tlb

Did someone publicly predict in the 80s or 90s that Welch would kill GE?

> If nuclear made sense, Microsoft ... would finance the construction of a few plants to feed their ever growing appetite for reliable carbon-free energy

Companies focus on their strengths. Microsoft isn't building its own chip fabs, or lots of other things it has an appetite for, because it's not one of their strengths. Companies can only be good at a few things. So them not building their own nuclear plants isn't a reason to believe it's a bad idea.

There are a few startups building small modular reactors. As I understand it, while data centers are certainly on their list of potential customers, they're getting more immediate sales commitments from remote mining operations where they're replacing diesel generators instead of competing with the grid.

Suppose they’ve estimated the risk of failure (and killing the astronauts) as 10%. Then

- returning Starliner empty is the correct decision

- a successful return doesn’t indicate they were wrong about the risk

- the experts will learn a lot more from poring over mission data

Iron age problems. Wasn’t a problem in the good ‘ol bronze age.

I don't get how that limits the packet size. If a sender's clock is 500 ppm faster than an intermediate node's, you need 500 ppm of slack. That could be short packets with a short gap, or large packets with a large gap.

Ethernet specs the IPG as a fixed number of bits, but it could easily be proportional to the size of the previous packet.

The treatment that makes people feel respected varies between cultures and genders, but the emotion is the same.

In general, it's useful to disentangle your emotions from what you think is causing them. For example "I feel criticized" doesn't name an emotion. It names one cause, but emotions are far from monocausal. Being able to articulate what actual emotions you experience is helpful in managing them.

The question is "Did you experience the following feelings during a lot of the day Yesterday? How about anger?" So it represents a pretty bad day. In my life, I think this would be true for less than 1% of days.

Inverters may be protected against changing settings, but if you can replace the firmware it can likely cause permanent hardware damage. Which the manufacturer, perhaps under pressure from its government, can do.

Travel speed matters too. At non-peak times, when queues are short, going 100 kph will give much shorter commute times than moving sidewalks (5 kph) or club cars (25 kph).

I'm not an expert here. But according to https://en.wikipedia.org/wiki/Diffraction-limited_system (see last paragraph of the intro) the diameter of a laser beam doesn't impose a diffraction limit on the optics it goes through. You still get the angular resolution you'd expect given the entire aperture of the lens even though the beam is only going through a small area of it.

Why don't welding/cutting lasers add more divergence with built-in optics? Would it hurt performance? It seems like you could add 1 mrad and it would hardly make a difference at the usual working distance, but spread out to a meter over 1 km, so you can't zap people across town.

“Profit margin” doesn’t subtract the initial development costs, just current costs. So drug profit margins should be large to repay the billions previously spent in development.

Most Bakelite products were a molded combination of resin and sawdust. Though there were also pure resin products.

Large for the old days, but a billion names is no longer a big file compared to a AAA game.

They could also just drop a device with a microphone in your bag and listen to both sides of every conversation. Involving bluetooth seems like an extra complication.

Though I guess you could save even more power by using transistors.

Is this known to work? I would have thought that running a whole tube at high temperature would cause the grid to emit thermal electrons, creating a current between grid and plate. Even though the grid doesn't have the low-work-function coating that the cathode does, its surface area can be larger and it's usually more negative than the cathode so I'd expect substantial currents.

If this does work, you could save a lot of power by building heated, insulated enclosures for entire tube appliances rather than heating each cathode and cooling every tube.

It would have to be able to fix your records in the IRS database, not just give you advice from the FAQ like most LLM support bots. Which could be awesome, but it'd have to be robust against prompt injection attacks and other bamboozlement.

If the error is the sum of many little errors, as it often is in mechanical assemblies, it's approximately normal due to the central limit theorem.

Many sanity checks are preserved with -O0, so they're still useful in debugging. And they serve as documentation of expected conditions. I often put in checks that I expect to be statically eliminated to remind myself what is supposed to be true.

Yes, in this case the title without the "how" is misleading, since it suggests they actually "handle" it and not just do whatever. So I think this title passes the howey test. Edited.

Clang has a mode (UBsan) that flags and optionally aborts if you execute UB. It runs several times slower because it has to check lots of things, but it might make sense for some programs.

Some of Reddit's early users were sock puppet accounts run by the founders, which are different than bots. The sock puppet accounts were, in fact, interesting to engage with since they were run by interesting & literate people. Since presumably a conversation is what you came for, you got what you wanted. Quite different from talking to a bot on a dating site where you came to initiate an IRL interaction which could never happen.

I sleep in a light room wearing a mask. So my eyes are dark, but skin gets some light. My understanding is that melatonin is regulated by light through the eyes, so that should count as dark for circadian rhythm purposes. But I’d appreciate any corrections.

Seems like this study measured it with a wristwatch, which is a poor proxy for either light on the eyes or skin.

Why wouldn't you ask for liability protection? No matter how safe, like even if the injection was actually pure water, thousands of people would be convinced they were injured and sue for millions which, even if you won every case, costs a fortune in legal fees.

No vaccine is perfect. They work by showing your immune system some example germ features, and most people's immune systems learn to defend against them. But immune systems are extremely complex and variable between people, and you can never guarantee a particular response.

Vaccines sometimes completely eliminate some diseases (smallpox is one). They do this not by giving every person 100% protection individually, but by reducing transmission enough that the germ dies out.

They do it because it makes more money. They've measured the amount of time spent (and ads served) to people with and without the feature, and they found that, averaged across the user base, the feature increases revenue. So everyone gets it.

At least, that's what every ad-supported platform thinks they're doing. There can be other factors, especially in large process-oriented companies like Google. Like, an ambitious PM wants to be able to point to a feature they added in their annual performance review, so they do some sloppy measurements to show that their dumb feature increases revenue even if the change is just noise. You can see why such a PM might not want there to be a way for people to disable the feature, since that metric could clearly show how unpopular it is.

Investors won't be interested in a one-shot project. They'd be interested if the company can grow big, either by selling systems to other cities, or expanding into related businesses. So figure out whether or not that's possible.

Most cities are terrible customers. They make decisions slowly and often impose painful requirements on suppliers. They're wary of anything that could fail publicly, because the media loves to show failures and blame it on the city government.

Another way that cities are bad customers is by not actually using things they buy. They might buy your product and then not get around to hiring anyone to operate it. Even though you got paid, you won't be able to iterate and improve the product.

City purchases are generally public records, so you should look for similar products the city has bought and go talk to people in the company that made it for advice and connections.

Yes. The good news, if you're trying to develop tech to solve the housing cost problem, is that there are several labor-intensive processes that are easier to develop automation for than building the structure of the house. Like, a robot to wire up light switches is much cheaper to develop and test than a robot to print entire structures.

Email apply@ycombinator.com for advice.

The cost of the structure of a modern wood house is a tiny fraction of the overall cost. Most of the construction cost is in the wiring, plumbing, bathroom & kitchen fixtures, windows, flooring, etc. 3D printing doesn't help with any of that. In fact, wiring and plumbing is harder than in a wood-framed building, where you can drill holes in a few seconds.

Also, aesthetic zoning rules are super-conservative. You can't build an unusual looking house in most cities.