Lidars use pulsed lasers with peak powers up to the kW range.

Yes but failure cases are still possible. A stuck mirror or frozen phased array can result in a continuous beam.

Engineers should design in a way where the worst case possible is the assumption not the exception

Silicon carbide fiber reinforced silicon carbide is also being increasingly used.

In production?

I think eutectic is referring to the ceramic matrix composites (CMC) used in the General Electric's engine LEAP. Here's some quotes from [1]:

  > The engine has one CMC component, a turbine shroud lining its hottest zone, so it can operate at up to 2400 F. The CMC needs less cooling air than nickel-based super-alloys and is part of a suite of technologies that contribute to 15 percent fuel savings for LEAP over its predecessor, the CFM 56 engine.

  > GE’s CMC is made of silicon carbide (SiC) ceramic fibers (containing silicon and carbon in equal amounts) coated with a proprietary material containing boron nitride. The coated fibers are shaped into a “preform” that is embedded in SiC containing 10–15 percent silicon.

[1] https://www.ornl.gov/news/ceramic-matrix-composites-take-fli...

From what I understand, shroud linings don't rotate, though. They are fixed to the engine casing. So they are not subject to the high centrifugal force that would make creep really problematic.

While you are right about the limited applications for this material, the reason cannot be creep, which should be negligible in this kind of ceramic even at the working temperature. Certainly it must be better regarding creep than the alternative metallic alloys.

In a rotating part, subject to high centrifugal forces and vibrations and shocks, I think that the risk of unpredictable fractures may be too high for a ceramic, even a composite one.

Silicon carbide ceramic has low toughness. A composite should be better, but still far from metallic alloys.

I have seen mentions of research about the feasibility of using silicon carbide composite ceramics for rotating parts, with the goal of reducing their mass and increasing their working temperature, in comparison with metallic parts, but it is unlikely that this has reached the stage of being used in production engines.

Ceramics, e.g. derivatives of zirconia, are frequently used for turbine blades, but only as ceramic thermal barrier coatings on metallic blades, not for the body of the blades.

This is why I love HN

But 'i' is not a car, it is the index of a car.

I like einsum, it's concise and self-explanatory. Way better than multiple lines of nested for loops would be.

I agree that nested loops are verbose. But einsum is unstructured and not extensible, hard to debug in pieces, hard to document etc.,

I got to level 8 by asking in rot13. I think I beat the bonus level too but I can't remember how.

Also, the problem is to find a global min-cut, which requires all-pairs max flow.

Not actually all-pairs max flow, you can fix the source and consider all possible sinks. In the AoC problem we also know that the min-cut is 3, so we can abort the flow algorithm as soon as we have found a 4-flow.

Different models have different inductive biases. There is no way you could build GPT4 with decision trees.

Is it not possible to compensate for inaccuracy, if you have a sufficiently precise measurement at the hand?

Here's a variant which gives 64-bit output:

  uint64_t x = 0, w = 1;
  uint64_t msws64(void) {
      x = x * x + (w *= 0xe9acc0f334e93bd5ULL);
      return (x = (x >> 32) | (x << 32)) ^ w;
  }

I wouldn't recemmend that without a 128 bit variables. The truncation is required to make the generator non predictable, otherwise you leak to much state.

Alternatively, you can just call the 32 bit one twice and build a 64 bit value from the results.

Edit: I didn't see that you changed the algorithm more than removing the truncation. It is honestly suprizingly good for exposing that much state, but it fails PractRand after 16 GB.

It is honestly suprizingly good and hasn't failed PractRand yet (I'm at >64 GB).

Your edit is confusing me. Which variant fails and which variant passes?

Your msws64 did not seem to fail

What if you don't truncate? I fell like it might still be OK.