venning

I'm sure that's why this article is listed here today.

This comment on Cap'n Proto came one hour before this article was re-submitted: https://news.ycombinator.com/item?id=12471541

Correct me if I'm wrong, but some of this sounds like blitting, except optimized for the in-memory structure and not the on-disk structure.

In 2008, Joel Spolsky wrote about 1990s-era Excel file formats and how they used this technique to deal with how slow computers were then [1]. Same technique, new problem set.

[1] http://www.joelonsoftware.com/items/2008/02/19.html

It's likely a usage-related calculation. If an intersection is known to nearly always have cars but rarely pedestrians, a walk signal makes sense.

In Washington, DC, most intersections do not have a traffic button, so pedestrians and cars are equally accounted for in the signaling. Only a handful of intersections, where pedestrians are less likely, have traffic buttons.

If I know that I'm going to try to compromise a system and access its username/password lists, is there an advantage to creating a number of accounts to which I know the password prior to the break-in?

Does this make it easier to break the other accounts once I have access to them encrypted? As in, I know that the account with username X has an unencrypted password Y, so now I have guideposts to tell if my cracking attempts are pointing in the right direction, trying to get back to password Y from the hashes. I imagine there would be something of an advantage to already knowing, say, 10,000 plaintext-encrypted pairs in a big list.

If this is the case, should one be concerned in managing a system that sees a dramatic uptick in new user registrations as a precursor to an attack?

I'm guessing not. Your comment is my first result in Google search for "keyboard walk" dvorak.

Edit: Ars Technica had a piece a while back on crackers and how they build their lists of probable passwords. I imagine if Dvorak-based passwords became common, they wouldn't last long as "secure" before appearing in those lists.

Edit: The Ars link follows. You'll notice that the password in the story title is a keyboard walk, which the writer confuses as being complex, when it is not: http://arstechnica.com/security/2013/05/how-crackers-make-mi...

This is known as a "keyboard walk" and its variations are one of the more common patterns tested for.

The pneumonia-asthma example seems to be an example of a Simpson's paradox [1]. The doctors acted on a strong (accurate) belief about asthma sufferers contracting pneumonia and acted in such a way that the data obscured an actual causal link (asthma as an aggravating factor to pneumonia). This is opposed to the canonical Simpson's paradox where doctors acted on a strong (inaccurate) belief about severe kidney stones [1a] and again produced lopsided data that hid the best treatment option until the paradox was identified.

Humans have a very hard time uncovering so-called "lurking variables" [2] and identifying such paradoxes. I don't see how a neural network (or other machine learning tool) could do so on their own, but I don't know that much about machine learning. So, I guess I have two questions for the experts out there:

* If all training data is affected by a confounding variable, can a machine learning algorithm identify its existence, or is it limited by only knowing a tainted world?

* Once we have identified such lopsided data and understood its cause, how do you feed that back into your algorithm to correct for it?

---

[1] https://en.wikipedia.org/wiki/Simpson%27s_paradox

[1a] https://en.wikipedia.org/wiki/Simpson%27s_paradox#Kidney_sto...

[2] https://en.wikipedia.org/wiki/Confounding

I know. I've edited my comment to reflect that.

FYI, DanG already detached a similar comment: https://news.ycombinator.com/item?id=12341452

I know: https://techcrunch.com/2016/01/08/y-c-switches-up-its-manage...

Edit: I didn't provide much context. He is Paul Buchheit, a (the?) managing partner at YC Core.

Thank you for your response, Mr. Buchheit. I really appreciate it.

This is the answer I was looking for, one that explains your driving intentions. I'm nothing like a venture capitalist or investor, so I appreciate that I may not have any kind of good read on these companies' potentials.

All the best to each of them and to you.

Not that my opinion matters for anything, but I don't want to call out specific companies as having "low" potential. Now is probably uncertain and stressful for each of them, and I wouldn't want to undermine anyone's confidence as some anonymous armchair investor on their big day. (I know one should not be affected by that, but we're all human.)

To your point, I wonder if there is some sort of calculus within YC: Company A has 1% potential of a $10B exit, Company B has 50% potential of $100MM/year ... what do we do? Cold logic says we invest in Company A over Company B (not that this is zero-sum), but they can only make so many investments and the real numbers are much smaller than one-percent. Whereas, a series of conservative investments keeps them solvent long-term, while the bigger exits slow boil.

That's not quite my question. I'm not referring to the divide between "pure software" businesses and, well, everything else. I'm referring to the divide between $10MM/year gross profit and $1B exit.

Dollar Shave Club had a $1B exit. Is YC willing to fund the former?

Honest question: have the specific funding goals of YC changed to accommodate more pedestrian "lifestyle" businesses? A number of these companies look like they could make viable businesses turning out respectable multi-million-dollar-a-year profits without ever being a target for a billion-dollar acquisition.

(My understanding is that YC specifically targets opportunities that have hundred-million- to billion-dollar potential, so perhaps my underlying assumption is wrong here.)

Does this mean YC is hedging their bets, or are they amenable to funding "low" potential companies now? That would seem to jive with their recent public statements about wanting to expand, since not all viable businesses see huge exits.

I do appreciate stories such as this. I don't draw any conclusions from them, but they are nice data points to have. New technology comes with new performance and safety envelopes, often unknown even to the engineers who design them. Reporting on otherwise unknowable events is helpful to build a picture of what those envelopes are.

Without such reporting, it's unlikely that Tesla would have added the underbody plating when they did. Perhaps this incident is an aberration, or perhaps it is one of multiple incidents where the plating can be pierced and it leads to an upgrade. Either way, it's nice to know.

Regarding ICE cars, Ferrari had a problem with a certain mid-engine model a few generations back where a fuel line would disconnect under hard driving and spray high-pressure fuel on the (hot) engine to obvious result. I don't believe any one was killed, but drivers sometimes reported barely making it out of the car in time. Enough reports of such incidents led to Ferrari investigating and making sufficient changes to prevent the fires.

Gmail did something like this back when it had ads. If it detected an email related to a death, it would attempt to block related ads from displaying above the content, so you didn't get pestered with ads for funeral services while receiving news of a loved one's death. From my understanding, it didn't always work, but was an interesting first step.

I don't know if other Google services now incorporate similar behavior.

(This is the earliest reference I can find to the behavior: http://staugustine.com/stories/060504/bus_2366168.shtml)

> "It's low risk, low cost," National Institutes of Health dentist Tim Iafolla told CNBC.

Isn't risk something that should be determined by study? I can think of a couple ways that flossing may actually damage a mouth.

RSI stands for "repetitive strain injury": https://en.wikipedia.org/wiki/Repetitive_strain_injury

(for those of us who didn't just know this)

I imagine Basecamp would consider their employees to be the future of their business, so these benefits are reinvestments.

This might be a better link, which includes a brief write-up of the work: http://www.stefankrause.net/wp/?p=301

Considering the machine was built in the US in 1965, it's not surprising that the machine used imperial units. Based on Wikipedia's table [1], it looks like metrication had not yet begun in a large number of countries by 1965.

Also, if your goal is to build a device that calibrates other machines, and most of the machines in your country (and, likely, most of the machines belonging to the majority of your trading partners) use one system of measurement, then that's probably the system of measurement you build the machine to calibrate against.

[1] https://en.wikipedia.org/wiki/Metrication#Conversion_process

> Most of all, the thing I'd be intersted in is how resistant to error is the design?

I feel like this is what keeps the most impressive research out of real world application. Yeah, manufacturing is expensive, but tolerances are just so much more controllable in a lab.

Manufacturing also produces revenue, whereas funding the years of research necessary just to figure out a new nanometer-precise process is a hard investment for most companies to swallow, especially if the end product isn't guaranteed to produce a profit itself.

Following up, it looks like Mahle has actually been tying to address the NOx production. The OP links to this article [1] from 2010, which is mostly taken from three SAE papers that indicate they have been able to constrain the NOx in jet injection engines to "single digit ppm engine out NOx emissions" [2]. The EPA seems to indicate [3] that less than 1 ppm NOx post-catalytic converter is targeted. I don't know if their advancements allow for three-way catalytic converters on a lean burn engine, but that would be pretty impressive.

[1] http://www.greencarcongress.com/2010/10/tji-20101027.html

[2] http://papers.sae.org/2010-01-2196/

[3, PDF] https://www3.epa.gov/ttncatc1/dir1/fnoxdoc.pdf

The biggest problem with road cars using valve springs (or indeed engines) that run at 20,000 RPM isn't technological; the issue is practical. Valve springs that function properly at 20k are going to be strong and produce serious resistance at low engine speeds (as a percentage of total power output). So much so that F1 cars with idle speeds of 7,000 RPM are normal. Low RPMs is where most road cars live. F1 cars don't want to be running at low engine speeds since that's not where the power is. Sacrificing power production at the low end (due to the power consumed working the valves) is an easy trade-off for a racecar.

Proper valve spring design is hard: it's a balancing act between power consumption at the low end and power production at the high end. F1 engines have a (relatively) narrow operating band, whereas a road car is expected to be usable from (a low) idle to redline.

(Expense is less of an issue here. Multi-million-dollar supercars with 20,000 RPM redlines would sell well.)

The plasma jet injection thing sounds pretty novel, but I imagine a lot of the efficiency gain is tied up in the lean burn. The brake thermal efficiency gains of lean burn are well known, as are the reasons we don't use them in modern road cars.

Burning so little fuel in so much air produces high levels of nitrogen oxides [1], bad for smog and acid rain and ozone. That's not a concern for an F1 car, but a serious issue for road cars that need to pass emissions tests since they can cause modern catalytic converters to stop functioning properly. See this note [2] on how unhelpful current NOx mitigation strategies have been.

I believe the addition of three-way catalytic converters was partly responsible for the stagnant fleet fuel efficiency somewhere in the 1980s, since they had to make engines run richer than before to not foul the cats; though, we've since developed better converters. (Also, safety equipment and luxury things increasing weight affected fleet numbers.)

[1] https://en.m.wikipedia.org/wiki/NOx

[2] https://en.m.wikipedia.org/wiki/NOx_adsorber#Market_use

A shorter, possibly easier to understand writeup from six months ago: https://www.mikeash.com/pyblog/friday-qa-2015-11-20-covarian...

I believe it was written by an HN commenter, and accumulated 67 comments: https://news.ycombinator.com/item?id=10606355

Damn Interesting has a piece on the race to produce the first liquid hydrogen over a century ago:

http://www.damninteresting.com/absolute-zero-is-0k/

From the article: "He did so with the full awareness that he was among the last of a disappearing breed—the “classical physicists” who had the luxury of simply banging on matter until it did interesting Newtonian things."

I assume that Uber is not competing with black market interest in bugs, rather they are competing against interest in other bug bounty programs.

The number of people who earnestly compare rates between bounty programs and black market buyers before deciding who to inform must be vanishingly small. However, the number of people who compare bounty rewards before deciding who to test against is likely much higher.

To your point, bombs are not chemistry, bombs are the application of chemistry. Encryption is not math, encryption is the application of math.

I don't see a support feature on the site, but since this is a YC company, I figure I can list an issue here:

@whois google.com is not working. It just replies: "@whois: Please enter a valid domain name. Example: @whois google.com" Same for other domains.

Edit: This only occurs over SMS, not through the site. My guess is that my carrier (Fi) is doing something to munge the domain name that Prompt needs to parse correctly.

If the phrase was "1.6x speedup" then I would agree and not have commented. But the phrase is "1.6x faster", which is a bit ambiguous.

That there is disagreement among the commentariat both ways is indicative of ambiguity. (Though, I am assuming no trolling here.)

While not exactly analogous, try replacing it with a percentage and re-evaluate: What does "60% faster" mean and what does "160% faster" mean?