michwill

What I really look forward with BECs (although simple atomic interferometers based on ultracold atoms would do, too) is super-sensitive magnetic field detection.

One can detect magnetic fields from a shielded laptop, even many meters away, and figure out what's happening inside (e.g. leaking secrets etc).

I think, not a terribly difficult experiment, but nobody has performed it: need ultracool physicists and hackers in one lab room

> you mostly can’t buy stuff with it

Sometimes I cannot buy legitimate stuff without them. Just today, I bought some books for which US credit/debit cards refused to work. And I see that every week.

> you can’t pay taxes with them

https://techcrunch.com/2018/11/25/ohio-becomes-the-first-sta...

Sorry, but this looks like a completely uneducated opinion.

There were many attempts to create virtual money which cannot be shit down, and all of them dated before Bitcoin failed due to government intervention. Bitcoin solved that, and the simplest solution was the wasteful proof of work.

Of course, when someone is using the same principle for something centrally controlled, which can be done by a database, that is a different story.

Well, this doesn't look right to me. Imagine that the moon is actually a filter at the path of the sunlight.

Sun's temperature is 6000 K. Moon's surface is pretty black: it reflects only 12% of the light.

So, effective temperature of the Sun reflected by Moon, considering that thermal radiation is proportional to T^4, is 6000 * .12 ^ (1/4) ~= 3500 K. That's quite enough to light up some fire! Of course, the spectral composition of the light will be not thermal etc, but the estimate should be close enough.

Why doesn't the Moon itself heat up like that? Well, the rocks on Earth don't heat up to 6000 K either... I think, it's partly that they are "not surrounded by sun", partly that the Moon is a giant cold heatsink

Meanwhile, Elon Musk acquired more than 50% of the space launch market (while 5 years ago it was Russia who owned it).

Don't troll, buidl!

Reg A+ is indeed easier than an IPO. I've seen some companies doing that, and it's 1 year pushback, to say at least.

One year is an insanely long time these days!

I don't think you're right. Under this impression, most of cryptocurrencies are securities. However, SEC said, they're not: https://twitter.com/mikebutcher/status/940487575491497985

You're not correct. I think, less than 50% of ICOs are not securities. But the way to prove is exactly this: ask a lawyer

All right, but if the coin is inseparable part of the resource, it cannot be a security. There is a thin line here.

Imagine a coin securitizing Bitcoin miners ("cloud mining coin"). Yes, that's a security.

Now imagine Ethereum. It was ICO'd, talking in today's terms. Is it a security? (hint: SEC just admitted it's not)

There are two things here:

* It's not OK to raise 200M with no team and product;

* Compliance only creates obstructions, not protection.

What's even more surprising is the position of Coinbase:

"Given how useful it is for criminals to have financial accounts that are seemingly unconnected to them, I view this feature of the legislation as an important next step in the fight against global money laundering."

Ms. Kathryn Haun Rodriguez of Coinbase

Oh wow! I didn't ask for the last one, but this is insanely interesting also!

I would apply it instead of garbled circuits here https://eprint.iacr.org/2016/591.

Although slower, it feels like TFHE would provide better security against an active adversary. So, at least, simple server-side encrypted queries would well be possible

Yes, I understand. I was thinking about:

a) smart contracts controlling something within the encrypted data based on publicly available data;

b) encrypted key-value store where you traverse the tree structure based on encrypted query and encrypted tree buckets, but get a publicly available result about which bucket is next (similarly to how it was done in Arx paper using garbled circuits).

That's a seriously cool thing to have in the toolbox!

Does it produce only encrypted output, or can it optionally produce unencrypted results also? Can it optionally use public data as an input?

Also I am guessing if it could be accelerated on GPUs. I worked with a guy who accelerated a standard FFT on CUDA 100..1000 times for scientific computations (and later NVidia copied his code, lol). I wonder if something similar can be done here

I think so. It is similar to CloudHSM, but you don't have to trust Amazon!

It basically allows some untrusted party to convert ciphertext which you can decrypt to a ciphertext which someone you granted access to can decrypt.

It's like "instead of decrypting my data and encrypting with Bob's public key I want Charlie to do this operation without showing my secrets to Charlie". Charlie is a miner in our case :-)

I think, the biggest latency hit will be from re-encrypting a symmetric key which encrypts messages, once in a while. This will result in occasional delays of ~0.25 ms (that's how much one operation with elliptic curve crypto takes).

The other bottleneck will be from the block cipher, when you actually encrypt/decrypt messages. Our open source version currently doesn't use AES-NI for that, so limited by performance of unaccelerated AES256 (few hundred k messages per second). This impact can be made negligible when using AES-NI (will come soon).

When using granular encryption, the performance bottleneck will shift to parsing messages (avro, for example), and will be limited by the performance of the parser.

Congrats with the launch!

Python doesn't force you to use IDEs. There are good command-line tools (and editors such as sublime, vim and emacs can jump to definitions etc).

There are "IDE guys" and "text editor guys". You can just determine who you are and be productive there!

Best text editors are sublime, vim, emacs; Best IDE is probably PyCharm.

It's much less of a choice when you code in Java for example: way better to use IDEs.

Same here.

Buying a windows device specifically for the VR toy? No ability to use VR for real work on non-Windows machine (if possible)? No thanks

I think, you can't simply push off zero-point fluctuations in vacuum w/o creating particles occasionally.

And if you admit you do, you're at level of photon thrusters at best.

Oh yeah, I wasn't clear, yes.

Not "everything is clear, the case is closed" - that's what I meant

Actually it's not as much useful as an engine as it may seem, at least at current parameters. Ion thrusters are pretty good as low-power engines. And it's not clear that emdrive is better than that even if it works. Maybe with powerful space-based nuclear reactors

The trick is that this is not a closed discussion. The thrust was suspiciously repeated in many experiments, and question is whether it's some fundamental flaw with experiments, or it is real.

I guess, the more info is available to the public, the better. Hopefully someone will test it properly. But no way it can be considered as "provably works"

Interesting what are the fundamental reasons for the thrust. It's certainly not throwing photons. Hypotheses:

* Flawed study. It actually throws something (like, cavity material slowly ablating) or repels of Eddy currents in surroundings;

* It repels of Earth (and the force has either gravitational or magnetic nature); then it's not suitable for deep space propulsion

* A sort of ether actually exists, and it throws that. Though, then there would be some anisotropy showing ether wind; in this case it is suitable to be used in deep space