swsieber

Here's a transcript: https://the-culture-shocks.blogspot.com/2022/03/evaluation-o...

I'm actually a really big fan of affine types and adts with great pattern matching. The performance is nice too, but I stayed for all the other nice bits that are put together really well (except async, but eventually...).

Rust has been my favorite language by far to refactor in.

I want to take a minute on rant about the insanity that is having node_modules used a caching location. That's the big reason for the incompatibility for between npm and any package manager that attempts to clean up the node_modules mess (though the impact varies by solution). Maybe I'm just spoiled, coming from java, where you can expect your libraries to be in a read-only archive, and that tools separate the runtime from their cache.

I was a yarn 2+ convert because of how much faster installs were (and the corresponding reduced disk bloat). It unfortunately doesn't play well with Angular, and lots of people fall back to the "node_modules" strategy which does away with much of the benefits.

Is there any good comparison out there of the JS package managers, and the trade-offs they make? I happen to really like the disk/time vs compatibility trade-off yarn 2 makes, but not the zero-install (which had been the default for a time). But hearing that Bun is faster, and better ergonomics? And other people mentioning pnpm? That makes want a package manager in depth comparison.

It did? I thought that proposal was basically stuck; I haven't checked in a while, but I haven't heard of it moving forward either.

Huge shout out for Antichamber. It had a wonderfully mind bending learning curve.

I'd love examples for things I can wrap my head around.

That's way slick.

Can I ask what your background is, and what things you're used to working with? I don't have the chops to build what you built, but I'd love to get there.

> Or did the authors count the amount of training data for the LLMs to the required training data for the destined/task-specific models?

Yes.

They are counting the amount of data you need to collect to solve your problem. I can grab a pretrained LLM, and the data I have to collect in that instance is what I need to fine tune it.

> also to anyone reading this. DONT POST any code to Zig issue tracker. post it to your own repo, and link to it if need be. any code you post to the tracker becomes MIT license, which I learned the hard way.

Any insight on this? How can this possibly be true?

Oh that's a bummer about the isolated or filtered. Thanks for the heads up!

For those wanting to do similar things, I found https://www.thinq.ai/RPi_Ethernet_USB_OpenWRT.html to be a pretty good page on how to do usb over ethernet in both directions with open wrt, which was a big chunk of the mystery for me.

Interesting. I did not know that about Animal Crossing. It sounds like that's because the last person is remote? I'd ideally be trying to resolve that quirk by using VPN (plus proxy-arp?) to put all the switches on the same VPN despite being geographically distant.

Hmm... that's a reasonably priced prepackaged solution. Thanks for pointing that out.

I am tempted to still use VoCore since it appears to run OpenWRT out of the box and I'm really only aiming to support a single device. Hmm..

Thank you!! I hadn't hear of proxy-arp, but it seems like from cursory google searches that reading up on it will lead me down a promising path.

So, something I've really wanted to do, but don't have the chops for is this:

Make a small dongle that plugs into an ethernet jack and sets up an wifi access point that is run over a VPN. e.g. I could give one to my friend and any device they connect would connect to my VPN.

a) Is that possible?

b) Would that get around Nintendo Switch NAT issues?

The last time this chip came up I got the idea, but I'm not sure how doable it is.

Edit: I forgot that a stretch goal would be to change the setup. You could have the VoCore connect to a wifi network, and then plug the USB side into a device (nintendo switch doc or computer) to emulate a USB ethernet adapter, but of course send the traffic over wifi through the VPN.

A single threaded Tokio runtime still has a separate thread pool for heavy tasks. Maybe block_on? I can't remember where I read that tho.

Edit: not block_on, but spawn_blocking. See https://www.reddit.com/r/rust/comments/16ebdi1/comment/jzy7f...

There's a few other options that I think in combination would make things much better.

1) Make using proc macros in your own application code faster: Encourage better macro re-export hygiene. Basically, one of the reasons serde + serde_derive is so slow is because for serde re-exports serde_derive to compile before it itself can be built. A solution (I didn't come up with this myself) would be to have a crate that re-exports serde and serde_derive together; see top comment herehttps://www.reddit.com/r/rust/comments/1602eah/associated_pr...

2) Make using libraries that use proc macros faster: Publish pre-expanded source code, such that no proc macros run for dependencies. Some work would have to be done around conditional compilation to make things just work (TM), but I think it could be done. (Who knows, maybe it'd be really hard to make the compiler deal with expanding proc macros on stuff behind conditional values. Also, there'd be issues regarding macro hygiene. Both solvable, but I'm not sure how much effort they'd take).

I belong to a religion that's really strong in western U.S. and less prevelant as you go east. Congregations are assigned based on geography and sometimes language.

I prefer the congregations out east that cover half a city because you interact with more doverse people across a lot more socio-economic classes just due to the wider geographic area.

I wish there were good incentives to build combined/intermingled housing to have more mixed neighborhoods.

Extracting data from documents.

Isn't the big issue not bots reading but bots generating? You could allow bots in a read only fashion.

I mean... you just batch those cents and have cash out limits. Preload accounts, etc.

* note that the word just does an obscene amount of lifting in that sentence.

Well... no. Proof of work doesn't mean you the human have to do work. Bitcoin is a proof of work system but you don't calculate the numbers by hand.

The question there is if an acceptable amount of work (e.g. cpu/gpu work) on a mobile phone is a large enough deterrent for a bot.

>> on a corpus of ... abstracts

Wait a minute. Isn't the abstract the summary? I've read accounts of people digging through a paper and seeing impossible, obviously fabricated numbers. I don't think those types of issues would surface in an abstract.

> [I]t is common sense to a _celiac_ that gluten causes inflammation throughout the body and brain even of _non-celiacs_

(emphasis mine)

I don't follow how this is common sense. It looks like extrapolating from your condition (celiac) and projecting it onto those that don't have your condition (non-celiacs).

Any idea how it was made? I'd like to do the same.

IIRC been some proposals floating around, but one major block is asncy drop. This is the article I'm thinking of specifically: https://blog.yoshuawuyts.com/tree-structured-concurrency/

I believe rust is sort of similar in that apart from the references, it's very heavy on the move semantics. And IIRC, sometimes it's doing extra memcpys. But I don't know if Rust is explicitly targeting const reference passing.

That said, if you're all in on it I imagine that the front-end could pretty aggressively target that, much like how rust uses no alias.

I had the same questions and found this article: https://snacksafely.com/2016/06/kelloggs-unintended-conseque...

Basically, there were stricter measures put in place called HARPC. From my linked article:

> The new directives mandate that the “Top 8 allergens” identified by FALCPA (peanuts, tree nuts, milk, eggs, wheat, soy, fish, and crustacean shellfish) must either be ingredients of the product and identified as such, or the manufacturer must take extra care (and cost) to ensure that there is no cross-contact with them. There is no middle ground or “out” for the manufacturer, which is why we believe “May contain” type label advisories are heading for extinction. And that poses a problem, at least in the short term.

> companies, when faced with the added burden of instituting and documenting cross-contact prevention measures as dictated by HARPC, may instead choose to add trace amounts of the allergen to the product, as doing so makes the allergen an ingredient of the product and obviates the need for preventative cross-contact measures for that allergen.

> it means that manufacturers will either take stricter measures to prevent cross-contact or add a trace amount of the allergen and list it in the ingredient list, thus eliminating the ambiguity that currently plagues us all.

> A compromise that might have avoided the unintended consequences of companies like Kellogg’s adding traces of allergens to their products is to have offered them a third option: A mandatory “May contain” label advisory for any product made on shared equipment or in shared facilities that did not meet the FSMA threshold for cross-contact prevention. Such label advisories are voluntary today, rendering them ambiguous at best, but a definitively worded and located advisory statement included on all such products would have provided a way for manufacturers to meet the requirements of HARPC without resorting to the addition of allergens.

So I'd be surprised if it was happening on an ongoing basis, but I can definitely see why people would be irked.

I mean... people who go to cancer clinic have cancer at a higher rate than those who don't. Cancer clinics cause cancer. And therapists create mentally unstable people, etc.

You wouldn't have to fork the universe. I think a randomly distributed trial would work, and could be reasonably made to work (woth Gov. help). Not sure though.

FWIW, I think we should prevent private equity in lots of places, especially hospitals.