milansuk

No need to use Ollama. LLama.cpp has its own OpenAI-compatible server[0] and it works great.

[0] https://github.com/ggerganov/llama.cpp#web-server

Higher volume legal actions can only be successful as "peer-to-peer". If it goes through a courtroom, there is no way they can handle this kind of volume(even with AI tools). Imagine that the CEO is informed there are 20 new legal actions and the first court hearing will start 10-20 years from now when the CEO will not be part of the company anymore.

It doesn't do that fancy "make this text more professional"

I looked into the Scramble code[0] and it seems there are few pre-defined prompts(const DEFAULT_PROMPTS).

[0] https://github.com/zlwaterfield/scramble/blob/main/backgroun...

True. The only solution is to keep your data outside cloud(aka someone else's computer) no matter what encryption you use.

But file syncing is a “dumb” protocol. You can’t “hook” into sync events, or update notifications, or conflict resolution. There isn’t much API; you just save files and they get synced. In case of conflict, best case, you get two files. Worst — you get only one :)

Sync services haven't evolved much. I guess, a service that would provide lower APIs and different data structures (CRDTs, etc.) would be a hacker's dream. Also, E2EE would be nice.

And if they closed the shop, I would have all the files on my devices.

This looks cool for v1! The only problem I see is most devices don't have much RAM, so local models are small and most requests will go to the servers.

Apple could use it to sell more devices - every new generation can have more RAM = more privacy. People will have real reason to buy a new phone more often.

I don't see any explanation for why they trained 8B instead of 7B. I thought that If you have a 16GB GPU, you can put 14GB(7B*16bits) model into it, but how does it fit If the model is exactly 16GB?

The progress is insane. A few days ago I started being very impressed with LLM coding skills. I wanted Golang code, instead of Python, which you can see in many demos. The prompt was:

Write a Golang func, which accepts the path into a .gpx file and outputs a JSON string with points(x=tolal distance in km, y=elevation). Don't use any library.

This is an implementation of a transformer and in README it's presented as text->text. Tokens are just integers going in and out.

Is it possible to use it to train other types of LLMs(text->image, image->text, speech->text, etc.)?

You can run Gemma and hundreds of other models(many fine-tuned) in llama.cpp. It's easy to swap to a different model.

It's important there are companies publishing models(running locally). If some stop and others are born, it's ok. The worst thing that could happen is having AI only in the cloud.

Agree. Spinning out the FSD and Optimus teams wouldn't accelerate progress on the new models. Also, they will make more money from humanoids than car robots.

Yeah, they probably don't want their 'demo' links on the HN home page.

Believe it or not ...

I believe and agree with most of you wrote ;)

The main problem with HTML/CSS/JS is programmers want more than these languages offer. With WASM you can pick up language(must compile to .wasm) that fits your use case best. This is the freedom most programmers want.

There will always be programmers who will draw their custom buttons(instead of modifying DOM from WASM) and ignore accessibility. They can do this with JS as well, but most of them don't.

Nice project! I love WASM. It's designed to be sandboxed and portable from day one. I wish WASM was invented instead of Javascript in the 90s. WASM will eat the world.

What I hope most is endurance. There are many programs that we are not able to run anymore. The best examples are probably older games. I hope WASM will change that, although I'm a little bit nervous about adding new features, because simple specs have a higher chance of surviving, but the future of binaries looks exciting.

Risks and problems with AI

One of the biggest risks right now is viruses created by transformers trained on known viruses and anti-virus binaries and databases.

Humans created relatively simple viruses which took down hospitals, etc.. Transformer can make new "super" viruses and keep fighting the anti-viruses/firewalls in real-time with high-quality custom attacks.

Does Nostr have a Wikipedia page? I'm wondering how old is it, the people behind it, some stories?

"we were pretty impressed with the overall performance on offer, despite being limited to a 65 TDP limit. Although power consumption on the CPU package was around 90 W at full load, this is still considerably lower than other processors the trio of 65 W Ryzen 7000 chips was pitted against."

This is from the Conclusion page. Still pretty impressive.

Sure developers have a hard time handling all that HTML, CSS, Javascript and connections(DOM, etc.) between them, but I see a much bigger problem in browsers. There are actually only a few browsers and most of them are based on chromium(or Webkit). Why? Because web standards are huge!

Their complexity is so big, that even great programmers have a hard time creating browsers from scratch. The latest example of this is probably Ladybird[0]. Andreas Kling and others have worked on it for years("Just under 1000 days for a bunch of hackers to build a new JavaScript engine"[1]) and they know it will require way more time to just catch up the Chrome[2].

A year ago I started to build SkyAlt[3], which doesn't have anything to do with the web. It's not just a browser, but also an IDE where you can create apps with few lines of code or just drag and drop stuff on canvas. It's written in C and it's only 25K LOC. Compiling takes a few seconds and binary is under 1MB. There are tons of features that need to be built, but I like its simplicity(relative to the web).

[0] https://github.com/SerenityOS/serenity/tree/master/Ladybird

[1] https://twitter.com/awesomekling/status/1596783757125898243

[2] https://twitter.com/awesomekling/status/1595387284035145729

[3] https://github.com/MilanSuk/SkyAlt

I was watching White Collar Advice[0] an hour ago and there was an interesting fact(25m30s) that she didn't work since December 2018. If she drove for Uber, it would probably influence the judge.

By the way, the whole video is interesting.

[0] https://www.youtube.com/watch?v=aFFxLwvGLhU

You have to login. Then on map there is "GPX" button.

I downloaded .gpx of the run and it's 6.7MB. He started with 4min/km pace, held it for 130km(9 hours) and then He "slowed down" almost linearly to 5min/km when he finished.

Graphics and Sound! Nothing is more motivating for children than something that moves, blinks and makes noise. That's how I got into programming in the 80s

Same for me, but in the 90s. Back then, games push PC hardware forward and made a lot of people code.

Today it's probably the 'metaverse' that converts consumers into programmers. Sadly, many think that programming is cool all the time(thanks Hollywood), don't realize the complexity and give up soon.

All HTTPS connections(most of web browsing last decade) are encrypted to and from the server.

Signal showed us that is possible to have E2E encrypted messaging, why don't they provide it, but with better UI/UX?

If Telegram doesn't read messages and doesn't provide access for 3rd parties(governments, etc.), It should provide E2E everywhere by default.

I wrote it in C with just one loop: for(i=0; i <= 10000; i+=37) and one condition: if(i > 0 && i%13==0 && i%19==0) and it takes only 0.000003 sec.

Is there any chance for another a16z Crypto Startup school? Their videos[0] from the last one(mid-2020) are still one of the best crypto/web3 resources.

[0] https://www.youtube.com/watch?v=2wxtiNgXBaU&list=PLM4u6XbiXf...

Do operation systems have APIs, so the program can send them what is on the screen? I mean just description like: There is the button with title "Push me" in the rectangle(50, 50, 100, 70), etc.

Or are there any "screen reader" libraries?

inefficient and costly

That is the best part of today's blockchains - huge space for improvements! I don't see the same opportunity in centralized services/protocols which are kind of "done" and control by few big tech companies with so many resources, which they use to keep that control.

Why first thing people think about blockchain is storage? I like ownership and transparency. Let me give you one example. Many HN readers complain about App store centralization, but what if there was decentralize app store, where code and data live on the blockchain and anyone can propose code changes, and then thousands or millions of users vote for what they like. If you think this can be done with a centralized server it can not because those who control the server have most of the votes no matter what. But it can be done with a transparent blockchain network.

The chart in the link says: Delta-V in km/s.

I remember watching "Everyday Astronaut" live with less than 1000 other people. Today, the numbers are over 100K. He is a great guy, who is very interested in and passionate about everything around space exploration and his articles and videos go very deep, the same for answering questions live. Only if more YouTubers were like him.

Can't wait for 2nd and 3rd parts.

His notes for 1st part: https://everydayastronaut.com/starbase-tour-and-interview-wi...