jplewicke

Even better, there is a major effort underway to legalize investing in startups by anyone. A bill (HR 2390 ) has already passed the House, and there's draft resolution in the Senate that would do the same thing. The main limitation will be that people won't be able to invest more than 1% of their net worth in any given startup, and that startups can't raise more than $1 million through it.

http://crowdfundinglaw.com/ has some more information.

The first thing that comes to mind here is the enormous red bullseyes that Target puts on top of all of its stores. They're visisble from airplanes and satellites, but their symmetry would make them less useful for calibration.

An example: http://craphound.com/images/googleeasrthad.jpg

Thank you.

I know it's an easy analogy to reach for, but I wish that writers would think twice before comparing some trivial inconvenience to being raped. You may have just ruined the morning of a survivor of sexual assault.

There are dozens of other things you could say that would have the same rhetorical effect without trivializing a severe problem, and it's beyond impolite not to. All you have to do is write "Nonstartups are just horrible to work with, like standing in front of an automatic baseball pitching machine."

If anyone else is wondering how to do this on Heroku, http://stackoverflow.com/questions/6179398/how-can-i-use-a-s... seems to be the canonical Stack Overflow question. I'm going to try the Rack::ReverseProxy suggestion myself sometime soon.

You can get names and Facebook IDs of user's friends for anyone who signs in with your app. I don't believe you can fetch their friends through the API, so the game companies should have access to only edges in which one person uses a game.

My major concern with using a former nuclear missile silo would be that it's hard to know whether Russia's intelligence service knows that it's been deactivated. The odds of of a nuclear exchange with Russia are low, but I'd prefer to know that my fallout shelter might not be targeted for a direct hit with an ICBM.

They're also going to be selling off real-time feeds of per-company sales data to investors:

MasterCard has launched a new product for investors leveraging its huge global panel of transaction data from over 1.7B cards. This panel is global, more representative, more timely (weekly data to get ahead of monthly comps) and more customizable than any other transactional database available to the buy-side.

There is a broad range of topics on which we can provide investors with timely insights with actual sales transactions. Here are just a few samples of the topics we have done work.

• Detailed Macro reads across 14 sectors in the US and 4 in the UK

      •         Department Stores – Total sales, transactions and Avg Basket size across segments


      •         Office Supply Stores – Total sales, transactions and Avg Basket size for ODP, OMX and SPLS


      •         Home Improvement – Total sales, transactions and Avg Basket size for HD, LOW, LL and Ace


      •         Consumer Electronics – Total Sales, Transactions segmented by price point (e.g. to get a read on flat-panel/laptop sales)


      •         Other Retailers across a broad range of segments – Total Sales and Transactions


      •         Grocery stores – Total sales, transactions and Avg Basket size


      •         E-Commerce—Segmenting direct v. store sales trends, analysis of e-commerce companies like AMZN, NILE and many others


      •         Hotels – direct vs. OTA bookings/international trends, highly correlative chain hotels RevPar, Vegas-specific hotels


      •         Airlines – forward looking bookings as an indicator of RASM


      •         Cable/Satellite – ARPU, Churn and Total sales

I work for a quantitative stock trading firm, and if this data is an effective predictor of stock returns you are underpricing it by a factor of about 4x to 25x. Most data providers charge at least $10,000/year, with the more effective ones charging in the six figures.

The primary thing a quantitative trader would want to see is a rigorous track record. Ideally this would be a "point in time" record where you record the information that would have been available as of a certain time in the past. If you don't have the data recorded for that, or it's for too short a time period, the next best thing would be to rerun your metric on all the historical tweets you can get your hands on.

Showing potential customers a mostly complete historical dataset of what you would have told them is a fairly common way of selling this sort of data. Don't show them anything too recent so that they need to buy the data from you. Firms have different investing processes, and so they often need to see a larger sample of data in order to know whether it fits well with how they invest.

Please feel free to contact me if you'd like to talk more about this -- my email is my handle at Google's mail service.

There's a tool on Erwin that allows you to jump to random categories. I have my homepage set to random Featured Articles: http://toolserver.org/~erwin85/randomarticle.php?lang=en&...

Boston, MA (not remote)

MDT Advisers - We're a small quant investing fund working with:

- machine learning

- discrete optimization

- C

- large datasets

You can find a job description at http://www.mdtadvisers.com/careers/qea.jsp . The people, problems, and pay are good, and we aim for sane work-life balance(e.g. 45 hour weeks, private offices).

The team and management come from a technical background and don't micro-manage you. You also would have significant opportunity to work on your own research ideas and projects.

You can email me at jlewicke@mdtadvisers.com with any questions you have.

When I read this article, my first thought was that this would make a great option in a/bingo. UCB1, which is one of the bandit methods discussed in this thread, looks like it'd be relatively easy to implement: you just calculate a simple formula for each alternative in the test and choose the alternative with the highest result.

While it would definitely take some time and testing to see whether the bandit method worked better in practice, it might be even easier to work with than the current state of a/bingo. Instead of writing one line of code to choose between alternatives and then checking back after a bit to see which one worked best, you could just write a single line of code once and not worry about it until you wanted to clean that program up and standardize on one choice.

Boston, MA (not remote)

MDT Advisers - We're a small quant investing fund working with:

- machine learning

- discrete optimization

- C

- large datasets

You can find a job description at http://www.mdtadvisers.com/careers/qea.jsp . The people, problems, and pay are good, and we aim for sane work-life balance(e.g. 45 hour weeks, private offices).

The team and management come from a technical background and don't micro-manage you. You also would have significant opportunity to work on your own research ideas and projects.

You can email me at jlewicke@mdtadvisers.com with any questions you have.

It seems to be based off an anti-spam heuristic run by Cloudflare: https://www.cloudflare.com/ . The most insulting part of it is that they offer to link to evidence of what your IP address has supposedly done, but don't actually provide any concrete details(e.g. links to the alleged spam).

If anyone from Cloudflare is listening, your CAPTCHA submission is broken in IE7.

I wrote a peer-to-peer credit server in Neo4j and Ruby: https://github.com/jplewicke/rivulet . There were two reasons I used Neo4j: I needed to be able to quickly find paths to route payments along, and I needed to be able to update properties on multiple edges as an atomic action.

Google can and does include a Javascript interpreter in their crawler: http://www.seomoz.org/ugc/new-reality-google-follows-links-i... . They don't need to solve the halting problem in order to run a Javascript interpreter for a certain amount of time(maybe 10-15 seconds) and see how that changes the page content or redirects to a different site.

If you want to make your visitors wait until after the heat death of the universe before you actually redirect them, I'm sure you'll be able to fool Google too.

It'd be cool if they added an option to subscribe for $10/year for a quick SMS and email notification if your account is compromised. I'd get it for myself and my family.

That's interesting -- it looks like news.arc automatically removed the "%27" from both of our URL submissions, maybe as an over-zealous anti-spam precaution.

Londsdaleite futures, actually: http://en.wikipedia.org/wiki/Lonsdaleite :)

I was originally talking about trying to get currently profitable trading strategies out of the same pieces of accounting information that millions of smart competitors have access to. It's definitely a bit of a Red Queen's race ( http://en.wikipedia.org/wiki/Red_Queens_race ) trying to find and clean up new data sources and improve our trading methods well enough just to stay current. It's also hard because even though there's lots of data available, it's hard to tell how useful the early data is and whether you can still make money with trading strategies based on it.

Boston, MA (not remote)

MDT Advisers - We're a small quant investing shop working with machine learning, financial analysis, and the hardest dataset in the world. We're mainly hiring for a general analyst position that’s about 60% programming and 40% financial and statistical analysis -- http://www.mdtadvisers.com/careers/qea.jsp . The people, problems, and pay are good, and we aim for good work-life balance(e.g. no 60 hour weeks).

You can email me at jlewicke@mdtadvisers.com with any questions you have.

I agree completely if we're talking about a recommendation engine that's supposed to always be able to find something of immediate interest for you. However, that problem is at least as hard as building a general AI, since not even my friends can always think of something I would find immediately interesting.

If you relax the scope of what you want a recommendation engine to do, they can do an awful lot. Directed Edge is absolutely killing it here -- http://directededge.com/ . By providing a white-label API so that developers can specify the data on which to base recommendations and by using some superior algorithms, they're able to come up with impressive results for recommendations like "People who might like to buy this might also like to buy this", or "These users might be your friends", or "You should read this other article."

Except for when it's not. According to the linked article, their business model is that they'll help protect you from the patent trolling of larger companies. For example, if you pay them $X,000 a month and you get sued by Microsoft for patent infringement, RPX will try to counter-sue Microsoft for infringing one of their patents.

It's not a complete solution to patent trolls, but it seems like RPX could provide a very legitimate service to companies that want to focus on building their business instead of worrying about getting patent-trolled by a competitor. It can't save you from getting sued by a completely non-practicing entity like Intellectual Ventures, but nothing else can yet either.

There seems to be a major divide between two-syllable non-meaningful words and three+ syllable ones. Etsy, Zynga, and Google don't sound that hard to remember compared to Wesabe or Geezeo. I suggest a simple test for domain name desirability: the value of a domain name is inversely related to its score in Scrabble.

How do people feel about longer combinations of meaningful words, like BingoCardCreator, SecondMarket, or AppointmentReminder? I see a tradeoff between the number of total words and the complexity of each individual word, e.g. CreditKarma vs. HotOrNot.

The only place I really miss having them is in older articles and on searchyc.com. I feel a much greater compulsion to engage each comment on its merit when voting and when reading, and I feel like reading HN has become more intellectually stimulating.

However, this is more of a problem when I'm trying to assess information in areas that I'm not already familiar with. If I'm searching for information on which DNS providers are best and I find an Ask HN from 4 months ago, I can no longer tell what the true community consensus on it is. I expect the current masking of comments will provide less biased voting, so I think displaying comment scores on stories that are more than two months old would eventually provide the best of both worlds.

Strangely enough, Groupon was originally a startup named The Point trying to save democracy: http://www.thepoint.com/ . Andrew Mason, Groupon's CEO, was pretty psyched about a campaign to build a giant dome over the city of Chicago: http://www.thepoint.com/campaigns/chicago-winter-dome .

I had to do an IQ test when applying to a hedge fund in LA: Madison Tyler Holdings / http://www.ewtcareers.com/ . I'm not sure if they'd talked to legal counsel about it, and I've never seen it anywhere else.

Boston, MA (seeking interns, not remote)

MDT Advisers - We're a small quant investing shop working with machine learning, financial analysis, and the hardest dataset in the world. We're mainly hiring for a general analyst position that’s about 60% programming and 40% financial and statistical analysis -- http://www.mdtadvisers.com/careers/qea.jsp . The people, problems, and pay are good, and we aim for good work-life balance(e.g. no 60 hour weeks).

You can email me at jlewicke@mdtadvisers.com with any questions you have.

The quant trading shop I work for in Boston is:

Boston, MA (not remote) MDT Advisers - We're a small quant investing shop working with machine learning, financial analysis, and the hardest dataset in the world. We're mainly hiring for a general analyst position that’s about 60% programming and 40% financial and statistical analysis -- http://www.mdtadvisers.com/careers/qea.jsp . The people, problems, and pay are good, and we aim for good work-life balance(e.g. no 60 hour weeks).

You can email me at jlewicke@mdtadvisers.com with any questions you have.

This seems to be the archive in question: http://replay.waybackmachine.org/20090227065224/http://colin...

Boston, MA (not remote)

MDT Advisers - We're a small quant investing shop working with machine learning, financial analysis, and the hardest dataset in the world. We're mainly hiring for a general analyst position that’s about 60% programming and 40% financial and statistical analysis -- http://www.mdtadvisers.com/careers/qea.jsp . The people, problems, and pay are good, and we aim for good work-life balance(e.g. no 60 hour weeks).

You can email me at jlewicke@mdtadvisers.com with any questions you have.