HNHacker News
TopNewBestAskShowJobs

hwillis

6,522 karma · joined March 21, 2012

submissionscomments
hwillis··on Linear, symmetric, self-selecting 14-bit molecular memristors (2023)
> graphene was supposed to be this wonder material, and now it's found in the soles of trail and hiking boots.

I mean, that's not because graphene has become a routine part of our material repertoire. It has no reason to be in those things, does nothing, and is just marketing fuel. We may put "graphene" in things, but we are not much closer to using its interesting properties.

hwillis··on The magic of DC-DC voltage conversion (2023)
You're overcomplicating it; you only need a single clutch and in/out springs[1] to do this. If you're spinning at 4000 rpm and your springs cover 6 degrees of rotation, then your clutch needs to be able to actuate at 4000 Hz.

When the clutch is engaged, the engine-side springs compress to supply the torque and match the speed difference. When it's disengaged, the springs expand back out as it returns to engine speed. The obvious problem is that clutches do not smoothly click on and off like a transistor.

However there are more specialized devices that use stick-slip dynamics like piezo actuators. Since there is a much more rapid transition between "on"/"off", they can be very efficient and allow relatively weak devices to exert very large forces. They're just only able to take very small steps.

[1] Labeled 4 here: https://haynes.com/en-gb/sites/default/files/styles/blog_lan...

hwillis··on The magic of DC-DC voltage conversion (2023)
> If you measure it across a high impedance then it looks like a voltage spike. If you measure it across a low impedance then it's not necessarily much of a spike.

"Disconnect" implies an open circuit and high impedance.

hwillis··on How to build a 50k ton forging press
It's plenty real, and it matters for more than just strength. Cold rolled grain oriented electrical steel has better magnetic properties than non-oriented steel and is used in some applications where the field is in a straight line.
hwillis··on How to build a 50k ton forging press
> Can you forge metals in a highly controlled and directed magnetic field where you can orient the grain/alignment of atoms/fields in whatever direction you want.

This is how you make magnets. "Soft" ferromagnets have small, round grains that rotate to reinforce outside fields. "Hard" ferromagnets have permanent fields of their own and long grains that can't reorient.

Forging with a field has a very low impact on the material properties because of how weak a magnetic field is compared to the forces moving atoms- same reason steel loses its magnetic properties when it gets hot.

> what happens when you make damascus from varying plated that have particular alignments/grains

"Damascene" is the layered look most often made from acid etching sandwiched and forge-welded layers of different steels. Damascus is a single alloy for which the pattern is named.

Since in both cases the material is melted together, it's far too hot for any magnetic properties to have any impact.

hwillis··on How to build a 50k ton forging press
annealing, which resets the grain structure, happens at a lower temperature than melting or sintering.
hwillis··on How to build a 50k ton forging press
The second moment of area is quadratic with thickness (t^2), so 250/8" plate. But in reality 2-3x more than that, because at that scale steel gets a bit... goopy.
hwillis··on How to build a 50k ton forging press
> Forgings have the added advantage of variable grain direction which generally can be tailored to the stress patterns of a specific design.

This is a super underappreciated fact! It's often repeated that forging is just stronger, but just squishing steel does NOT make it stronger. Forging a part is so much more than just smashing it into a shape.

Steel cable is made of pretty ordinary steel which is stretched 100s of times its original length. That process alone makes it 2-4x stronger in that direction. You stretch steel and it gets stronger in that direction.

Do you see how complicated that optimization process becomes? The process steps are not just trying to take it to the final shape. Your piston rod needs to be strong lengthwise, so you actually want to start with a short fat ingot and stretch it out instead of one that is near-final size.

Think of making an I-beam. You could hammer out the middle, making it thinner. That would give you a bit of strength there but very little on the edges. If you instead pull the edges out, you create a long continuous stretch that will be very strong against bending. Where, how, and in what order you stretch makes all the difference. You may want to leave extra material and cut it off later, so that your grains are all oriented together instead of tapering to a point.

For any moderately complex part, this process is as complicated as modern engineering problems. With poor steel you genuinely need to understand how to foster and bring out those continuous lines or your corkscrew will unwind like playdough. Blacksmiths had a legitimately intellectual job back in the day!

hwillis··on How to build a 50k ton forging press
> like cast iron doesn’t have fantastic high-performance material qualities

Cast iron is a material, not a process. It's an unfortunate legacy term for very high carbon steel (>2% by weight, or <11 iron atoms per carbon atom). For reference "standard" steel is ~.08-.18% carbon, and high-carbon steel is ~.8% carbon.

The >1% carbon precipitates out into graphite within the steel, causing it to behave totally differently. Less rusting, but weaker and much more brittle when solid. Less viscous when liquid, so you can cast long and thin parts.

> Are there different, higher quality casting processes, and I’ve only seen the bargain bin results?

There are, but it mostly is independent of the material. Some turbine blades are cast as single crystals for heat stability; you can't really cut a single crystal without introducing cracks and issues. There's also vacuum casting and spin casting (using a centrifuge to force liquid into the mold), which lets you cast metals that react with air or cool too quickly for normal methods.

Most of the variation in process is about the final form you cast into, though. Engine blocks are sometimes cast into a one-off ceramic shell that is sprayed onto a sand form. It's an expensive process but it lets you do the whole thing in one step.

> is it all very complicated and material specific?

It is very material specific. Fundamentally its all about shaping the grains. In many steels you can physically alter grains. In others, like precipitation grains (aluminum alloys, some steels) the structure is determined by the cooling and you can't physically shape them. In that case you may often get a better structure by casting since you can choose how to cool parts down, while a billet will have a homogenous structure that is usually worse towards the center.

hwillis··on Federal appeals court finds geofence warrants “categorically” unconstitutional
I wanted something slightly more realistic. It works just as well if the garage is an apartment building where everyone keeps their apartment unlocked.
hwillis··on Federal appeals court finds geofence warrants “categorically” unconstitutional
Search warrants are usually public record after they are unsealed. If they aren't included in evidence its a Brady violation and if caught the case is essentially dead. If the defense sees that the search warrant was given or executed first, they probably have a very strong argument that all the evidence is tainted. Parallel construction doesn't work very well when proof of tainted evidence is publicly available.
hwillis··on Rodney Brooks’s Three Laws of Robotics
Human fingertips and skin in general are extremely precise high resolution pressure and vibration sensors that robots almost completely lack. Tactile feedback takes up the second most human sensor bandwidth behind vision. Imagine doing anything without it- interacting with the world through a puppet.

The joints in a human arm are direct-drive series elastic actuators with pretty incredible sensitivity. If you were holding a big heavy box and I bounced a ping pong ball off it, you'd probably feel the vibration. That's kind of a big deal. You may not be able to measure absolute weight well, or even relative weight, but you can detect quick changes with amazing precision. Lifting any heavy or large object is inherently wobbly even for a robot, and that sensitivity is important.

hwillis··on Federal appeals court finds geofence warrants “categorically” unconstitutional
https://www.law.cornell.edu/wex/fruit_of_the_poisonous_tree

> Fruit of the poisonous trees is a doctrine that extends the exclusionary rule to make evidence inadmissible in court if it was derived from evidence that was illegally obtained.

hwillis··on Federal appeals court finds geofence warrants “categorically” unconstitutional
A suspect runs into a garage with several unlocked cars, including yours. The suspect had drugs when he went into the garage and didn't when he came out. The drugs may be in any of the cars or none of them.

You would not get a search warrant in that case unless your car was the only one in the garage. The police don't know if there is three or even one phone connected to the tower in that time, and they should not get a warrant.

hwillis··on Thunderstorms Have Caused $45B in Damages in the U.S. in Six Months
The cost of thunderstorms has doubled from 5 years ago.

https://www.swissre.com/press-release/Insured-losses-from-se...

hwillis··on Thunderstorms Have Caused $45B in Damages in the U.S. in Six Months
Thunderstorms damage things like house cladding, roads, and water and power infrastructure. Things built 20+ years ago. Inflation is a reasonable control since it may make the repairs more expensive.

The stock market is not a reasonable control because it is related to the production of these assets. Production that is delayed and smoothed over decades.

hwillis··on Thunderstorms Have Caused $45B in Damages in the U.S. in Six Months
> Losses from severe thunderstorms have steadily increased by 7% annually in the last 30 years. 2023 marks an increase of almost 90% compared to the previous 5-year average (USD 32 billion), and more than doubles the previous 10-year average (USD 27 billion).

https://www.swissre.com/press-release/Insured-losses-from-se...

from the previous year's report.

hwillis··on How do jewellers capture every last particle of gold dust? (2017)
it's because "megagram" and "milligram" are a hassle. With the added bonus of mm/MM being used to abbreviate millions, I'm personally very grateful for tonnes.
hwillis··on Pattern of brain damage is pervasive in Navy SEALs who died by suicide
> .50 cal sniper rifles and shoulder launched weapons like Carl Gustaf, the M72 LAW or the AT4 can [produce shock waves that induce cavitation in brain tissue]

The US military, and militaries in general, do not use weapons like the M82/M107 as sniper rifles very often. The M82/M107 in particular has a recoiling barrel (the entire 2.5' barrel slides back when shooting) and isn't a very precise weapon.

They're used for blowing up ordinance or disabling light vehicles. They are sometimes used for hostage situations because they're more likely to immediately disable someone.

The US military has pretty rarely used shoulder fired weapons, since they very rarely have to worry about tanks or aircraft. SEALs in particular wouldn't be doing that.

hwillis··on EPA rule limiting certain 'forever chemicals' in drinking water is challenged
It was a health advisory in 2022, and it was revised from .004 ppt (note that you are off by an order of magnitude) to zero. Per the EPA:

> Drinking water health advisories (HAs) provide information on contaminants that can cause human health effects and are known or anticipated to occur in drinking water. *EPA's HAs are non-enforceable and nonregulatory* and provide technical information to drinking water system operators

The update was because of additional evidence of cancers. The EPA always acknowledged that .004 ppt is not reasonably detectable, but the point of an advisory is to note the level that is evidenced to be dangerous. Its a goal, and the new goal is "as good as possible". It should be obvious why the EPA should set requirements to reduce toxins while also noting what is a safe level.

https://www.epa.gov/system/files/documents/2022-06/technical...

hwillis··on The Delusion of Advanced Plastic Recycling
plastic packing peanuts vs paper stuffing, paper/plastic bags, blister packs vs cardboard boxes, paper-backed tape, paper/plastic plates, straws, etc
hwillis··on EPA rule limiting certain 'forever chemicals' in drinking water is challenged
2 of the 3 lawsuits were filed jointly by two large water utility trade associations, and the lawsuit (which is a petition for review) is here along with their statement: https://www.awwa.org/Resources-Tools/Resource-Topics/PFAS

The argument appears to be basically just this:

> According to a report commissioned by the American Water Works Association (AWWA) and prepared by Black & Veatch, drinking water utilities will need to invest more than $50 billion to install and operate treatment technology over the next 20 years in order to comply with new PFAS standards.

Which is ~65% more than the money the EPA has already allocated. That seems... very silly. The increase would be passed to consumers at about $5 per consumer annually. A study they commissioned themselves can only find a pretty modest difference. Hopefully this is shot down relatively quickly for being baseless, but who knows.

hwillis··on EPA rule limiting certain 'forever chemicals' in drinking water is challenged
Wrong. The EPA limit is 4 ng/L for PFOA and PFOS: https://www.epa.gov/system/files/documents/2024-04/drinking-...

An off the shelf Agilent spectrometer with EPA approved methods can detect PFOA at .28 ng/L and PFOS at .78 ng/L:

https://www.agilent.com/cs/library/applications/application-...

This limit is not even the lowest EPA limit in drinking water. Radium is limited to ~5 picograms per liter, 1000x lower that the PFOA/PFOS limit. Dioxin is 0.03 ng/L.

https://www.epa.gov/ground-water-and-drinking-water/national...

hwillis··on Amazon fined $5.9M for breaking labor law in California
1. where's the case law that makes it apply to corporations?

2. how is 100x the crime excessive, but $150,000 per pirated song isn't? That's a lot more than 100x the cost of a song.

hwillis··on BYD and CATL aim to launch new EV batteries with 6C charge rate
That means 400+ kilowatts- doable, but pretty disruptive for grid operators. Typically that will mean expensive (>.5 MWh) onsite batteries. Gas stations are inherently capital-intensive but making a $200k+ bet may still be a struggle.
hwillis··on BYD and CATL aim to launch new EV batteries with 6C charge rate
I totally agree, but it does only apply to drivers who have chargers at home. If you're parking on the street or in an apartment garage without enough chargers, things are different. That's less range anxiety and more range PITA though.
hwillis··on A fragment of human cerebral cortex reconstructed at nanoscale resolution
Capillaries are around the same width as the central bodies (soma) of neurons
hwillis··on A fragment of human cerebral cortex reconstructed at nanoscale resolution
There are 10^18 voxels in that cubic mm, so a petabyte (10^15 bytes) is not too bad. Must be a lot of empty space in there.

Put another way: this is pretty close to a football-field-sized cube, filled with sand. Each grain of sand is a voxel. The central bodies (soma) of neurons are about a yard wide at that scale.

hwillis··on Giant batteries are transforming the way the U.S. uses electricity
No, you're actually reading it wrong. Its a unit of power, not energy. You can follow the link on that line and see the EIA also using watts.

It's talking about the ability to supply power to the grid, not how long it can supply that power. It's arguable how useful a measure that is but they're certainly using it correctly. IMO they're also talking about it in a reasonable way- capacity and power are pretty closely related for batteries, and the ability to pick up demand is critical to handling instability.

hwillis··on LLMs can't do probability
The LLM isn't being perfectly reset. It chooses words randomly; internally it should be slightly different every time. That's the whole point of temperature.
← PreviousPage 7 of 34Next →