wouldn't it make that much more dangerous?

yes... the only advantage of Hydrogen I can imagine is the abundance, for the rest I see only downsides. Helium is relatively abundant and technically renewable: it's a byproduct of gas extraction, since it's continuously produced by the alpha-nuclear decay inside the earth. Of course doesn't mean the offer is adequate with the market request, especially after the Hormuz crisis

Under this interpretation models themselves do not have copyright protection either

Funnily I think copyright protection for LLM model is 15 years. As to me it just looks like very fancy database and that is the protection for those...

Not a layer but a bit nerdy on the law but specially on the copyright side. Copyright is a framework to protect creative work. One could easily argue that an LLM model is just weights automatically produced from data and there is no creative work involved. The counterpart would argue that the body of training is itself a curated body that is reflected in the model and that the pipeline training is a creative process required to obtain the final model, like some artist work; But it feels like a weak argument to me.

But then again, let's stop pretending law in US has any meaning other than what the billionaires or the dictator wants.

> No one is seriously going to leave GitHub because of reliability

had a discussion with the team yesterday, literally every one wants to leave and looking for the best alternative

Wanting to leave is not the same as actually leaving. I can imagine small teams leaving because moving a 10-30 member team is easy. If you have a dev team with 500 members and the team has been on GitHub for 10 years, you will just take it because migrating is just a PITA.

> If you have a dev team with 500 members and the team has been on GitHub for 10 years

This so closely describes my company. We’re so fed up with GHA that we’re charting an off ramp. It’s been happening in the background for a few months already but whenever there is a major outage I swear we add one more person to effort.

Migrating sucks but so does having 500 engineers who can’t do any coding work for a day every {month,week}

I think I was imprecise in my last message. When GitHub actions has an outage, engineers at my company can and do still write code. No one can merge though and we aren’t able to deploy (like we normally do) so we very quickly get a huge backlog of PRs that need to be tested / reviewed etc. People generally just pivot towards reviewing code at this point (or doing other non coding work)

When GH has an outage like yesterday, we cant even formally review code. Sure, people could pull others branches and read them and discuss in chat but a dedicated tool for distributed code review is pretty important for most teams. It wasn’t that long ago we were a small company and GH’s reliability wasn’t such a nightmare so using it for what it offers made a lot of sense for the business.

The calculus has changed recently.

Yes, I heard that plans many times. Same as in another case, people have sworn to leave for Canada, if. No one actually did.

I'd be really glad if you do actually, but my hopes are weak.

i beg to differ, in an ideal world a system possibly is a binding law and high end models are starting to be really aligned to the exact system prompt. The instructions must be simple to follow, if you start doing complex rules it'll call apart, but I'll usually follow the stringer interpretation.

"I beg to differ, it is my opinion that reality should be different to what you have observed"

in reality even the mention of a prohibition is enough to make the model reject that no matter what

That’s just not true. There are bypasses that happen all the time.

the fact that sometimes someone finds a way to workaround the rules doesn't make them a ""suggestion"" they are very solid, just not absolute

or completely deleting the app and use the browser version for desperation

Browser version eats about as much ram sadly. It’s not only about the Chrome dlls but also about the memory leaks, needlessly loaded resources etc etc.

Disagree with the conclusion, even without carefully curated context every high end LLM perform just as well, maybe with an extra detour. In contrast if even one of the learnings is not up to date or doesn't apply to the current situation you find yourself with a long detour or even a failure.

Just go online rent a few servers, download K3, ablate it, run the thing, shut it down. Will probably cost a few hundreds bucks to ablate, but f* this non-sense paternalistic sh*.

I wish i had the time to do it and blog it, "in your face" kinda style.

FYI:

There have been reports of much smaller qwen derivatives being used for 0day research.

Yes, given a direction, they are pretty good at "fuzzing", trying out everything and eventually find something. Super useful, but it doesn't have the same level of precise targeting you could expect from a high end model.

there is a simpler way, make a complete mental model of the changes and ask questions to confirm your understanding. so much faster.