scott00

The key difference seems to be German private health insurance contracts are long-term affairs. Multi-year, or perhaps even lifetime? US health insurance contracts are typically a year at a time. So German companies have to reserve for costs projected to occur far in the future because they are liable for them, while US companies have no idea if their customer will still be around in 20 years.

My guess would be there's a healthy dollop of regulation pushing the German insurance market into that shape, otherwise you would probably see short-term insurers outcompeting long-term insurers since they wouldn't have to do old-age reserves and could therefore charge lower premiums. Consumers tend not to be nearly as good at rationally planning for long term expenditures as are actuaries.

It mentions reduced spending on doctors. My thinking on the mechanism is that freeing doctors from noncompetes makes it easier for them to leave big practices and start small ones. Small practices have less bargaining power with insurance companies and will have to charge lower rates.

Doesn't make intuitive sense to me why you'd need a whole new airplane for this. Can you not rig something up to let one of these blades ride on top of an existing plane like the space shuttle transporter? Or would Sergey Brin's giant dirigible work? Or two helicopters flying in formation? Or just make the blades come in two pieces assembled on site?

Just seems very hard to believe a project as massive as a whole new airplane is the best solution to this problem.

I was curious on some of the details, so I did a little digging.

The town in question (Salisbury, MA) is built on a narrow strip of land between a marsh and the ocean. Mostly between 2 and 5 houses wide. Doesn't seem like you need a geology PhD to determine you're going to have some erosion problems here.

As best I could determine, the beach replenishment in question was from access points 5 - 11, which covers 1.6 miles and about 150 homes. If I'm right on that, they put down enough sand to extend the beach 3-7 feet[0]. So the first thing to note is that this was a very small beach replenishment project. The senator is probably right that they should go bigger next time.

The cost per home for that would have come out to about $3333.33. Honestly, even if you triple it and do it every year, I don't find that to be an unreasonable expense to impose on owners of $1-5 million houses built on a sand dune. These guys need to quit whining and raise their _local_ taxes the relatively modest amount necessary to preserve their town. Something like $5k/year for beachfront and $1k/year for the rest would get the job done.

[0] This is assuming a constant slope to the beach. The low end is extending at 3 feet above sea level, the high end is extending at 6 feet above sea level.

For the benefit of readers, here's the quote from the article indicating no controls:

"Factors that may also play a role in health, outside of daily duration of eating and cause of death, were not included in the analysis."

Contractors pay self employment taxes in lieu of the social security and Medicare payroll tax, and as a result are eligible for both.

The health insurance situation is not great, but it's available through Obamacare. In my experience the actual cost is not much worse than employer provided, but the true cost is often subsidized, so employees don't always realize how much salary they are giving up for health insurance.

No eligibility for unemployment or workers comp though.

'motivate customers to visit' - Some people like cheap stuff. Some amount of people who are not willing to pay x for a meal at noon might be willing to pay 0.9x at 2 pm. Other people have money and tight and inflexible schedules. Some of them may be more willing to visit at noon and pay 1.1x than they would be to visit at noon, wait in a 10 minute line, and pay x.

'enhance customer and crew experience' - Neither customers nor crew like it when the restaurant is busy enough that the line gets long. By making it more expensive to eat at peak times and less expensive to eat at off peak times, they think they can smooth out the demand schedule. Of course whether that's a net positive to any given consumer depends on their relative preferences on meal time, wait time, and meal cost. But the potential is there at least. On the crew side, a smoother demand schedule means they can either schedule fewer people on longer shifts, or if they keep schedules the same reduce the amount of "crunch time" during each shift.

I think what this work does is establish a new, and lower, upper bound on the number of points that need to be explored in order to find an exact solution.

From some of your other replies it looks to me like you're confusing that with an improved bound on the value of the solution itself.

It's a little unclear to me whether this is even a new solution algorithm, or just a better bound on the run time of an existing algorithm.

I will say I agree with you that I don't buy the reason given for the lack of practical impact. If there was a breakthrough in practical solver performance people would migrate to a new solver over time. There's either no practical impact of this work, or the follow on work to turn the mathematical insights here into a working solver just haven't been done yet.

Yeah, the discretization interacts with the oscillation for sure. Full implicit is better than CN with regards to oscillation for instance, but I don't think would be a net win. Running a few implicit steps before switching to CN might help, though I've never tried it.

Is that method extensible to discrete dividends?

Crank-Nicolson is probably the least objectionable part of the method, but I prefer ADE.

There are two numerically painful parts of the problem: the advection term and the oscillation inducing terminal condition (because it has a discontiuous derivative). I like to deal with advection by transforming the equation to an advection free equation. I'm under NDA on the best solution to the oscillatory terminal condition so I can't give that one away unfortunately.

This method will work but will require a large grid and consequently be quite slow. And order of magnitude or two faster than this is possible if you are clever.

Dark pool trades are reported to the FINRA TRF within at most 10 seconds and appear on the consolidated market data feed.

If you are genuinely in the market for multiple racks of servers you (a) know how much a rack of hp/dell gear costs, which gets you within an order of magnitude of what this is going to cost, and (b) would not buy one of these without a sales call even if you could.

Drivers doing nothing productive are expensive, but drivers working ride hail are somewhere in the range of cheap to slightly profitable.

The comparison with Uber depends strongly on how similar the dispatch profiles are, and I would not be quick to assume that they are similar. If they are limiting the self-driving cars due to weather, or time of day, or any property of trip type it could easily have a substantial impact.

The existing studies have decent metrics given the sample sizes, IMO, I would use something similar. For the Cruise study that was recently released it was collisions, with sub-analyses for collisions with stationary objects, low speed collisions, and high speed collisions.

I believe that you are correct that scale is too low for analyzing injuries or deaths.

Been a flurry of self driving car safety news lately. Both Cruise and Waymo are reaching for more sophisticated comparison samples than nationwide stats to demonstrate the safety of their systems.

But the obvious way to prove this, to me anyway, is to run a randomized controlled trial. Put them in a dispatch system with human driven cars, randomize whether any given assignment goes to a human or a robot, and you've got the statistical gold standard.

Anybody understand why they're not doing this?

I tried to figure out what the economics on this looked like from the Wilson, NC Mass Transit Fund budget: https://www.wilsonnc.org/home/showpublisheddocument/5592/638...

The numbers were confusing. Among them, for 2021-2022:

Ridership: 156,904

Revenue Mile: 315,409

Total expenses: $2,022,634

Fares collected: $3,255

The total fares and ridership from the budget do not seem consisten with the unit fares discussed in the article.

My best guess here is that the fares discussed in the article aren't showing up in the financial statements, and the ones in the statements are for another service entirely, maybe disabled transit.

Across their entire system, they're spending $12.89/ride. I'm lacking data for this assertion, but my gut is the non-microtransit stuff is fairly negligible here. Even if we knock it down to $10/ride we're still talking an 80% subsidy.

But 2019-2020, before the switch to microtransit they were at $28.03/ride. So they do seem to be picking up substantial operating efficiency, even if it's still heavily subsidized.

Covid's probably skewing that substatially...2018-2019 only $15.15 subsidy per ride. So still some improvements but not that amazing. And they were collecting $0.73/ride of revenue, so that subsidy rate was 95%.

Depends on your definition of a good password.

The experience of one victim mentioned suggests it's economical to crack at least 50 bit entropy passwords for at most a $3.5 mil payout.

50 bits isn't great, but it's not terrible either.

The elephants in the economic incentive room are:

(1) Taxation of environmental externalities caused by fossil-fuel driven transportation are nowhere near the socially optimal level.

(2) Fossil fuel production is massively subsidized, even beyond the lack of taxation offsetting environmental externalities.

They do not literally mean that money collected from ratepayers can't be used to pay political expenses.

I'm not an expert in this, but roughly speaking the way utility rate regulation works is that rates are set at a level that gives the utility a certain return on capital. (roc = profit/equity = (revenue - expenses) / equity).

What they probably want is for political spending to be excluded from the expenses section of that formula for the purposes of the rate setting calculation. (They would still appear there in GAAP financial statements.)

So the utility would pay those expenses by accepting reduced profits.

It's not 600k per year per person. It's 606k additional spending since 2016, per additional sheltered individual since 2016.

From what I can tell, SF had about 18,548 shelter beds in 2022[0], which puts it at $36,230/bed/year. In 2016, there were 11,017 beds[1], which gets you $20,332/bed/year.

And of course not all help for homeless people is just paying for housing. Still seems expensive, but not nearly as bad as your math indicates.

[0] https://files.hudexchange.info/reports/published/CoC_HIC_CoC...

[1] https://files.hudexchange.info/reports/published/CoC_HIC_CoC...

TP-Link enterprise wireless access points (the EAP line) don't support IPv6.

Messagease is an on-screen phone keyboard optimized for single finger use:

https://www.exideas.com/ME/index.php

I get about 30 words per minute with it.

Plaid isn't a solution to a technical problem. It's a way to deal with the fact that banks don't want their customers to bypass their websites/apps and the cross-selling ads within.

I don't get the security benefits of this device over any other ARM computer. It seems like a complicated enough device you'd need to run full blown linux on it, and it would communicate over BLE and USB. Are those stacks much more secure than the TCP or UDP stacks for some reason? You'd have the benefit of nobody opening random email attachments or visiting sketchy websites, but the same would be true of any device treated like an appliance or server.

You're right, I was not factoring in the effiency of the generation process. 100 W is definitely moderate or higher for most people.

This was the first time I had seen a cycle desk and I find the idea really interesting.

Having done a little more reading on the subject, my sense is that you may have a better handle on the ergonomics than the existing cycle desks on the market. It seems like most people are using cycle desks to do a pretty limited amount and type of work while doing a short workout, whereas you are using it for a long time each day, which I imagine means you can do most types of work comfortably.

If that's true, and it's true for a sizeable number of people, not just you, I think that's what I would focus on.

Are the other people for whom you've built one able to use it as much as you?

Very cool!

I'm curious about the effect this has had on your fitness. 60 W output on a bike is pretty low intensity exercise for most people, but 15-20 hours a week is a lot of volume.

Did you have to work up to that? Where did you start and what did the journey to where you are now look like?

Do you have a sense of how this has affected your capacity for higher intensity exercise? (If it helps to have a more concrete idea of what I mean by this, were I designing an experiment I would probably track max power output and power output at VT1 and VT2. But any kind of fitness metric would be interesting, as would be what you think the effect has been, even if it's not based on hard data.)

A similar problem exists with that analysis. Tesla ownership is probably correlated to the amount of time/miles spent in driving conditions where Autopilot is effective.

You also add a whole bunch of other confounding factors if you're not comparing same-driver accident rates. Accident rates vary by location, city/highway driving mix, driver attributes (age, gender, credit score, etc), and probably many other things that may be correlated with Tesla ownership.

The data analysis is not straightforward. You would expect a driver's decision to use Autopilot to be correlated with the ability of the Autopilot to handle the conditions. So just comparing Autopilot accidents per mile to human accidents per mile won't cut it.