1f60c

It's pretty easy to understand why Apple doesn't want its models to reproduce racial slurs, but what’s wrong with "Boris Johnson?"

(See, e.g., here: https://github.com/BlueFalconHD/apple_generative_model_safet...)

Is it connected to the network?

searchcode doesn't seem to work for me. All queries (even the ones recommended by the site) unfortunately return zero results. Maybe it got hugged?

https://searchcode.com/?q=re.compile+lang%3Apython

Oh gosh, I'm so sorry. My comment wasn't meant as criticism of you or Nping; I had just never seen a DNS server intentionally return obviously incorrect results before.

I wonder why www.google.com resolved to 31.13.94.37, which IPinfo tells me belongs to AS32934 Facebook, Inc.

https://ipinfo.io/31.13.94.37

I wonder what this means for Support using ollama as an inline_completion_provider https://github.com/zed-industries/zed/issues/15968. ':]

I hadn't heard of Baseten before (it seems to be in a hot niche along with Together.ai, OpenRouter, etc.) but I'm glad I did because I was actually noodling on something similar and now I don't have to do that anymore (though it did teach me a lot about Fly.io!). Yay economies of scale!

(2021)

I think you're moving the goalposts. I didn't know this problem existed until OP's article taught me about it, but it seems pointless to continue this discussion.

Wait really? Are you saying this Google search[0] returns no useful results at all?

The prompt I used isn't a particularly good search query, and I got:

  * A Stack Overflow question about this issue with a proposed solution;
  * A Reddit thread offering a workaround to non-technical users;
  * This thread on HN (which OP of course couldn't have seen, but anyway);
  * OP's article;
  * A blog post about OP's article;
  * A blog post with an elegant solution using pyftsubset

What am I missing?

[0]: https://www.google.com/search?q=Why+would+flag+emojis+render...

Did they even try? 3.6 Sonnet, R1, 4o, and o3-mini-high all correctly diagnosed the issue given the same basic prompt, albeit with varying levels of utility.

Heck, even pasting the "prompt" into non-AI Google and Bing worked.

That reminds me of the debate surrounding Star Trek Into/into Darkness https://en.wikipedia.org/wiki/Wikipedia_Star_Trek_Into_Darkn...

I had the same exact thought. It's good to remember that writing ≠ publishing[0].

I'm not sure if it matters much in this case, but the OP is (2017).

[0]: https://news.ycombinator.com/item?id=23316443

The number of skills one has to master for a project like this is just mind-boggling, and the part about the tortilla press made me laugh out loud. A labor of love in the most literal sense!

That said, I would feel pretty nervous about actually selling these, as I wouldn't be surprised if it is legally considered a medical device.

Genius! I assume it's intended to be used to moderate the input to and output of OpenAI models, but I think it will moderate anything regardless of the source, and it's free (aside from one network call, and potentially sharing data with OpenAI?).

I like Sindre Sorhus' Scratchpad (syncs across Mac, iPad and iPhone using iCloud).

Why isn't the app available in the Netherlands?

No need to resort to ad hominems; I just wasn't sure what you meant by "CAPTCHA as a use case" and wanted to point out that it's possible to do much better than $0.1/CAPTCHA solved.

If you're talking about CAPTCHA solving as a service, that already exists, and the cost is measured in mere dollars per thousand CAPTCHAs solved.

I'm actually really disappointed in Wikimedia. Why didn't they fight this at all? Or, y'know, just ignore it?

AFAIK the owner's manual says you have to keep your hands on the wheel and be ready to take over at all times, but Elon Musk and co. love to pretend otherwise.

I got that feeling as well. In addition, I suspect it was originally written for an internal audience and adapted for the 'blog because the references to SLOs and SLAs don't really make sense in the context of external Uber customers.

Resinator's error messages look amazing! I also feel like I’ve gained a lot of cursed but useless (to me) knowledge, so thanks for that. :-)

I don't have a horse in this race, but regarding FONT resources, I would like to humbly suggest not supporting them at all. Radical, but from what you wrote, they do seem pretty weird and ripe for accidental misuse. Plus, they are obsolete and it seems like Resinator already intentionally diverges from rc.exe in a few cases anyway.

Just use scientific notation. :-)

One thing I’ve always felt is that the relative difficulty of each task seems to have flipped? I could write a bird classifier in my sleep using fastai, but I have no idea how to do a GIS lookup.

Nice writeup! It reminded me a bit of Julia Evans' blog in terms of content (learning by teaching).

> Umbra has an open data programme where they share SAR imagery from 500+ locations around the world

I think they're referring to this: https://umbra.space/open-data/ (warning: most files are absolutely ginormous)

https://archillect.com/archive? Sadly it seems to have gone quiet.

This has been invented a number of times. Facebook's version is called Open Graph.

https://ogp.me/

Bob addresses this in the introduction^:

> We will abstain from using [parser generators] here. I want to ensure there are no dark corners where magic and confusion can hide, so we’ll write everything by hand. As you’ll see, it’s not as bad as it sounds, and it means you really will understand each line of code and how both interpreters work.

^ https://craftinginterpreters.com/introduction.html#the-code

It has to be satire because of Tom's complete overreaction and the fact that comments are actually one of the easiest things to handle when building a lexer (usually, you just discard them). Eval'ing them makes no sense.

That said, I suppose stranger things have happened.