yeah i got an off by 1 and decided that was good enough :-D

freshrss will do that for you. It has a built in web scraper.

just curious.. are there languages that are better or more efficient to build LLM's with other than English?

For some definitions of better, yes. Chinese is more token efficient for representing fixed text, for example, although this does not always lead to better performance on downstream tasks.

Increasingly exhausted GenX'er here... I'm just happy to come into work these days. I no longer have any desire to push for promotion and early retirement is a very tempting goal right now.

My eldest is now 18 and starting to to into the workforce and my youngest is still about 3 years from that.

if (and thats a big if) we dont have a GFC part 2 in the next 10 years I'll be fine I think... but jeez Trump/AI is making it hard to get there.

so... AnythingLLM? https://anythingllm.com/desktop

but with extra steps

ha! cool my cheap and cheerful MG4 passes all these. (well... "Physical controls for temperature and fan speed." is mostly physical. so a .5 score for that I would say)

I have fond memories of spending long nights recompiling with new flags to try and get slightly better FPS in games... this was over 25 years ago so not in the proton heyday we have now.

These days I'm a fair bit lazier, throw Fedora on and use it happily. update frequently and it almost never causes me any issues.

The Gentoo Forums were a super fun and friendly place back then, I hope they haven't lost that spirit.

Gentoo has pre built binaries for years now - you can be totally lazy about that ...

And still benefit from packages/system being smaller/faster because they are not built to cater for all possibilities.

And you can still override options (use flags) and compile some things exactly as you want/need them.

It is super nice to be able to mostly use binary packages but build from source anytime you want different options:)

oh thats nice. maybe if i ever feeling rebuilding i'll look into it again.

but truly, I've been using this install since 33... and am on 44 atm. never have i has such a slick trouble free experience. I've had one nvidia related issue once... I also had to rebuild a mirror once because i wanted to update more than i wanted ZFS, and they were updating the kernel faster than the zfs package was updating. not a big deal. I've moved on happily.

it'll probably be a favorable event for SpaceX's IPO.

Surely not as favorable for the IPO as SpaceX’s own recent explosion and multiple engine failures?

Blowing up on the pad is incredibly worse from a design data collection perspective, a risk to life perspective, and a downstream impact to future launches perspective (nobody can use that site for a couple of months).

To be fair the last Starship to blowup on launchpad/ground was less than a year ago. It is a set back but it appears nobody has avoided this issue yet.

https://www.livescience.com/space/space-exploration/spacexs-...

That was on a test stand, NOT a launchpad.

Vastly different destroying each of those.

I will give you that. Better to blow up in a test rather than ready for launch.

not to mention 7 days before it was meant to deliver a payload to space... a proper commercial payload. not just a POC payload.

The entire point of SpaceX’s recent launch was an explosion. They were aiming for that outcome. They wanted that outcome.

The fact they did it with pinpoint accuracy even with engine issues and an in tact heat shield is a monumental success for a test flight.

SpaceX also had a massive explosion on the ground not that long ago.

Absolutely, they were running a test on a test stand.

For BO it’s much better to have this now when there is no payload or people on board so they can correct whatever the issue is.

If anything their one engine out unexpected but successful test boosts their position a bit.

After separation that turned into all engines out on the booster, so perhaps not.

I meant the engine on the Raptor. But yeah good point about the booster. However that's nowhere near as large of a set back as this explosion

Sat through a demo of a tool that now has AI shoehorned into it yesterday.

At one point i noticed that the demo had a column that showed how many tokens each AI query was using.. one used 250 THOUSAND (for a single analysis of a single device). I asked the company who is footing the bill for those tokens.. they say they are (*for now). I pointed out the rather high number of tokens being used and does that mean we get a budget or quota or is it all you can eat.

they said it was a good question and they'd have to find out... and then one of the execs said OUT LOUD "Why is that column even there? Why do we show this to the customers"

I lol'd. It wasn't taken very well.

>"Sure they can, we can just ask them."

I'm convinced this is in large part because of the anthropomorphism of AI.

AI is not people, the moment you start treating it as people you start believing and thinking of it as people and you now take the human safety check out of the loop.

If you take the human out of the loop you have an unsecured weapon/dangerous object on the loose. anything from excel spreadsheets through to actual weapons. it all needs a human in the loop.