Hacker Newsnew | past | comments | ask | show | jobs | submit | happyraul's commentslogin

FEIE is only one of the options for avoiding federal income tax. The other is the Foreign Tax Credit, which has no such limit: https://www.irs.gov/pub/irs-pdf/f1116.pdf. If the place an American lives and works has a higher income tax rate than the US one, in practice he will not face any tax liability, regardless of income level.

No, the point is that you can set up the testing exercise without using an LLM to do a simple find and replace.


Its a test. Like all tests, its more or less synthetic and focused on specific expected behavior. I am pretty far from llms now but this seems like a very good test to see how geniune this behavior actually is (or repeat it 10x with some scramble for going deeper).


This thread is about the find-and-replace, not the evaluation. Gambling on whether the first AI replaces the right spells just so the second one can try finding them is unnecessary when find-and-replace is faster, easier and works 100%.


... I'm not sure if you're trolling or if you missed the point again. The point is to test the contextual ability and correctness of the LLMs ability's to perform actions that would be hopefully guaranteed to not be in the training data.

It has nothing to do about the performance of the string replacement.

The initial "Find" is to see how well it performs actually find all the "spells" in this case, then to replace them. They using a separate context maybe, evaluate if the results are the same or are they skewed in favour of training data.


What sucked about it? I've been considering it for a long time as an alternative to flying across the Atlantic.


It’s a Carnival cruise. If you like going on cruises you’ll probably like it. I don’t like cruising - and we went with our four year old, and there is literally nothing for kids to do on board, (there’s a “kids club” but it was closed for the duration) and they aren’t allowed into the dining room, so we ended up living off the crappy buffet and the very limited room service menu.

Our stateroom also hadn’t been cleaned when we boarded, and they had problems with the black water system so several decks just stank of shit. Our balcony had a persistent leak above it, so we couldn’t use that either.

The tickets themselves weren’t that much, but then it’s about €1,000 for internet access, and they nickel and dime you on absolutely everything, while not allowing you to have the things you already paid for - our bill on departure was about €4,000.

Oh, and at disembarkation in Southampton they destroyed two of our suitcases, and told us to go cry harder - and then it’s a three hour wait for a taxi.

Honestly, the whole experience was pure crap.


The reason there will never be enough lanes is not population growth. It's that we generally don't provide people with alternatives to driving, and adding more lanes just makes even more people have to drive, since things get even farther apart.


Are you saying on iOS there is no way to browse the Internet without an ad blocker? Maybe I should get an iPhone to stop spending time on my phone...


Can't answer that authoritatively since I don't know if there's something iOS-specific, but Firefox is the only browser on Android that lets you have uBlock Origin. And I know that Firefox on iOS doesn't.


That's simply good writing practice. I find it more taxing to read digits than prose.


Thank you. To me after reading the parent comment the numbers option was so evidently better that I didn't even consider that someone like you could exist. My conception of humanity has been slightly enlarged.

If I may ask: Do you also find numbers more difficult to parse when doing math pure math operations? Is this:

Two hundred thirty five plus one thousand eight hundred twenty two

Also easier for you to parse than this?

235 + 1822

Or do you have two "parsing modes" ("text" and "math"), and going from one to the other is the difficult part?


I was taught numbers up to ten should be spelled, the rest use digits


Chicago Manual of Style (though it says 1 to 100, er, I mean one to one-hundred). I try to use a CMOS subset for my professional/technical writing, mostly for consistency, but, partly so that I don't need to argue with people with subjective opinions about how I'm writing it wrong.


I can recommend The Story of Art: https://en.wikipedia.org/wiki/The_Story_of_Art


To me the question of what activity/method is more "valuable" in the context of art is kind of missing the point of art.


This may be a good summary of the blog post, but in my opinion it is an uninformed understanding of Elm and Elm's history. JavaScript FFI isn't and never has been a "critical language feature". Rather, people discovered an implementation detail that allowed them to create thin Elm wrappers around JS libs (think Elm interface for d3, leaflet, moment, etc...). This was (rightly IMO) seen as undesirable for multiple reasons and 0.19 closed off the loophole at the compiler level.

As for being banned from the community, I keep seeing this claim but have never seen or heard of such things happening. Sure, there may be a negative reception to folks who keep wanting to re-litigate a decision that was made over and over, but nobody has been banned from anything.


But do either of them manage to show a legible street name at a convenient zoom level?


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: