eob

Thank you!

Hi skrowl --

We know and understand your hesitation. It's coming this week. We'll have a free tier that we're aiming to make pretty broad in capability.

This is something we've been hearing from a lot of folks -- spreadsheet-driven websites are actually a really popular practice that we think should get more love and attention. I'd love to help you package that site you made for cloning and reuse on Cloudstitch so others can use it, too.

- Ted from Cloudstitch [ted@cloudstitch.io]

It's a fair question, but from my heydays doing ML research even good libraries can be pretty revolutionary. The engineering-side of ML is still in relative infancy.

For a great example look at all the papers Andrew McCallum's group has been able to publish by building on top of FACTORIE: they get to focus their time on the problem at hand rather than all the math required to solve it. Basically they can write code that generates a model dynamically but the framework handles all the inference. Compare that to how these things are built without such a framework: you spend most of your time painstakingly hand-deriving update rules and then implementing them as code.

IMHO, the exciting thing is that ML is getting closer and closer to being an everyday tool for engineers rather than something that requires you to be a full-time math person to use effectively.

I think I started, and then shelved, Anathem three times before finally powering through the first third of the book and like you was hooked once I did.

Interesting.. so what's the backstory behind the company that got the contract?

It may just be the PIs have a good track record..

The Shakespeare sonnet wasn't written by a computer. IIRC Nate just had a text editor with a type-ahead suggest box that was weighted by an n-gram lookup into your corpus of choice. It was up to the human to choose which word to use.

This would be against the spirit of the offer, but I wonder if there is any easy way to arbitrage that credit into cold hard cash with low effort.

E.g., boot up a bitcoin mining cluster (though this particular idea would obviously yield low return since it would be running on off the shelf hardware)

You're comparing things of different scales. Your body has many cells, but cancer in one can take down the whole system. It's more like a self destruct button on each AWS machine bacause they have a nasty habit of propagating a worm that spreads across the rest of the data center

CS PhD here: In our field, yes.

That's pretty awesome. I just tried it out to see how you make money and noticed that you have both banner ads and also open popup-tabs. Are you comfortable sharing how much of the revenue is from the banners versus the popups? Are those the only two streams?

FWIW according to the listserv traffic at MIT CSAIL right now, some members of the community seem to believe that the TOR node lives in CSAIL, not LL.

The repayment with respect to industry work is in credentialed autonomy, not money. And for many, that's worth it.

If a permission-enabled filesystem was bolted on top of BTSync, this could result in all sorts of fascinating log-based applications.

Wow. Just as another data point to show how monopolies stifle progress, in Cambridge MA I pay $50/mo for 6Mmbps. Comcast is the only Internet company available.

We did this with sqlite about six years back and got awesome performance results. It was a system called SyncKit. Still would have preferred relational storage in the browser, but glad to see at least we're moving toward something.

http://edwardbenson.com/papers/www2010-synckit.pdf

Re: challenging the network effect.

A telling moment that happened two nights ago for me. A friend from ~8 years ago came to visit and we wanted to send a joking message out to a few people from the old group of friends we used to hang out with.

The conversation was literally:

A: "Post it to Facebook and tag everyone." B: "Ehhh. It would be funnier to find an old email thread and reply to it."

We went with B. It seemed like a better idea, plus we weren't sure if folks in the group checked Facebook. But everyone reads email -- and sure enough, we got life updates from the whole gang within 12 hours.

Your observation isn't necessarily in contradiction, though.

People hate Comcast, and yet we all use it because (for many of us) there are no other options.

Sentiment is an important measure because it indicates the activation energy required to leave if a sufficient replacement emerges.

Speaking entirely divorced from my opinion on China politically (as someone with strong US and TW ties, FWIW), I think the Great Firewall has been a brilliant economic move in retrospect. While it may have been an accidental side-effect, blocking so many foreign sites has enabled a flourishing of domestic internet companies of the likes that nobody but the US has seen unless I'm mistaken.

And for a "middle class" of blocked sites, while not technically blocked, the packet delay is so great most people (except a dedicated few) will just shrug and figure that cross-continent internet connections must be too slow to be practical. Accessible, but only if you're willing to wait a minute. A brilliant political play, because it's hard to pin a charge oppression on anyone, since the information is technically there, but so slow nobody will access it.

Is development on FACTORIE ongoing? I played with it a bit a long while back and loved the "imperative-declarative[1]" idea you all were pushing for model construction toolkits.

[1] edit: Because I'm not sure if I just made that phrase up or if it came from one of your papers, the idea that ML libraries that take declarative model descriptions are great, but what's even better is if we also have an imperative API that can dynamically generate those declarative specs for us, even based on train-time inputs, so we can essentially "program" the structure of a model but still benefit from keeping everything generalized and declarative at the base.

Agreed. After reading several paragraphs about how speed as the only toggle was a brilliant feature, I started wondering, "Wait a minute. Am I reading reporting or an aggressive attempt at framing the lack of features by a PR team?"

Agreed. Now that the "get rich quick" phase of mining is over, there aren't many incentives for all the costs incurred.

So if drivers pooled together, using the same account from Lyft's perspective, they could pull off a $6,666 / month guaranteed salary for 40hr weeks. I.e., four accounts working 60 hour weeks are staffed by six people working 40 hour weeks. (I doubt the terms of service would smile upon this activity)

Without paid advertising, how do you advertise the service? I've created a micro-service like this in the past, and where I really fell short was getting the right people on the site. AdWord conversions were also too low in my case.

Looks awesome, guys! I've been wanting something like this for a long time.

One question: can I deploy to my own server (for free) or only through your hosted service? It wasn't clear from the documentation I read online.

Do any of the audiophiles on HN know if it's just fashionable to hate Bose or are there actually double-blind studies done in which music listeners consistently rate Bose as worse than price-comparable alternatives?

I'm not an audiophile by any means, but I also get the sense that many around me seem to have disdain for Bose. I also stumbled into possession of QC headphones and some Bose speakers for my house and think they're great. So I've always wondered at what the precisely stated and justified beef is.

This is dependent on what field you're in and where you graduate from, of course. CS is less heavy on postdocs [1] and I'm biased by only having had the experience of top-tier schools.

As a tenure track professor, you don't really answer to many people. You have to beg for money (grants), constantly prove yourself (publications), and always be recruiting (students). Of course you also have to teach classes, do community service, and participate in department politics and initiatives. I'm not counting these as "answering to people," however, because none of them involve someone telling you to do something. They're all activities you must decide to do yourself, and then it's completely up to you how to do them. Is this what you meant by answering to lots of people, or something else?

The relevant academic joke is: "It's the best job ever! You can do whatever you want with your 80 hour weeks!"

[1] In CS, from my experience, postdocing is usually a 1-year "reset your brain" time before taking an academic job. I know in other fields it's often a mandatory many-year grunt-work experience.

One challenge with the PhD->CS Industry sidestep is that PhD candidates are extremely picky about the kind of jobs they're willing to accept, and many companies are suspicious of the value/cost ratio of a fresh PhD.

If you're a graduating PhD headed into academia, you will be the boss of your group, with almost nothing above you in the org chart telling you what to do. So it is natural for a PhD headed into industry to start from a place of expecting the same---from their perspective.

But from the perspective of industry, the PhD's value isn't judged the same way. They've just spent six years doing research, not shipping code. Their management experience is managing students and classrooms, not employees.

So this mismatch results in a perception of "I'm being undervalued!" from the perspective of the PhD and "He's overvaluing himself" from the perspective of the company.

If the student and company don't have high social IQ to navigate this situation, I can see how it would result in no offers being made.

Source: I've made hiring decisions about PhDs when I worked in industry, and now I'm a graduating PhD student going back into industry, so I can see it from both sides.

At one point the E4X standard (back in 2007?) enabled a cool multi-line string hack by promoting XML elements to language entities. You could say:

    var htmlFragment = "" + (<r><![CDATA[
    l(a
    
    le
    af
    fa
    
    ll
    
    s)
    one
    l
    
    iness
    
    - e.e. cummings
    ;
    ]]></r>);

And then `htmlFragment` would contain your multi-line string. (The addition with the empty string coerces the <r> element to output its `toString()` value, which happens to be the innerHTML).

This is largely a historical factoid, though. I don't think browsers beyond Mozilla ended up supporting E4X.

Really clever hack with the function's toString method!

I wonder how long it's going to be before Dropbox announces another vertical. They have an amazing product, but it's basically been unchanged for years now and been allowed to succeed (and thrive!) as such because competitors have always treated their own Dropbox-style clients as a hobby instead of a product.