ziedaniel1

What if you released the script under a permissive license? (Contrary to ericjang's suggestion :) )

Would this eliminate the risk of a lawsuit?

I think you're underrating the difficulty of this kind of improvement. Rockets already have to be extremely well-optimized to reach orbit, so there wouldn't be the kind of low-hanging fruit like removing 300 pounds of insulation without side effects. Also, if you want weight reductions to give payload increases in a 1:1 ratio, the weight has to be taken off of the final stage (in F9's case, the second stage); removing it from earlier stages gives you much less benefit.

I count in Gray code using my fingers. It's twice as fast as binary because the LSB switches half as often! I've also found patterns that are even better, but they're harder to learn.

It's possible to do a lot better than TOFU or Keybase. Andres Erbsen and I have been working on dename, a key distribution system based on the anytrust model: as long as at least one of the servers you rely on is good, you're safe. See his writeup here: https://andres.systems/blog/2015-07-22-another-take-at-publi...

Seems like XOR is better for approximating a random oracle, and appending is (negligibly) better for ensuring collision resistance. People often are not clear about which of these two very different properties they actually want out of a hash.

Another thin client technique I haven't seen mentioned much: VerSum. https://people.csail.mit.edu/nickolai/papers/vandenhooff-ver...

97% environmentalist -- the same way he called himself 97% feminist: http://www.scottaaronson.com/blog/?p=2091#comment-326664

For what it's worth, once you're a customer, future purchases are fulfilled in 1-2 weeks (even if you've e.g. cancelled your subscription).

If anyone's interested in proving distributed algorithms correct, they should check out the Verdi project (https://github.com/uwplse/verdi), which has proved Raft correct in Coq. I imagine handling Byzantine faults and the full complexity of SCP would be quite a bit harder, but probably doable.

To me, though, it would be more interesting to prove the implementation correct. Rather than trying to prove an existing C++ implementation correct, it's probably more feasible to reimplement the algorithm within Coq and extract to runnable code. Verdi already supports that, but unfortunately it doesn't support disk state.

The numbers in the article are all wrong. SATA's ceiling is 600 MBps (megabytes per second), not 600 Gpbs (gigabits per second). SAS goes up to 12Gbps, not 12GBps, which is actually the same thing as 1.5GBps. At least the PCIe numbers look right.

It's cool that the .NET version actually JITs specialized serialization and deserialization code at runtime. This is one place where managed languages really shine, because emitting bytecode is easier and more portable than emitting, say, raw x86. It's also safer -- the runtime can verify the memory safety and type safety of the code.

Why wouldn't they? Tech internships tend to be well-paid and a lot of fun.

The bathroom server was up until just a few weeks ago. I'll have to ask someone here what happened. (We've had the system for over ten years, I think, so some of the equipment is getting old. )

It's a shame Udacity hasn't worked as well as Thrun hoped. I'm convinced that the failure of existing online courses has been primarily a problem of motivation, rather than an inherent pedagogical problem in online coursework. It's very hard to motivate yourself to consistently work on coursework if there are no deadlines and dropping the course just means you keep going with it sometime later. Still, I can't really explain why the SJSU courses failed as well; it sounds like students had actual deadlines, and were "required" to complete the courses to the same degree they're required to complete their regular courses. It's possible that the online tools to ask for help didn't work as well as in-person interaction, or that sitting in front of a computer doing things at random times is less motivating than actually attending lectures in person at regular times. There's one thing I'm certain of, however: it's a huge waste to have universities pay professors (who are selected for their research ability, not teaching skills) to repeatedly lecture the same material to small groups of people. A hybrid model might work best. Lectures could still be online (but possibly at fixed times to prevent procrastination), but they would be supplemented by in-person recitations, where students would meet in small groups with instructors and have the opportunity to ask questions or review particular topics, with the instructor as well as with peers. This model is still far more expensive and less accessible than online courses hoped to be, but it might actually work.

The Portuguese one also had a word sounding like "portuguese" in it that made it trivial for me.

I'm no graphics expert, but I'm pretty sure premul is not just an optimization technique -- it's a fundamentally different approach at alpha blending, and it makes some new things possible and some behavior more correct. See http://blogs.msdn.com/b/shawnhar/archive/2009/11/06/premulti... . Sure, if you don't need the extra features premul provides (like the built-in ability for additive blending), then you can do some clever computations like IE apparently does to get the same result, but it's much simpler just to do things with premul.

Actually, check this out: http://blog.joa-ebert.com/2011/11/11/project-hiddenwood/

Audiotool has been working on using only HTML5 and JavaScript (although who knows whether they'll ever start using that in their real product).

40/75 (53.33%)

Wow, there are so many languages I still need to try.... I'm proud I got ones like Piet, Io, and Racket, though.