you are saying holds up. In practice(I'm running these 24/7 and generating my income solely from them) this has not happened yet, and is showing zero progress towards happening.


Can anybody explain the fundamental reason behind "resistance/support lines" in technical analysis?

I'm guessing it is a self-fulfilling prophecy (everybody believes in it, so it becomes true), but perhaps there is a logical explanation that I'm totally missing.


Quoting from Bruce Kamich's How Technical Analysis Works:

Support and resistance areas form because market participants remember price levels, and they tend to react as a group when a stock returns to those prices. A simple example will make the concept easy to understand. Let's imagine that you and other investors bought a stock at its initial public offering price of $20. People who did not buy the stock at the offering price are interested in buying it at that level, and they buy the stock at $20 or perhaps above $20 as interest in the stock builds. Other buyers come in and the stock trades up to $22.

The stock may trade back and forth between $20 and $22, but let's imagine that at some point the stock slips down to $16. Some traders will hold on to the stock, hoping that its fortunes will reverse and it will trade back above $22. But if the stock remains depressed, owners of the stock who bought it in the $20 to $22 area will begin to think it would be nice just to break even. If the stock does trade back up, the desire to get out at break-even will become even stronger as the stock approaches the $20 to $22 area.

To put this price zone into perspective, the more sideways trading that occurs, the greater the supply of stock will be. There will be more people who want to "get even," and therefore there will be more resistance to the stock's advance.

Support and resistance are not precise concepts. When support develops during a decline at a price short of the exact level, it is usually because traders are anxious. They remember the prior resistance level at $22, but they start buying on the way down at $22.75 and $22.50 because they are anxious or even fearful they will not have the opportunity to buy again at $22.

On a few occasions, eager traders might push a market too quickly through a support level, and support might not develop until just beyond the support area. Whether support is found short of the expected level or just beyond the level will tell you how eager or fearful traders are.

We have seen that when support is broken on the downside, it then becomes resistance. The opposite is also true; resistance, once broken, becomes support. This reversal of roles is due to the memory of traders and investors who want to get out of their losing trades at break-even, and traders who want to increase winning positions by buying more stock at or near support.

The role reversal of support and resistance leads to the formation of trends, because in an uptrend, market pullbacks or reactions will tend to find support at the last resistance level. In a downtrend, reactions or rallies will tend to find resistance at the last support level.

All support and resistance levels are not equal. The strength of a support or resistance level depends on several factors, such as the number of times the level or area was tested, the volume of trades transacted there, how long ago the formation appeared, and whether it was a round number of a "big figure." Even knowing the type of security will help in determining the validity of the support or resistance area.

The more times a level or zone of prices is tested, the more important that level becomes. A level that is tested six times and holds tends to be more important that one that was tested only twice. The more times a support or resistance level is tested, the more traders will remember it and the more traders will be likely to be committed near it.

When a large volume of trading occurs in a support or resistance area, that adds to its validity because a greater number of traders and investors will remember the level, so their commitment to the level will likely be greater.

The further back in time the support or resistance area was formed, the dimmer the memory, and the more likely that people have already acted on new information and may not respond in the same way again. They moved on by liquidating their positions. An area of support or resistance that was formed recently tends to attract a greater number of people who are still committed to the level. Thus, these nearby levels have more validity or potency.

When a support or resistance level forms at a round number or a big figure, such as $100 or $1,000 or DJIA 10,000, many more people will remember the level because it is easy to remember and obvious. In turn, the more people who remember and act at that level, the stronger the support or resistance will be.


This is a good explanation of the psychology when trading these "imaginary lines", but for example it does not explain why distances between horizontal resistance lines have the ratios of the Fibonacci sequence. Horizontal lines are not that interesting, even though they are powerful. What's much more astounding are the angled lines and things like the Gann fan, or the Fibonacci speed resistance arcs.


Thanks. I think this only explains resistance/support lines which are horizontal, but in technical analysis these lines may also have a slope.


From eye-balling the chart, it seems to consistently lag. Am I mis-reading it, or is this not actually a predictor?


You're mis-reading it. The blue line is subsequent 10-year performance, so it's already shifted back (left) several years from its natural alignment on the chart.


If I'm reading this correctly, the indicator is currently at 0.42? Somewhat above the lifetime average of 0.35, and well below the 0.50 mark which would indicate stocks are overvalued relative to other assets?

I believe that corresponds quite accurately with the assessment we are currently at the early to mid stages of a secular bull market in equities analogous to the runs during the 1950s and 1980s. But as others have pointed out, it doesn't say anything about the possibility of a correction in the near term. Be careful out there!


When the yield curve begins to dip, a flight to quality will begin into crypto's.


Good luck with that. Usually there's a exodus from speculation the minute things get hairy in the market. Nobody speculated in crypto right?


Like there was in 2008/09 when most people in the world lost confidence in banking systems followed by the bailouts and Quantitative Easing.


> in 2008/09...most people in the world lost confidence in banking systems

You mean when everyone sold securities, moved to cash, and bought Treasuries, CDs and bank deposits?


2008 was great. Everything was on sale, and I doubled down on my stock investments...


With money from where?

If your money wasn't in the market, you might have been lucky with timing but statistically if you replay that scenario, you're losing out on gains by not just parking your money in the market in the long run.


I did get lucky, to be honest. I had low 6 figures in a total stock market fund and a similar amount of cash sitting around.


> With money from where?

Your salary, invest you savings in the market.



> And Bitcoin

Given Bitcoin's "capitalisation" failed to hit even $10bn until well after the crisis [1], it's safe to say that more capital moved, in almost any single U.S. state, from stocks to corporate bonds, than from anything to Bitcoin. Being optimistic about the future of a technology is fine. Being delusional about its history is not.

[1] https://blockchain.info/charts/market-cap


Ignoring correlations to market downturns and ignoring the data is delusional https://www.google.com/search?site=&tbm=isch&source=hp&biw=1...


Bitcoin came into existence in 2009. Nobody was buying Bitcoin in 2008 because it did not exist in 2008. We don't have data to support the claim that people would rush Bitcoin amidst stock market crashes. We do have data supporting the claim that they rush into traditionally safe assets, i.e. insured bank deposits and Treasuries.

Regarding Cyprus, wealthy Cypriots--by and large--bought German and Greek government bonds, not Bitcoin.


The crash came into existence in 2008 and Bitcoin began to rise in 2009. https://www.washingtonpost.com/news/the-switch/wp/2014/01/03...


> Bitcoin began to rise in 2009

Bitcoin came into existence in January of 2009. What was it supposed to do? Start stealing its owners' money and go negative?

Also, everything rose in 2009. The S&P 500 was up 23%. Bonds were up, commodities were up...that's what happens after you scrape past a global financial meltdown. Sure, Bitcoin was up like five thousand percent, but that's the difference between a tens of millions and hundreds of billion in market capitalisation, things that are very volatile and things that are not, and venture versus mature.


Everything rose but not before everything crashed in 2009: https://www.google.com/finance/historical?cid=626307&startda...

Those who've hedged with crypto's as a trading vehicle are likely to win again during the next inevitable downturn.


AFAICT, the source you cited (dated 2013) quantifies the investors merely as 'many Spaniards', and from the grandparent's source, we know that the Bitcoin market cap around that time was around $1bn, which is not even rounding error on the scale of financial markets. You're not ignoring the data, you're imagining it.


The religiosity on the side of "traditional" assets is exactly why they missed the boat. Whenever traditional models are threatened, the pitch forks come out. The VC industry is having it's Uber-moment. The disrupters are being disrupted.


People are downvoting you, but you're not wrong. Bitcoin is legitimately seen as an alternative to gold. So it's not crazy to expect people to buy BTC when shit hits the fan in the fiat markets.


No it's not. The only people that view Bitcoin as an alternative to gold are the people buying/trading in Bitcoin. No one else is that delusional.


> Bitcoin is legitimately seen as an alternative to gold.

Bitcoin may be an alternative to gold but it is not the same.

1) Gold has intrinsic value.

2) Gold is a tangible asset.


Gold having intrinsic value is a straw man argument for it being different than Fiat currencies. Consider this thought experiment. You are going to live by yourself in the forest for a month. Would you rather have A) a weeks worth of food or B) 1 oz of gold. I think this highlights there is no intrinsic value or at least much lower than what people claim. Obviously, there is place for gold in electronics and circuitry, but aside from specialized applications, the need for gold is pretty low.

Now, I know that I've committed a straw man argument myself by using a contrived example. To that, I'll say that the only "value" one has by having gold in the real world is that other people will trade you for it. But this is exactly the opposite of intrinsic value. The value of gold being entirely fabricated by people's desire to hold it.


Gold's price is not supported by its instrinsic value, I'll give you that. But the "forest" test is absurd. By that definition computers, chemotherapy and candy have no instrinsic value.

Gold's value comes from its (a) millennia-long history of stably holding value across cultures and technological domains and (b) its tangibility and physically-enforced scarcity. Its intrinsic value is a fraction of its market value, in part because the inflated price deters lots of uses.


But point a and b in you're assessment of gold's value are two of the most common arguments for crypto, specifically bitcoin, as well.

a) crypto holders believe that the value will hold because as more individuals use it to store their net worth, the harder it will become to manipulate. ie. a history of price growth/eventual stabilization will occur in time.

b) its algorithmically-enforced scarcity, which many people believe is as valid as physically-enforced scarcity. So long as hash-based cryptography always works.


I think there's potential for cryptocurrencies to find homes in the modern financial landscape. But your counterpoint (a) is based on network effects. Distinguishing between short-term bubble behavior and long-term resilience can only be done after knocking the system with crises. Is there hope? Sure. Is it demonstrated? Absolutely not.

There is one feature gold has over Bitcoin that cryptocurrencies cannot replicate: resilience across technological domains. Gold holds value without computers or electricity. Bitcoin does not. Gold, on the other hand, cannot travel at the speed of light. TL; DR each solves different systems of trade-offs.


If we somehow get to a world without computers and electricity, I doubt gold has much value either. Water and bread though, those will be valuable.


> Gold's value comes from its (a) millennia-long history of stably holding value across cultures and technological domains

That's recursive - you're basically saying that gold has value because it has historically being valuable. Which begs the question of why it has been valuable.

I'm not saying that this doesn't add some (most, in fact) value to gold. I'm just saying that this isn't a property that is enabled by anything specific to gold.


> this isn't a property that is enabled by anything specific to gold

With money, you want five things: fungibility, durability, portability, cognizability and stability [1].

On fungibility, gold is an element. It can only be extracted, not produced in a conventional sense. In fact, before the 1669 discovery of phosphorous, humans had only purified, from oldest to newest, copper, lead, gold, silver, iron, carbon, tin, sulfur, mercury, zinc, arsenic and antimony [2].

Out of those, lead, gold, silver and mercury are chemically stable, though only gold and silver are also physically durable. Both are easy to recognize, though more metals are "silvery" in color than yellow.

Gold won due to stability, in large part because of a few flukes. For most of human history, growth was flat and gold mining was minimal. When we actually started growing, the major powers were using gold. The rate at which they added to global gold supplies happened to mirror their economic growth; this gave gold a few decades of price stability. That memory, together with the millennia of use, forged a cultural memory in the furnaces of the industrial revolution that remains, vividly, to this day [3].

> That's recursive

Cultural memories, like trust, are re-enforced by network effects. Your observation is correct. Gold got where it got, in part, due to luck and then just stuck.

Can that be replicated? Perhaps. I personally think our obsession with gold is silly. But engineering that properly means understanding why it happened in the first place. At least amongst Bitcoin enthusiasts, I come across the types of comments you see others making in this thread, as opposed to bona fide introspection and defenses.

[1] https://en.wikipedia.org/wiki/Money

[2] https://en.wikipedia.org/wiki/Timeline_of_chemical_element_d...

[3] https://core.ac.uk/download/pdf/6252203.pdf


So what you're saying is that gold is JavaScript - it got significant early adoption because the alternatives (like VBScript) were worse, and then that just amplified until it boosted it into an unassailable position. ;)


I do admit that my forest example is absurd. I just wanted to elucidate, as you pointed out, the price isn't supported by human's necessity to survive.

Your last point brings up an interesting tangent, if gold wasn't treated as a store of value, what would its price be? Would it be similar to diamonds where synthetic ones are half the price? Are you aware of any sources that try to answer this question?


I'm not sure gold has been stable recently. it has regularly had 30-50% swings up and down over the last 10 years. recently well off it's highs.


What if trade is essential to survival, and gold checks all the marks as the medium for trade? Of course this could only be the case in a primitive society. If you believe contemporary society could revert to a primitive one, then you would be compelled to hold gold as a hedge.




Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact

Search: