CNTK contributor here - Keras indeed is pretty high on our list of things to cover soon. But then, all our code is out there on GitHub and we welcome PRs :-)

Many years ago, when I did my first steps in Emacs, I wrote Richard a mail asking, why it has such unusual and sometimes awkward keyboard shortcuts. He replied: "For fast and efficient editing."

I like that guy.

Exactly. And even if SO goes evil some day: one cannot point out the fact often enough how cool it was that SO shared the user generated data. We at TwoToReal.com benefit a lot from it (http://twotoreal.com/site/credits/)!

Any ideas/opinions why XMPP's mult-user chat (MUC) is still not widely used?

The article nailed it. Chances are that the world of QA sites is switching from narcism based reputation systems to the real web currency: user attention. That is, if I spend some of my attention on your problem, the system will remember it the next time I have a question.

This is what I am doing with http://www.TwoToReal.com: Unlike SO/Quora/etc. it is a synchronous QA site for questions where you didn't get an answer on the the former QA sites. As it works by pulling in other expert users into a real-time web chat, I had to be very careful to find a reasonable trade-off between disturbing the expert (i.e. asking another user by IM to participate) and benefiting the questioner. Rule-of-thumb at TwoToReal: the more you are allowing yourself to be "disturbed" the more will be asked for participation the next time you have a question. Soon it will allow you to "invest" more of your attention on one question by letting the system pull in more than one expert. Thus the currency of attention.

I believe that's part of the motivation behind bounties: if I've invested a lot of time and accumulated karma, I can spend that karma to improve the quality/speed of answering my question.

Re: TwoToReal. It's an interesting idea, although it seems like it adds burden to the answerer. One quality I like of SO is that it encourages people to ask better questions (keep concise, provide relevant details, etc), but in a system where I'd be getting real-time help I might become lazier about posting my question. "I have a problem with node.js" rather than "I'm getting this output when trying to use node in this way".

That's a totally valid point, gbelote. Nevertheless, there will always be questions where you don't know you to frame a question - even after having searched around for a while.

E.g.: if you are new to Linux and X doesn't start - how could you ask if you are new to the whole field? A bit of guidance by an expert could help in 5min. And then, sometimes it is just like the xkcd strip at http://www.twotoreal.com/site/about/

I view it as a last resort - after SO/Quora/etc. could not help.

That's a good point. I was thinking about it as a direct replacement of SO, but like you said it doesn't have to be. Good luck!

That's why my wife and I abandoned TV long time ago. First week was hard, but after that we have always been happy with how we spend our time.

Not sure whether it is in your vision as well, but wouldn't it be cool if Grove.io offered a library of such IRC agents in the same manner as HuBot from GitHub?

I think this is a super cool idea and it's something I've been thinking a lot about. Right now, existing IRC bots work fine with Grove but it would be cooler to have more integration.

Could even open a market of such bots for you: Customers of Grove.io could then offer their bots in your IRC bot market place ... but now I'm just dreaming ;-)

Cool idea. However, I am missing the method to auto-group chats by topic. Would be nice to have it with IRC.

Currently, I am doing that (not with IRC, but with Jabber/XMPP) via my side-project http://TwoToReal.com (beta): You ask questions (or raise topics) for which automatically determined experts are then pulled in via IM into a real-time web chat (most of the knobs/switches are tuned by machine learning).

What are your pricing plans?

Indeed. E.g., I am using git for more projects than I use mercurial. To these projects, however, I just submitted a limited number of change sets. Most heavily I am using mercurial instead, which I also would choose for future projects. Maybe it would more accurate to measure the number of submissions per time interval.

Also interesting that the Economist continues to come up with the same kind of story every week. Just check the last Economist links that were posted on HN over tha last weeks and you will see that they all have the same message. To be honest, I would expect them to be more diverse.

That's what the Economist is all about, though - extremely well written and produced with a very predictable and consistent message and opinion.

Zwieback - that's the coolest alias I've seen for a long time :-D Are you plattdeutsch?

No, I grew up in Stuttgart and now I'm an Oregonian. My mom grew up in Mecklenburg so that's pretty close.