samwillis

This is really cool, I love the retro design.

If the devs are here, what does the tech stack look like?

What's the strategy for conflict resolution, i.e. two people edit the same area at the some time? Is that resolved on the server or on the client?

--

Edit: to somewhat answer my own question, it seems to be open source, client and server here: https://github.com/OurSources

Server is JS running on Node, need to dig deeper...

You touch in an important topic here, how our understanding of AI/ML/LLMs will influence our "understanding" of the human brain and intelligence.

My fear is that we will ascribe too much human behaviour to that we see in and understand of our AI inventions, and that this will result in the dehumanisation of people.

So essentially my fear is what we justify doing to each other due to AI, rather than what "AGI" could do to us.

This is both an excellent demo of what you can build with Textual, and just a brilliantly ridiculous project, it's exciting to see.

I'm looking forward to seeing some commercial products adopt Textual, I hope we see some soon!

I suspect punctuation adds significant meaning to the models, that could be why so much computation is applied to it.

That's not to say a pipeline couldn't be effective.

If I'm following correctly, does this mean that with this change along with a model being quantized, we could see models that are 5% the size (on file system) and memory usage but almost identical in output?

No, but also a lot of users enjoy them. It's like marmite, you either love it or hate it.

Also note that most product managers love them.

Quite often they are a symptom of a UI that isn't descriptive to particularly discoverable.

Apple and windows are trademarks for software. You would be unable to get the trademark "Apple" for a company selling fruit, or "windows" for a window fitter, and neither could Apple or Microsoft.

𝕏 is a pre-existing symbol used by, and part of a standard character set for, software. You could probably get a trademark on it for a company selling apples, but I highly doubt you could for a software company.

Trademarks are against a "class", these can be quite wide. They would clearly want to own that exact symbol for "Class 9: Computer and software products and electrical and scientific products."

Unicode already contains exactly it, it's a mathematical/scientific symbol, even if they did get the trademark in that category (which I doubt) they will never be able to enforce it. With trade marks, if you don't successfully protect them, you loose them.

Yes, but those "classes" can be quite wide, there is for example no way they could trademark it for "Class 9: Computer and software products and electrical and scientific products."

It makes typesetting easer, pretty genius really! /s

We're going to see a trend of companies using Unicode characters for their logo now aren't we?...

Absolutely.

They can probably trade mark it for social media (unless someone already has), but are highly unlikely to be able to trademark it broadly, which clearly is the intention.

Bloomberg have been doing this for years, I believe it's just a relatively simple automated rewrite/republish from various news wires. Most larger newspapers/sites have some level of automation from news wire to published article.

This isn't some new LLM based system replacing the work of journalists.

This running in the browser via Emscripten by Georgi Gerganov of llama.cpp fame:

https://ggerganov.com/llama2.c/

Via his Twitter with ongoing thread: https://twitter.com/ggerganov/status/1683174252990660610

This and the original is all absolutely awesome, it's obviously only a proof of concept with a tiny model, but local first LLMs are really exciting. I particularly love the idea of being able to build webapps with local inference.

With optimisation, research into ways to make smaller models, partial downloads, and then the opportunity to use WebGPU we potentially have the start of an exciting new way to build privet local LLM based apps.

It's never going to be up to the same capabilities of hosted LLMs on massive clusters of top end GPUs, but there are so many use cases that this sort of thing will enable.

It's currently only running a tiny tiny model, but the point it that this is possible in the browser with WASM and no gpu. It's not going to be CharGPT but there will be use cases that this sort of thing enables.

Agreed, perfect Sunday HN material.

That's settled, we're even.

> Why doesn't the United Kingdom have the name of the country on its stamps?

> Because the United Kingdom had the privilege of being the first country in the world to introduce postage stamps, meaning that they did not need to be identified as coming from that country, especially when used domestically.

I think it's particularly interesting that the US use .gov and not .gov.us (as a Brit). I'm sure there are oversights on who can acquire an inherently international .gov domain, but for example here in the UK .gov.uk domains have a strict application process [0] managed by central government.

It just seems to me that it would be more secure, and more reassuring to citizens and visitors that they are on the correct site it's under a cctld that's clearly affiliated to and managed by that government.

0: https://www.gov.uk/apply-for-and-manage-a-gov-uk-domain-name

--

Edit: turns out .gov is exclusively for the US, not sure I feel good about that, particularly as .com and .net are very much not just for the US.

The possibility of the US government creating a .gov specifically to confuse uses in a foreign country isn't ideal.

I get it, you invented the internet, but the special status you have over it is a little frustrating.

https://www.whitehouse.gov/wp-content/uploads/2023/02/M-23-1...

There's nothing cool about using LLMs to generate content for marketing.

The most successful product lunches on HN are the ones where people can try it out immediately, no paywall, no registration, just straight in and have a go.

Waitlists, or "pre announcements" very very rarely get voted into the front page.

I imagine fairly similar to what happens with a gas/petrol vehicle, similar amount of energy to release. Maybe with a bunch more toxic fumes...

I think we will see solid state battery's in portables (Phones, Laptops) long before we see them in cars. The yield required, cooling, structural integrity, is all very hard to solve. Starting on a smaller scale for the significantly larger per unit market seems to me to make more sense.

The insanity of what's made possible with CAD and CNC for manufacturing is incredible.

Effectively these are two towers that twist/spiral around each other, a double helix. I think the sky walks double a structural braces between the two sides.

The structural engineers will have had fun!

Something like 90% of burglaries are looking for car keys. That's why I keep my keys by the front door in eyesight on anyone who comes in. I have insurance, and don't want a thug wandering around my house.

Or the third to to illustrate the argument in another context.

I usually avoid analogies of HN as you are often called out on the technicalities of them. And you are right that can often damage an argument.

However, I have argued that changing peoples behaviour, even when their current behaviour is bad for them, is often impossible despite all evidence, from the beginning.

Is the cost of lung cancer lower than smoking?

Some people will just never change a behaviour.

> I think GP's father doesn't have that awareness, and never will, by choice.

This! I mention password managers, two factor or just try and get him to improve his passwords and his eyes glaze over and can see him beginning to nod off. Probably somewhat my fault for always bring them up after a meal when he's probably already half way to napping...

Exactly, they are in recent years moving to address these previous issues, I applaud them.

I would like it if they were more discoverable, but not necessarily with the obnoxious pop ups. It would be awesome if they let you list a PWA in the App Store for a fee or the 30% cut.

It is when they are all a variant of each other. I can't imagine him successfully writing down a bunch of none alphanumeric characters and not giving up.

This is why I hope PWAs take off more. If there is an implicit "save first party cookies forever" for PWAs that have been "installed" and we make them discoverable it's a net win.

Gosh, you're fun.