ijustlovemath

Perhaps, but the profit motivation is completely different. One's there to serve and the other to bill hours.

Inflation only happens if the US starts paying down the debt -- locking up the tariff money in interest payments. Otherwise, there's fundamentally no increase in the supply of money. I'm guessing there still will be inflation for consumers, especially seeing as most don't have the funds to invest in bonds.

Anyone who's played cards or an MMO know those are insanely good odds -- virtually guaranteed!

thanks for clarifying that!

I hear what the author is saying, but I think if composability is your goal, you should look no further than traits/interfaces. Accomplishes the same goals but actually far more general (which, according to the author, is a good thing)

c# also has a runtime you have to ship with any binaries

It would be a huge loss in capabilities. The only place to build the most advanced process semiconductors we have (like the ones in NVIDIAs flagship AI cards) is at TSMC, in Taiwan. A lot of the equipment there is highly specialized and tightly controlled. Considering the EU is not exactly interested in US foreign policy right now, you now have to take into account that the most advanced EUV systems are made solely by a company in the Netherlands... you see how the US might never be able to replace the current state of the art chips.

It's possible foundries in SK or the new ones we're building will get there one day, but it wouldn't be for a while.

You could do a simple weighting, probably adjusting for sample distributions.

If you added half stars and stuck to a 5-star system, there's less user surprise for functionally the same thing.

Semi related: when I'm using a dict of known keys as some sort of simple object, I almost always reach for a dataclass (with slots=True, and kw_only=True) these days. Has the added benefit that you can do stuff like foo = MyDataclass(*some_dict) and get runtime errors when the format has changed

For those interested, see slide 4 of section 7: Hertzian dipoles

https://courses.grainger.illinois.edu/ece350/ece350lecture_n...

I definitely think it takes a Maxwell to describe the physics behind displacement current (which I think was not around in Maxwell's initial drafts), but the derivations of the four classic Maxwell equations are fairly straightforward multivariable calculus problems that start with fairly simple assumptions. Multivariable calculus was relatively young in his era, but I think a motivated student from the current era could discover it given a summer of thought experimentation.

See for yourself: https://m.youtube.com/watch?v=AWI70HXrbG0

If you want some true genius, behold the shear terror of the near field of a dipole antenna. All shall view her and despair.

What are the limitations on usage? What's the approximate usage percentage, say, per kilotoken of context? Is there a point at which users are not allowed to query (profitably)?

Re: open source, what's your general attitude/commitment towards the community? Is it more like SQLite (no contributions accepted), or more like Rust (let's get everyone involved)?

The physics of the 1800s had a lot of low hanging fruit. Most undergrads in physics can show you a derivation of Maxwell's equations from first principles, and I think a fair few of them could have come up with it themselves if they were in Maxwell's shoes. The hard truth is that the physics/math of today is just much further afield, and much harder.

Which would probably raise less than a modest wealth tax on billionaires, but we know that will never get traction (even though its an extremely popular policy)

Agreed that the value of a CGM is in the change information, and that adding a CGM is probably the biggest quality of life increase for anyone with diabetes. Highly pro CGM if you can get it!

The issue with spectroscopic approaches is the amount of noise can be really hard to disentangle, to the point that you might get really unreliable trend information, where it might even be dangerous if you're making dosing decisions off it. And even if you aren't, getting incorrect trend information doesn't really help you any more than just not knowing it.

My container using is admittedly pretty simplistic (CRUD app with some REST services), but after initial setup I've found it to be extremely reliable and simple to use. They strive for 1:1 docker compat so I think it should be pretty easy to migrate.

we're building an artificial pancreas for hospitals, so I know a good bit about CGMs. Noninvasive blood sugar sensing is horrifically difficult. Every few years, people come along and say "oh this is just some simple DSP on spectroscopic information, piece of cake" before inevitably running up against:

- skin conductivity changes over time

- the ways in which skin tone changes signal absorption (which itself changes over time)

- the ways in which different levels of fitness affect blood flow, material density etc.

You also can't use it in a hospital setting, due to how your skin and bloodflow changes during serious conditions like sepsis (though I'm guessing they're not thinking about that market).

Really smart people have been trying to use Raman spectroscopy to solve this problem for decades at this point (early patents go to early 2000s). Apple is an extremely strong hardware vendor, and I wish them luck, but I would not hold my breath for this. Plus, I'm guessing they will not open the signal up for looping, which would really leave the T*DM community out to dry.

Sure, but if privacy matters to you, you have the option of buying credit anonymously and applying it to an anonymized account number. And if your threat model includes nation states, you're definitely not buying anything with a credit card. I also think if you're after payment details, there's more lucrative targets, eg Stripe.

Meaning they're achieving their privacy goals without any inherent trust in their systems (eg no databases of user info, etc)

Mullvad is near the cutting edge on zero trust deployments; allowing user traffic to pass thru, with guaranteed no logging, assumption of compromise guiding system architecture, etc. Nobody can withstand a nation state, not even other nation states, so I feel like they're doing the best that can be reasonably expected of them

Adding some anecdata, I feel like emulator is mainly used in the context of gaming, in which case they actually care a great deal about accurate reproduction (see: assembly bugs in N64 emulators that had to be reproduced in order to build TAS). I haven't seen it used much for old architectures; instead I'd call those virtual machines.

definitely agree on simulator though!

I think that when you, as most HN commenters are wont to do, are trying to achieve a certain level of mastery in your craft, you're going to want to dive deep, understand the abstractions, have full control, be able to start with nothing but a UDP socket and a steady hand, but that approach misses a huge number of users of these languages. You don't want scientists to have to worry about setting up a venv, you want them to analyze their data and move on with their lives. Sure, people like you or me will be able to setup and source a venv in no time, without expending much mental energy, but we're not who this product is for. It's for the rest of the users, the 99% who aren't even aware places like this exist.

All that said, I'm pretty skeptical of using uv until their monetization strategy is clear. The current setup is making me think we're in for a Docker-like license change.

How are they handling the returns VCs expect through this free software? If it's so easy to deploy, surely we should expect a Docker-like licensing model in the near future?

What's remarkable to me about criticism like this is how quickly it's rebutted by people claiming "where did they say they would do this," as if these people don't make incredibly rushed and poorly planned decisions all the time. It's like an idea immune system that rejects any criticism or self reflection. It would be sociologically fascinating if it wasn't being combined with a dereliction of congressional power and an unchecked executive.

I thought it was Timsort?

it's fundamentally how these things work; learning token distribution given prior context. The expected output over time is the mean value of that distribution. Regression to the mean is the danger I'm talking about.

This is the biggest risk of AI imo; almost by definition your thoughts regress to the mean when using it

you'd only have to test worst/median case scenes, which you could find with a bit of profiling!

static consts in C carry their identity through their (fixed, unchanging) pointer address. Lets say you have a business rules engine, that's only meant to ingest threshold values from a certain module. You want to know if the 3.0 you're using is coming from the correct place in the code, or if there's a programming error. With a define, there's not enough additional information to be able to. With a static const, you can just have a const static array of pointers to the valid threshold constants, and use that in your tests for the rules engine.

I work in a highly regulated field, so often this level of paranoia and proof helps us make our case that our product works exactly the way we say it works.