by comatose_kid on 9/12/14, 3:46 PM with 391 comments
by dang on 7/15/23, 8:20 AM
by akkartik on 9/12/14, 4:25 PM
I've been working on this for several years, though a startup seems the wrong vehicle for it. I think the description in the RFS is misguided:
"We’re interested in helping developers create better software, faster. This includes new ways to write, understand, and collaborate on code, and the next generation of tools and infrastructure for delivering software continuously and reliably."
There's a blind spot in prose like this that gets repeated all over the place in our community: it emphasizes writing over reading. I think we have to start with reading. My hypothesis is that we need to reform the representation of programs to address this use case: I download the sources for a tool I use, wanting to make a tweak. How can I orient myself and make a quick change in just one afternoon? This is hard today; it can take weeks or months to figure out the global organization of a codebase.
You can't "deliver software continuously and reliably" until you rethink its underpinnings. Before the delivery problem there's a literacy problem: we programmers prefer to write our own, or to wrap abstraction layers over the code of others, rather than first understanding what has come before.
More on my approach: http://akkartik.name/about
by impendia on 9/12/14, 5:34 PM
http://www.amazon.com/Discrete-Mathematics-Applications-Susa...
$264.39, for students that work part time jobs at $7.00 an hour (before taxes).
Not only students are angry about this. Professors are angry, and authors are angry too. Bitter fights between professors and publishers are common.
Everybody wants to see the big players in this industry fail. Please, someone, make it happen.
by maxcan on 9/12/14, 5:51 PM
Sorry, this is not the right problem in financial services. Companies like Vanguard are already doing a great job of this and the costs are extremely low. Its a commodity product with razor thin margins that actually serves the needs of its customers well. Maybe there's a marketing issue where they aren't educating enough people, but that's not a technology problem.
As an alternative: Lower the Costs to IPO, disrupt Investment Banks
Sarbanes-Oxley, minimal competition between investment banks, and heightened SEC scrutiny have made the fixed costs to an IPO astronomical. These days a company, for the most part, cannot IPO for less than a $1 billion raise. This means that the broad public, including those index funds YC loves, is prevented from enjoying any returns at all for younger, high-growth companies.
There is room for startups to disrupt part or all of the process. It would be capital intensive and hard as hell. But, you're not looking for easy right?
by atonse on 9/12/14, 4:58 PM
I'm no longer a mid 20-something that can live on Ramen and 16 hour days. I'm married and have a young child.
Are there YC founders in this phase of life that were able to make it work in YC? What did you do differently? Is YC interested in working with these kinds of founders? (it's certainly a different kind of "Diversity")
by frandroid on 9/12/14, 4:48 PM
> Healthcare in the United States is badly broken. We are getting close to spending 20% of our GDP on healthcare; this is unsustainable.
That's mostly a policy problem, not a technology problem. Countries with single-payer healthcare spend massively less on it per % of GDP than the United States with its pro-profit healthcare system, and American doctors and healthcare corporations end up being fabulously more rich than in those countries. (And they still have private healthcare, like in Sweden, which competes with public healthcare organizations.) The other reason healthcare costs are getting higher is that people are getting older and thus more sick. That's a generational bump, there's very little we can do about that. Not that I'm opposing the types of ideas YC is after in this sector (preventative medicine and better sensing/monitoring), just that the premise is wrong that it's a technological problem.
> At some point, we are going to have problems with food and water availability.
That's because we dedicate most of our water and land resources to feeding cattle that we then eat. Innovations that will have the most impact in that sector will involve weaning people from animal products. Stuff like Beyond Eggs and lab-grown meat.
> It’s not a secret that saving money is hard, and that people tend to be bad at doing it. The personal savings rate has largely been falling since the early 80s.
Sure, some super-low-cost index funds would help, but the main problem here is two-fold: 1) real incomes are stagnant, due to government policies favouring corporations and 2) government/pension funds are much better at providing good ROI on investment than individuals can. Once again policy change is much more likely to have a massive impact than trying to improve the individual worker's investment returns. Collect retirement contributions at the source, and have the best investors in the country manage them. Without taking a profit for themselves. It's done elsewhere.
by ealloc on 9/12/14, 5:12 PM
These have traditionally been domains requiring a huge research apparatus with tremendous manpower, for only very long term gains. Not good for startups. In AI, how can a startup hope to succeed when academia has had almost no success in 50 years (and I am doubtful throwing more CPU/neuron layers will 'solve' the problem).
In addition, the people with the skills necessary to make progress are going to be advanced researchers with PhDs, who are good enough to remain in academia if they wish or who have already developed a proven-enough idea through their research career that they don't need Y-combinator-style money.
I am not trying to be a downer on the idea, contrarily I hope there can be success. Really I am fishing for anyone with a good perspective (or an answer) to these points.
by bravura on 9/12/14, 6:02 PM
I'd like to invite people to try the early release of Empire API, which is one API for every enterprise SaaS:
Empire is an API for accessing enterprise SaaS services such as Salesforce, Zendesk, Google Apps, etc. It provides a uniform, database-like interface to every service that it supports. Empire makes it easy to integrate data from multiple enterprise services into your own enterprise app.
You can click Login to create an account, and we'll send you an API key. Or you can just sign up for the mailing list.
by cik on 9/12/14, 4:34 PM
Between the staid companies that have been providing tools for decades that can be better, the tools that don't really exist that need to - I think we're ready. Similarly, with the security world starting to consolidate (FireEye buying Mandiant, likely goings public of companies like Rapid7 and TripWire), I'd think it's an ample rate/return option.
by wmeredith on 9/12/14, 4:08 PM
It could also be written like this: the government is a very bad customer with very large software.
by hazz on 9/12/14, 4:40 PM
Doesn't this already exist, in the form of the bicycle?
by jblow on 9/12/14, 9:42 PM
I came to Demo Day in 2010 (as an investor) but left without investing in anything, because I was so demoralized by the way it seemed everyone was trying to start lame web sites doing relatively trivial things.
If Demo Day looked like the stuff on this list, I'd be banging down the door to get in again.
by startupfounder on 9/12/14, 6:01 PM
Generation - Solar & Wind
Transmission - Distributed Grid
Storage - Batteries
Consumption - Electric Vehicles
> We believe economics will dominate - new sources must be cheaper than old ones, without subsidies, and be able to scale to global demand.
The world uses a huge amount of energy and it is vital that any technology is 1.cost competitive and can 2.scale on a globally. These are no small feats, but like Airbnb the assets already exist, but our access to them does not. This is a distribution and financing problem, not a creating new technology problem.
by zeratul on 9/12/14, 5:54 PM
Also it's a very difficult field of science. Now you need to be proficient in AI, machine learning, computational linguistics, linguistic corpora research, cognitive sciences, statistics, and sometimes physics if the text changes over time. Of course, you also need to be a good programmer. This combination of skills is very rare. Thus, very slow progress.
I suggest to start with well defined practical problems. For example, no one seems to do much with user generated reviews. There is some sentiment analysis but that is just a binary text categorization problem - not even close to general purpose AI.
It would be much more interesting to show a seller a time ordered stream of clustered reviews that depict only the most representative review for each cluster. This way a seller can see how his/her fixes/changes impact user reviews. Also it would be a great source for features and bug fixes requests. This is an ideal testing bed for clustering, novelty detection, categorization and mild inference. The inference is required because of sparseness of data.
This would create a good data set for a more general purpose AI. We would have reviews and text documenting changes and improvements of a new version of a product. Now the computer could start learning the dialog between users and product developers. Then, we are just one more step from statistical inference based question-answering system. Not a brute force system like "Watson" or a hand crafted rule base system like "Siri".
[EDIT:] I was thinking more about a decision support system that can recommend product changes. But in a way that maximizes customer satisfaction and minimizes the cost of implementation. The dialogue between past changes and customer reaction would give us the surface that needs to be optimized. This would generalize well to other domains where there is a text for request and a text for response - just to name one: clinical text in healthcare (position 5 in the RFS).
by aresant on 9/12/14, 4:02 PM
The http://www.reddit.com/r/oculus, http://www.reddit.com/r/oculusdev/, and https://developer.oculusvr.com/ are jam packed with excited hackers cranking out their projects and with the Oculus Connect conference coming up I'd love to see some of this talent pointed towards Y-Combinator
by nerfhammer on 9/12/14, 6:29 PM
There already several startups in the "personal saving" space largely based on index funds, though some of them have large minimums. Complex schemes may not be worth the effort for those with only a few bucks to spare:
https://www.futureadvisor.com/
It would be cool if Vanguard had an API so we could do the same thing open-source rather than incurring the extra management fees from these companies which are mostly based on Vanguard funds.
by bambax on 9/12/14, 10:19 PM
Oh yes, yes, yes. Everyone is talking about the quantified self but human augmentation would be so much cooler. I don't care if a watch can tell me my heart rate at all times (I know when my body is tired, or out of breath, because I live in it!!!)
But there are so many senses that I would like to have; for example, be able to always know where the North is relative to me. A device that would let me feel the North would be so cool and useful (I wear a Tissot T-Touch for that reason, but it's a very poor solution to this problem).
I think I heard the Apple watch will be able to do this, in some cases; but it sounds like an afterthought. I would pay serious money for a wrist bracelet or some other wearable that would do only that, but do it well.
by Permit on 9/12/14, 3:50 PM
A few of the accelerators I'd applied to in the past don't see the business opportunity present in developer tools (Who pays for those?) so it's a relief that YC recognizes the opportunity there.
by foobarqux on 9/12/14, 5:01 PM
This exists already, there are a ton of discount brokers with very competitive pricing.
by DodgyEggplant on 9/12/14, 4:06 PM
by foobarqux on 9/12/14, 4:52 PM
by mbesto on 9/12/14, 6:35 PM
I still am curious about this one. Is there any startup that has successfully done this in recent history?
Keep in mind, there is a big difference (IMO) between creating new jobs and shifting existing ones.
by foobarqux on 9/12/14, 4:55 PM
Isn't this like asking what comes after spoken language? Arguably the answer is nothing because language is what structured thought is.
by tsax on 9/12/14, 8:46 PM
by jolan on 9/12/14, 4:58 PM
by anon1385 on 9/12/14, 4:01 PM
This is a bad thing. Replacing opens standards with proprietary APIs locked behind access tokens hardly makes the internet more accessible.
Walled gardens are great for making money though…
by foobarqux on 9/12/14, 5:12 PM
Why does YC think that it will be different now?
by Jun8 on 9/12/14, 4:23 PM
1. It has often been commented that the next Google will come from the company who develops a palatable version "next generation TV". Hollywood 2.0 really covers only part of this.
2. A YC generator. YC emerged as an anomaly in the VC field and became widely successful. How can its success be replicated, both for US and in other countries. This was on a 2008 list that pg posted (http://old.ycombinator.com/ideas.html).
by agorism on 9/13/14, 6:16 PM
my implementation: https://github.com/zack-bitcoin/Truthcoin-POW
Science/Drug development: truthcoin allows for a new method of funding public goods that is far more impact per dollar than taxes.
Government: http://hanson.gmu.edu/futarchy.html
One million jobs: Successful traders can do it as a job. Every job that involves nothing more than decisions making can be replaced by a trader. Doctors, Governors, CEOs, various boards and counsels, Congressmen, Supreme Court Judges, etc. are all in danger of being replaced by anonymous, potentially uneducated, traders.
Diversity/Developing Countries: It will give modern financial instruments to everyone with internet connection. Insurance to hedge everyone's risks.
Enterprise Software: prediction markets are very unconnected today. Each business has a separate in-house prediction market built. They have to be heavily regulated. Truthcoin makes prediction markets as available as google or facebook.
Financial Services: prediction markets are a financial service. They are illegal. Use of a blockchain circumvents the law against prediction markets.
by orky56 on 9/13/14, 4:07 PM
Our favorite use case is a medical office that has made the jump to EMR. Much of their data is digitized but there is little benefit directly to them. However, once that digital data becomes a part of their workflow they actually see the benefit in it.
We are trying to think more generally about each of these problems and providing guidance to our individual customers specific to their vertical. Although it may seem we are still finding product-market fit, the small business market is always fragmented and diverse so this is just the reality of the situation.
by pptr1 on 9/12/14, 4:23 PM
I had my doubts about @sama but he is pretty much on the right track and and seems to be the right person for the job. They are fighting the typical SV stereotype about not funding big ideas. Go YC keep on disrupting!
by colmvp on 9/12/14, 3:50 PM
Isn't that already catered to by experiment.com, a YC company?
by pja on 9/12/14, 8:06 PM
I think that's called the bicycle.
(Of course there are plenty of barriers to people actually using the things, especially in the US, but those have little to do with the machine itself and everything to do with the social context in which it's used. If a startup can manage to solve those problems then more power to them. Start by talking to the Dutch perhaps.)
by mqsiuser on 9/12/14, 7:01 PM
I have been working (and still am working) with incredibly expensive software from IBM. Now it's somewhat trivial what it does and I have open sourced it (http://www.use-the-tree.com). Feel free to contact me. My dream is to find a team, apply to YC and crush IBM (I am somewhat serious).
by kfcm on 9/13/14, 3:05 PM
While a great list, the mistake is being made of assuming disruptive startups (through technology, business processes, or both) will solve problems in several of the sectors. Some being Healthcare, Pharmaceuticals, Energy, Transportation & Housing, Telecommunications, and--most of all--Government.
It's not because smart people haven't developed solutions to the problems they see in these sectors. It's that government--through policy and/or bureaucracy--often prevents those solutions from being implemented.
For example, I have family who are in the pharmaceutical industry. PhDs and all. The amount of money spent on the bureaucratic steps to obtain government approval (FDA and others) to bring a new drug to market is the vast majority of development costs.
I won't even go into my experiences working with various levels of government on technology projects. Suffice it to say, never again.
You want great, efficient solutions to hard problems? Get government out of the way. Government involvement hinders progress, and makes what progress there is extremely expensive.
by chrisacky on 9/12/14, 3:53 PM
"An important trend is the API-ification of everything. As more and more businesses are accessible with a web API, the Internet becomes more and more powerful."
I think a POSTman style Zapier, love-child would go down very well. Also products like Mashery where you provide APIs as a service and charge.
Notably missing from the list...
- Anything related to travel.
- Anything related to storage
- Anything related to video (Maybe that's Hollywood 2.0?)
by dobbsbob on 9/12/14, 7:23 PM
Actors are now being hired based on the amount of followers they have on twitter. If you have a lot of worldwide followers you're guaranteed to be casted since that's where all the money is these days.
>Government
The problem with this is dealing with government technocrats who will never deploy your software as is and will demand all sorts of complexity, basically creating the same garbage they were using before. If you can somehow survive this insanity there are gigantic contracts up for bid, for example many post offices are using Microsoft Mobile devices and looking into some kind of wearable scanner that doesn't charge hefty MS licensing fees. Then there's the Integrated Case Management software contract for $182 million that still doesn't work http://www.vancouversun.com/technology/government+million+co...
by pbiggar on 9/12/14, 11:55 PM
That said, I had more success building what wasn't in the RFC (dev tools in 2011), than what was (journalism stuff in 2010).
by taigeair on 9/12/14, 4:00 PM
by wj on 9/12/14, 7:34 PM
Ultimately I think something that provides a whole financial picture is what is needed (I think Learnvest is trying to do that). I picture something along the lines of a combination of Credit Karma, You Need a Budget, and Vanguard as being the way to go.
My email and twitter are in my profile if you would like to talk.
by rhspeer on 9/12/14, 8:37 PM
Learn by making awesome things to solve problems, then attach rockets and see what happens. It's a good approach, that's more fun for all and more accessible to folks that don't learn from the book & lecture model well.
I really like the prototype, simulate, 3D print approach. Being able to take home a flying model at the end of a lesson is just awesome.
Full Disclosure: Chris, the lead on the project, has been one of my best friends for the last 25 years so I'm biased.
by walterbell on 9/12/14, 5:00 PM
We need business models which support partially-open graphs, e.g. object IDs and some metadata are open, some metadata is closed via API. Open metadata can be cached offline and standardizes models within a vertical market, driving demand for paid metadata. API revenue grows as the open graph grows.
Semi-open knowledge graphs reduce the cost of adversarial algorithms. Better to have many competing skynets than one big skynet. The open part of the graph lowers the cost of consumption. Multiple, closed annotations on the graph compete to support AI use cases, and can optionally become open as value moves to higher-order representations.
by bfe on 9/12/14, 7:31 PM
by foobarqux on 9/12/14, 5:05 PM
That's not because saving is complex or low-yielding, it is because middle class non-disposable expenditures have gone up. (see Elizabeth Warren)
by vishalzone2002 on 9/12/14, 4:10 PM
by bradleysmith on 9/12/14, 4:14 PM
by saosebastiao on 9/12/14, 6:04 PM
by pyb on 9/12/14, 5:29 PM
by kibaekr on 9/13/14, 12:36 AM
by npostolovski on 9/12/14, 11:40 PM
by mbenjaminsmith on 9/13/14, 6:25 AM
by dasmithii on 9/12/14, 7:29 PM
by bfe on 9/12/14, 6:47 PM
Maybe somewhere in the direction Meteor and Light Table are heading?
by flipside on 9/12/14, 4:15 PM
If anyone else is interested in Hollywood 2.0, hit me up (check profile). Tinj is reimagining content ratings, reviews and recommendations.
by jdp23 on 9/12/14, 5:46 PM
by dharbin on 9/12/14, 3:53 PM
by edawerd on 9/12/14, 6:23 PM
by _tlf7 on 9/12/14, 9:52 PM
by ivv on 9/13/14, 1:28 AM
by desireco42 on 9/13/14, 6:50 PM
by mladenkovacevic on 9/12/14, 7:58 PM
by napoleond on 9/12/14, 4:44 PM
by chris123 on 9/14/14, 8:21 AM
by webmaven on 9/12/14, 4:40 PM
by calebm on 9/12/14, 4:37 PM
by AndrewKemendo on 9/12/14, 6:14 PM
by yaur on 9/12/14, 6:30 PM
I have a side project where I am aggregating celebrity generated content and one of the things that surprised me, though it really shouldn't have, is how bad the underlying content is. New tools aren't really going to help here beyond creating an incentive for celebs to create better content.
by lettergram on 9/12/14, 4:10 PM
by dhillonj on 9/13/14, 8:35 AM
by miguelrochefort on 9/13/14, 6:33 AM
In practice, this probably means something like a "Semantic Marketplace + Contract and Reputation Management".
When are they going to explicitly ask for it? I don't know. Time will tell.
by Joshyuen on 9/13/14, 6:22 AM
by joeguilmette on 9/12/14, 6:48 PM
by graycat on 9/12/14, 8:37 PM
I mean, reading over these requests, my guess is that, while some of them, if successful, would help lead to a better life for nearly everyone and save the world, so far not many people would "want" the results in the sense needed by a startup.
Some of the requests are nearly hopeless: E.g., the US Federal Government via DoE, NSF, and NIH have been spending billions on research in energy and medicine for decades. The idea that a YC start up could do a lot better makes most long shots look like sure things.
Next, a lot of these requests ask for some darned challenging research projects, and just a research project is one of the worst insults passed out by the venture capital community. Instead, no matter what the Web sites of early stage venture firms suggest, such firms want to see traction, not research projects, not even projects to write software for research already successfully done, not even to go live with software already written from research projects already done.
Example? Okay, want a research project to solve a big problem? Okay, consider security and reliability of the complex systems of large server farms and networks. We'd like to do better, right? For this, the first step is essentially near real-time monitoring for ASAP detection. So, we want detectors.
First big problem is getting a good combination of rates of false alarms and missed detections; I'm correct here; think for a few minutes and otherwise trust me on this one. Or, we're trying for a good combination of false positives and false negatives. Or for a good combination of Type I and Type II error.
Right, you guessed it, oh how you guessed it: Such detection has just two ways to be wrong -- a false alarm where we say that the system is sick when it is healthy and a missed detection where we say that the system is healthy when it is sick. Inescapable. Have any doubts, then think for two minutes. With me again now?
Okay: So, right, such monitoring and detection is a case of ASAP, essentially real-time, statistical hypothesis testing. I know; I know; you don't want anything that is just statistical. Neither do I. Tough stuff. We're necessarily, inescapably stuck-o never the less. Or, want a detector with no false alarms? Got that one for you -- just turn off the detector. Want a detector with no missed detections? Got one of those, too -- just sound the alarm all the time. Yes, some detectors occasionally make correct detections and have essentially no false alarms, but such detectors will detect only problems of a very narrow kind and otherwise have a high rate of missed detections. Arguing is futile -- the hard stuff isn't here, and I'm correct here. We're talking research here, like YC now seems to want to see, and research can be tough stuff to swallow. Keep reading ....
Now, what the heck to do about this? Okay, we've got some good news that, right Andreessen Horowitz should understand quickly: We can get data on each of several variables, maybe dozens or hundreds, at data rates of a point each few seconds up to hundreds of points a second. At a big, complex system, we're talking big data. So, we want our statistical hypothesis test to be multi-dimensional. Sure, go to the library and find a lot of those, right? Wrong. You won't find much. Next, most of the material you see on statistical hypothesis tests wants the probability distribution of the data when the system is healthy. Tough since for these complex systems there's no theory that will give you means of finding such distributions (we're talking multi-dimensional), and, even with big data, anything like accurate estimates of multi-dimensional distributions is agony with the curse of dimensionality. So, now what? Okay, we want to be distribution-free, that is, have a statistical hypothesis test that makes no assumptions about the probability distribution.
So, how many multi-dimensional, distribution-free statistical hypothesis tests did you find in the library? Not a lot. Maybe the only ones you found were mine. Mine? Yup.
But, when the dust settles, we do get a (large class of) genuine statistical hypothesis tests that are both multi-dimensional and distribution-free. So, right, with meager/standard assumptions, as is standard we can calculate false alarm rate and set it in advance and get it exactly in practice. For detection rate, as is usually the case we don't have enough data to use the best possible Neyman-Pearson result, but there is good reason to regard the detection rate as relatively high.
So, any large server farm or network doing important work and interested in security and reliability, , that is, nearly all of them, will be interested, right? And any VC firm, too, right? Nope. Don't hold your breath waiting. I only wrote nearly every information technology venture firm in the country, indeed, some months before the bubble burst in 2000. Responses? Even during the days of big bottles of Wonder-Bubble, none or f'get about it.
Lesson: Research, even for a big problem at important enterprises, even done research, even with algorithms to make the computing fast, even with prototype software running, even with a research paper that passed high quality peer review, doesn't get venture funding. I learned that lesson. Here HN and YC can learn it now or learn it later. Now is easier.
by ycskyspeak on 9/12/14, 6:23 PM
by pinaceae on 9/12/14, 9:49 PM
great marketing, kudos.
by maximem on 9/12/14, 4:08 PM