ntoshev

I wonder if statistical machine translation approach could be usefully applied here. Get tons of source from github, compile with every compiler available, train on the result. It would be challenging to compile automatically at scale, to align the code with the source, or to get a source representation invariant to identifiers, but should be doable.

The cstore_fdw extension uses the Optimized Row Columnar (ORC) format, which seems to be PAX in your nomenclature. The improvements he accomplished seems to be related just to the first "fact" that you listed, which applies to all column stores.

"Google" does not think that, a cofounder of the Dart project does.

I thought LightTable is an IDE, not rethinking programming :) Anyway, great thinking, I've actually come to a lot of the same conclusions. I'd suggest an inspiration: spreadsheets. Can you implement the separation of data model from indexes/optimizing physical model in an ad-hoc spreadsheet-like UI? When thinking "spreadsheets" you should think about not just Excel, but also Lotus Improv (which is a more structured spreadsheet, which is both a strength and a weakness).

Fairly obvious extension: when arrived at a composite image, choose the one image from the set that is closest to it and show that instead of the composite to get a sharp version.

Rural Northern Greece is awesome, the people are awesome, but I haven't seen it there. Where were you, more specifically?

Dominant assurance contracts might be interesting here. It's a mechanism like Kickstarter, but if a project can't reach it's funding goal, the entrepreneur pays something the people who tried to support it. This makes waiting for others to fund a project that would benefit you as a side effect an inferior strategy to joining to fund it yourself. It would also make marketing the project (to the people that would benefit from it, in order for them to fund it) easier.

http://en.wikipedia.org/wiki/Assurance_contract#Dominant_ass...

Imagine a consultancy that organizes this type of deals. Identify projects worth doing with well defined and easy to reach beneficiaries and offer them this kind of contract. If that's properly done, they would want in and fund it. Then hire someone to execute. Repeat.

Perhaps this would be easier to do with geographically local projects first. Also it seems like the expertise required to organize this type of deals would be very different from the expertise required to execute them when they are funded, hence the separation of jobs. Fortunately, the projects don't have to be very sophisticated, they can be something as simple as a playground in an area that needs one.

There seems to be a fundamental trade-off between latency and throughput, with stream processors optimizing for latency and batch processors optimizing for throughput. The post says stream processors are actually used for reprocessing in production at LinkedIn, and simplifying things might well worth it being slower, but do they know how much slower this is than using Hadoop?

I have a vehicle routing solution (minimal routes via multiple destinations, with time windows, capacity constraints, weekly scheduling; it's a website service on top of Google Maps) that I would be happy to provide for free to social impact projects. Email is in my profile if you're interested.

Kivy does too.

http://kivy.org/

I guess "working correctly" for cyclic dependencies means you need to introduce the concept of time and be careful to update everything in the proper moment? What is this useful for?

I wonder if this technology will eventually become good enough to measure the heart rate of lying politicians from just video records. Bill Clinton would make an excellent example. Makeup might fool it though.

Just having the right culture is not a solution. You can have a culture that routinely goes for Pareto optimality. If that's not an evolutionary stable strategy, the moment you introduce agents that choose the Nash equilibrium instead, they would start winning and gaining ground.

I wonder about a different solution though: can you design a mechanism that transforms the Nash equilibrium of a Prisoner's dilemma game into the Pareto optimal outcome? Something similar to escrow?

[1] http://en.wikipedia.org/wiki/Evolutionarily_stable_strategy

The Wikipedia entry is quite unclear. The paper itself states at this point it's a proposal of academic and not practical interest, since a vote buying attack has been discovered.

http://people.csail.mit.edu/rivest/Rivest-TheThreeBallotVoti...

Is the table rendering the result set an image, or is there a way to render arbitrary HTML in a notebook from your Python code?

Came here to write this. Promise pipelining is an especially interesting attempt to solve latency in RPC (although it doesn't always work).

It might make sense to run nginx on a separate server / vps just for better security.

Well, sort of. With this architecture (not using a real database) it is fairly easy to save the state to S3, every minute. I don't have an automated way to rebuild the server in case of failure yet, but this is coming as part of https://t1mr.com

It should soon be possible to recover automatically, in a different data center, in matter of minutes. That's tolerable for me.

If you are just starting, you should have the simplest setup - everything on one server - and scale it only when it becomes necessary. Premature scalability adds complexity and slows down your iterations.

My setups usually consist of an nginx serving static content and proxying applications requests (doing gzip, etc). The data tier is initially collapsed into the application as described in http://www.underengineering.com/2014/05/22/DIY-NoSql/ This architecture allows very fast iterations while providing enough performance headroom; it can serve 10k simple (CRUD) http requests per second on a single core.

At https://t1mr.com we don't use SMS for notification, it's too unreliable. Besides delivery issues, there are benefits if the monitor knows if someone has really seen the notification - we know when somebody answers a call, but we can't know if he has actually seen the SMS.

It's a nice hack, but it's better to use external service for this. We run https://t1mr.com

One problem is you need separate infrastructure to host your monitor. You also need to monitor the monitoring service, or it is easy for it to be quietly failing until your real site fails without warning. We run a separate instance that only monitors the public instance of t1mr.

Attention to detail matters and you really want to be focusing on your product. We are quietly handling other stuff, like calling multiple people until someone really answers the phone, or checking if your ssl certificates are about to expire.

Right, so the value of the simulation is somewhat questionable. Exponential slowdown is enough to kill all intuition about what works and what doesn't. You only get some correctness gurantee and that would be for very small problems / test cases.

I don't really find the tree view useful. But I wish there was a way to see the code weight by individual files and whole repos: as KLOCS, size, anything. Is there such an extension?

Can you give good examples of the both kinds of software frameworks?

What if you tried to transcribe the music solely from Fourier transform of the audio source? I expect the piano has an abundance of harmonics, but there should be some way to distinguish them from the keys. Hasn't someone done it already?

You really need a data driven approach in this. Gather statistics of actual vehicles going from point A to point B so that you know the best way to do it in the future.

I'm running a routing optimization service (http://fleetnavi.com if you are curious, but it's only in Bulgarian for now, although Chrome translate works really well on the site). The routes a mapping service like google maps will give you are not that great really, in many cases the data needed to do a perfect route are just not available. We aren't using the stats from routes in the past yet, but it's a next step that would clearly be beneficial.

I believe the accepted definition for big data is data you can't handle on a single machine and need a cluster to process. So Moore's law makes it a moving threshold.

It was about time someone optimized charging/discharging of batteries. I didn't know ordinary batteries have such a potential for optimization, I thought that would be the batteries in UPSes. Current UPS batteries die in 3 years, having been used just a dozen times, apparently because the UPS insists to keep them fully charged at all times. Someone fix this, please.

Both HN and Facebook are not regular forums (they have distinct features). If you are trying to change politics with software, perhaps a good way to approach the problem is to try solving something specific, then build the community and write the code to do precisely that. One example is gathering support for specific campaigns. I live in Bulgaria, we have active protest movement against the current government, and a lot of the coordination and debate happens on Facebook. I had a specific campaign in mind that didn't happen for legal reasons, but it would have been very interesting otherwise. Send me a mail if you want to know more.

I think it's a mistake to think about real and virtual world as separate things. We live in a single world, and the online part grows more important with time. There are bits that are quite independent from the offline part, like interacting on HN with pseudonyms. Other bits, like Facebook, are typically much closer and interconnected with our offline life. Some mobile apps are even more integrated with the real world, although they are not very social.

The internet does rewire the fabric of society. Facebook has a big role in modern revolutions. I'm just skeptical that the forum format has the potential to bring further material advances.