msoad

The idea from the proposal is that types are used by other tools to type-check and runtimes would ignore them. It's not final yet but it's very likely that in future you can `node -e 'function foo(arg: string) {}; foo(42)'` but if your code has `enum` in it, Node.js or browser will throw an error

After almost a decade of TypeScript my recommendation is to not use TypeScript enums.

Enums is going to make your TypeScript code not work in a future where TypeScript code can be run with Node.js or in browser when typings are added to JavaScript[1]

Enums results in runtime code and in most cases you really want type enums. Use `type State = "Active" | "Inactive"` and so on instead. And if you really want an closed-ended object use `const State = { Active: 1, Inactive: 0 } as const`

All of the examples in the article can be achieved without enums. See https://www.typescriptlang.org/play/?#code/PTAEFEA8EMFsAcA2B...

[1] https://github.com/tc39/proposal-type-annotations

why Dropbox when you can rsync, huh? ;)

I'm sure a few decades ago people would say that for not fiddling with actual binary to make things work.

This is really fast! Serialized 50k lines in 500ms on my Mac

To people who are experts in AI TTS:

Why elevenlabs has such a lead in this space? It sounds better than OpenAI and Google models

Not sure why you say this. I know multiple 10+ yo MacBooks still operation.

"problem". Looking at the $NVDA chart is not showing any problems to me

Hating Stack Overflow is fashionable but it really helped me solve me many problems that would've been much more effort otherwise. I agree that it's a victim of its own success.

Higher ups love to be "a data guy"! Just reduce all their responsibilities into a few numbers so they don't have to really understand what's going on underneath them. Have you notice how in love they are with their "progress dashboards"? They love to kick back and put their legs on the office desk watching minions making "progress" on OKRs so they can report that to their manager. It's too much work to really understand things and be on top of them anyways...

> On a more meta level it is tragic that we wont be able to use our computer online without being signed in to an online and be verified.

Fully agreed! Again, for most people this is the way of life. They type in Google in search bar to go to google Google and then search...

> If the site being visited get info about whom that specific user

My understanding is that you only get to know user is legit without knowing anything about them.

The cryptography that Apple uses for _their users_ is something novel and useful. I'm sure there would be completely open solutions that doesn't require Apple servers. At the moment this is Apple specific. I'm guessing most users will user Apple/Google solutions (just like how Passkeys are being used today) but for the security and privacy conscious there is always ways to keep your own private keys. Most people are handing over their private keys to Google and Apple for Passkeys today. That's not necessary a bad thing. Most people reuse passwords so it's an improvement.

When I try to log into AWS console and have to solve their captcha I always think that the target audience that this is designed to avoid can easily automate it (Some open source models can solve this without all that "AI safety" gatekeeping) while the majority of audience are simply suffering from this "feature".

Apple has proposed a solution to captcha[1] which I can't wait to be standardized and widely used.

[1] https://support.apple.com/en-us/102591

Your snobby message didn’t help. This is for completion of ollama cli tool itself. I did search and couldn’t find the docs. Maybe the person responding to me also not understanding what I want. I want to prompt a tiny model for completion

I had no idea! Where can I find docs to set this up?

I love fish and I’ve been a user for years. In the wake of AI, I am really interested in getting out of completes via a local tiny AI model

Does the fish team have any plans for integrating AI models for other completion?

in my opinion there is room for small and fast and large and slow but much smarter models. Use cases like phone keyboard autocomplete and next few words suggestion in coding or writing need very fast models that should by definition small. Very large models that are much smarter are also useful, for instance debugging issues or proofreading long letters.

Cursor really aced this. The Cursor model is very fast to suggest useful inline completions and then leaves big problems to big models.

I remember when the more tech savvy folks were migrating from Altavista and Yahoo to Google because they understood it is better sooner than others. The same thing is happening. I consider myself tech savvy (still!) and I have to admit that I rarely use Google for all sort of information.

As a side note: I am using Safari and I noticed that Apple's search is also replacing my Google searches. In the past if I knew name of a company or organization but not their website I'd Google it. Now I put it in the address bar and Safari very often finds the website for me.

Let's not assume that in a world where Apple has 80% market share artists are getting paid better...

Gemini for coding does not work for me. It gets so many things wrong

You can run a model that can beat 4o which was released less than 6 months ago _locally_! I know this requires a ton of hardware but OpenAI will not be the leader in 2025 I can assume. Always bet on open source (or rather somewhat more open development strategies)

The math and coding performance is what we really care about. I am paying for o1 Pro and also Sonnet, in my experience beside Sonnet being faster, it is also better at many tasks. In a few instances I got answers from o1 Pro but it's not justifying the price so I am cancelling and going back to $20/mo.

I am currently paying for Cursor, Claude, ChatGPT and v0! The productivity I am gaining from those tools are totally worth it (except for o1 Pro). But I am really hoping at some point those tools converge so I can pay less. For instance I am looking forward to VSCode Copilot improvements so I can go back to VSCode and once Claude has no limits I rather pay for one AI system.

There are new research where chain of thoughts is happening in latent spaces and not in English. They demonstrated better results since language is not as expressive as those concepts that can be represented in the layers before decoder. I wonder if o3 is doing that?

There is a trend towards less colors overall. Something is changing

https://lab.sciencemuseum.org.uk/colour-shape-using-computer...

No. I literally thought people in Portugal are speaking Russian at first!

I don't think anyone else in Europe rolls their Rs as much as Bulgarians. It's a give away!

It recognizes my accent correctly too btw

I am using o1 Pro in that chat

I paid $200 but it can't figure out something this simple. Sonnet 3.5 gets this question right

https://chatgpt.com/share/67528a78-c61c-8007-b6fa-1d1deb8d84...

I'm going to give it one month. So far I'm inclined to not pay this crazy fee if the performance is like this

React works because I can hire engineers that can be productive quickly. A lot of tech conversations leave the talent pool aside as if it's easy to find good engineers in all of the options you might have.

The ecosystem is another aspect. You have so many great options for off the shelf components and libraries with React that you might not have with other frameworks.

Before anyone mentions Custom Elements and Web Components, I must say I tried it. Gave it a lot of effort to be a solution but it is not!

Article says BYD sells Sodium-Ion cars too. Are those cars super low range with quicker charging times? If they are much cheaper batteries it makes sense for an urban setting to have such cars.

Norway has a good oil refinery industry though. They refine oil from other places because they're so good at it.