Cool project, but they should really upgrade at least the downloads to https...

This would break so many websites. There are valid uses for the history API, I often do modals/popups as shareable URLs, and using the back button closes it.

There's a (very) big difference between reading weekly and reading a book in the past 12 months. I used to read a book or two in the past 12 months, back then it was on vacation to have something to do while lounging on the beach. Today I read a book every month and it's something I do instead of watching TV. I would not call my younger self a reader.

Interesting choice to change the time of the comment, a deja-vu can be weird enough without staring at a comment with a recent timestamp.

Didn't know about this project, looks useful. Thanks for bringing it to my attention, FAKKU.

> Why even use else if with return...

What is the problem with that? How would you write that snippet? It is common in the new functional js landscape, even if it is pass-by-ref.

Using guard clauses. Way more readable and easy to work with.

  export function extractSearchToken(completionToken: {
    token: string;
    isQuoted?: boolean;
  }): string {
    if (completionToken.isQuoted) {
      return completionToken.token.slice(2).replace(/"$/, '');
    }
    if (completionToken.token.startsWith('@')) {
      return completionToken.token.substring(1);
    }
    return completionToken.token;
  }

It doesn't try to do anything. It doesn't work like that. It regurgitates the most likely tokens found in the training set.

That is so reductive of an analysis that it is almost worthless. Technically true, but very unhelpful in terms of using an LLM.

It is a first principle though so it helps to “stir the context windows pot” by having it pull in research and other shit on the web that will help ground it and not just tell you exactly what you prompt it to say.

They are amazing tools, but when people try to give them agency someone has to explain it in simple terms.

Hmmmm i didn't know that... so a machine is not human is your point? Look, i know it doesn't try, just like a sorting algo does not try to sort, or an article does not try to convey an opinion and a law does not try to make society more organized.

Hey! How can I make this about LLMs?

(Many countries' laws are already available online and included in the dataset they're trained on. The project is very cool for humans though.)

And if you set `set editing-mode vi` in ~/.inputrc (readline configuration) you'll have it in even more places.

What is it like to be this proud of not learning the tools you use? Do you really think several paragraphs to an agent that may or may not be correct is the "easy" way compared to just checking the manual for the flag you want?

I will never understand people like you.

Tools are means to end.

They don't matter much to me.

That's fine to a point. But if you don't learn your tools, you will have rendered yourself unable to catch when the LLM gets things wrong (and it will get things wrong because it doesn't understand anything it touches). That, in turn, will mean that you're going to struggle to reach your desired ends but won't have the understanding to figure out what's wrong.

"Tools are just a means to an end" means you need to not get attached to a tool if it's not doing the job. It doesn't mean you don't need to understand your tools.