New Requests for Startups

Not another choice, just a simpler way to choose funds, with recommendations, etc. Wealthfront and a few others have offerings that do a pretty decent job at this, but they don't cover 401k plans, and their fees are high for something that should be set and forget. Target date funds are actually pretty good at doing this, and some have relatively low fees.

A lot of the 401k providers just dump a list of 40 funds on your lap, and let you figure it out, but it's information overload.

I really want something that can help me decide what type of risk I can handle, and then properly choose funds / balance between all of my accounts. Financial advisors are able to do this, but software can do it better and cheaper.

Basically, I want something like Mint that connects to all of my investment accounts, and sends me an email once or twice a year with specific instructions for rebalancing.

It's a good point. I was also thinking SalesForce too, but then thought to myself "actually I know tons of former SAP people who now just work for SalesForce/partners and thus just shifting the jobs elsewhere". Selling CRM software in 2001 is not dissimilar to selling SaaS CRM in 2014.

I'm genuinely curious about YC's mission here and some examples. I'd love to position a concept towards that goal but can't quite grok it.

What about the YC motto "Make something people want."?

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.

And two big pieces of the cost savings are:

1 - using buying leverage to negotiate prices, something republicans owned by the drug and device industry specifically banned (see eg [1])

2 - rationality about end of life care, which we spend a lot of money on -- 20%+ off the top of my head. As many doctors have shared, they often choose not to aggressively treat terminal illnesses and focus on quality of life. Unfortunately (remember Sarah Palin's death panels, and let's all thank John McCain for bringing that snowbilly grifter to the national stage), attempts to do things like pay doctors to sit down with patients and have end of life conversations, explaining what is happening have been successfully yet stupidly fought off. Whereas when doctors talk about how they die, they often chose to undergo very little treatment [2,3]

   Almost all medical professionals have seen what we call “futile care” being 
   performed on people. That’s when doctors bring the cutting edge of 
   technology to bear on a grievously ill person near the end of life. The 
   patient will get cut open, perforated with tubes, hooked up to machines, and 
   assaulted with drugs. All of this occurs in the Intensive Care Unit at a 
   cost of tens of thousands of dollars a day. What it buys is misery we would 
   not inflict on a terrorist. I cannot count the number of times fellow 
   physicians have told me, in words that vary only slightly, “Promise me if 
   you find me like this that you’ll kill me.” They mean it. Some medical 
   personnel wear medallions stamped “NO CODE” to tell physicians not to 
   perform CPR on them. I have even seen it as a tattoo. [2]

This doctor summarizes his choices as

   my physician has my choices. They were easy to make, as they are for most 
   physicians. There will be no heroics, and I will go gentle into that good 
   night. [2]

A different article

   Research shows that most Americans do not die well, which is to say they do 
   not die the way they say they want to — at home, surrounded by the people 
   who love them. According to data from Medicare, only a third of patients die 
   this way. More than 50 percent spend their final days in hospitals, often in 
   intensive care units, tethered to machines and feeding tubes, or in nursing 
   homes. [3]

There is almost always something that a doctor can do, but patient comfort is approximately priority F.

   More typical was an almost eighty-year-old woman at the end of her life, 
   with irreversible congestive heart failure, who was in the I.C.U. for the 
   second time in three weeks, drugged to oblivion and tubed in most natural 
   orifices and a few artificial ones. Or the seventy-year-old with a cancer 
   that had metastasized to her lungs and bone, and a fungal pneumonia that 
   arises only in the final phase of the illness. She had chosen to forgo 
   treatment, but her oncologist pushed her to change her mind, and she was put 
   on a ventilator and antibiotics. Another woman, in her eighties, with 
   end-stage respiratory and kidney failure, had been in the unit for two 
   weeks. Her husband had died after a long illness, with a feeding tube and a 
   tracheotomy, and she had mentioned that she didn’t want to die that way. But 
   her children couldn’t let her go, and asked to proceed with the placement of 
   various devices: a permanent tracheotomy, a feeding tube, and a dialysis 
   catheter. So now she just lay there tethered to her pumps, drifting in and 
   out of consciousness. [4]

And finally -- you should read all of [5], though it's heart-wrenching -- many terminal patients don't want to be aggressively treated when outcomes and the experience are fully explained. A close family member had to make similar choices and chose to die at home. The surgeons and oncologist where happy to keep going, but he was dying, and nothing the doctors could do would change that. They could only prolong for another couple months the inevitable, at the price of excruciating pain, repeated surgeries, and drugs that made him feel terrible. He chose to die at home. And not only is this far more humane, but far cheaper.

[1] http://www.politifact.com/wisconsin/statements/2012/sep/04/t...

[2] http://www.zocalopublicsquare.org/2011/11/30/how-doctors-die...

[3] http://www.nytimes.com/2013/11/20/your-money/how-doctors-die...

[4] http://www.newyorker.com/magazine/2010/08/02/letting-go-2

[5] http://www.newyorker.com/magazine/2010/08/02/letting-go-2

Looks like the Jobs quote was a rumour, at least according to BusinessWeek in 2001:

"Other stories claimed that Apple Computer co-founder Steven P. Jobs got an early peek and made the wacky prediction that cities would redesign themselves around the device. (Jobs denies he ever said this.)"

http://www.businessweek.com/stories/2001-12-16/commentary-gi...

But then, I wrote "reportedly", and it was indeed widely reported at the time :)