Wer kein Open-Source einsetzt, ist nicht wettbewerbsfähig. Auch kommerzielle Software verwendet Open-Source Bibliotheken, i.d.R. hunderte davon. Diese Komponenten neu zu entwickeln würde schlicht niemand bezahlen, weil ein anderer Anbieter, der auf Open-Source setzt, nur einen Bruchteil des Preises verlangen müsste.
Letzte Woche kam ein Kollege zu mir und meinte, er hat jetzt einen Meeting-Raum reserviert, den er sich von so einer Übersichtskarte ausgesucht hat. Dann habe ich gefragt, wo man das im Intranet findet, und hatte schon mit einer Wegbeschreibung gerechnet, wie genau man durch die Links navigieren muss.
Daraufhin hat er geantwortet – immernoch sichtlich perplex – dass er das tatsächlich über die Suchfunktion gefunden hat. Der Kollege arbeitet seit 2011 in dem Laden und das war eine der wenigen Male, dass das funktioniert hat.
Natürlich musste er mir dann aber das exakte Suchwort nennen, mit dem die Suchfunktion die korrekte Seite ausgespuckt.
What makes sense to me, is that unlike capitalism, communism requires a government to function. Well, and how do governments fail? By turning into a dictatorship.
I know that QMMP has a built-in visualizer, and the webpage says that the visualizer is call projectM, which you can apparently also run standalone: https://github.com/projectM-visualizer/projectm
Ah, the post is written by the president of Signal, that’s why it keeps raving on about it.
But yeah, I think that’s a bit optimistic of a timeline. I can see that investors might be fed up with financing the unprofitable AI technologies next year, but then it’ll start with increased prices, which will lead to broad adoption of cheaper, worse models, which will lead to a generally worse image of LLMs and then possibly 2026 or 2027, we’ll see LLMs being taken out of all the webpages where they weren’t terribly useful. I do think that’ll hurt Big Tech, because they set quite a lot on this one technology this time around, but most do still have their cash cows to fall back to.
But yeah, we’ll have to see how this actually plays out. There is some use-cases, where LLMs are genuinely the right tool and companies might pay quite a bit of money to use them.
One critique of the article:
and even tech titans such as the VC powerhouse Y Combinator, which is singing in harmony with giants like a16z in proclaiming fealty to “little tech” against the centralized power of incumbents.
That’s kind of the job of Y Combinator. They find small companies, invest in them, then hype those up. As such, they will sing the praises of little tech, but very much with the goal of turning it into the Big Tech of tomorrow.
I thought I heard so, too, but when I tried to research it, all that came up is that you can publicly hide your phone number and instead give people your username, but you still need the phone number for sign-up. I really do not know, though, if search engines are failing me again…
> yellow is ugly
> 🇸🇪
???
Hui, bekloppte Frage, aber was hält ihn eigentlich davon ab?
Die müssen natürlich verkaufen wollen, was man aber mit einer groß-genugen Geldsumme bestimmt bewältigen kann.
Und theoretisch könnte die Wettbewerbsbehörde etwas dagegen haben, aber jetzt ist ja bald die Trumpete an der Macht, mit der die Muskete regelmäßig heiße Liebesnächte verbringt, also wird die wahrscheinlich auch nicht dazwischen gehen.
I’ve definitely also thought about, if our government gets taken over by fascists, how do you organize a rebellion?
And yeah, Signal definitely has some weird fucking shit going on. As far as I’m aware, they don’t allow you to use their centralized servers, if you don’t use their provided build of the app. They don’t seem to have a mechanism to enforce that, so you could still use a self-compiled build, but if all your friends are on a compromised client, you can’t talk to anyone anyways.
Well, and then there’s also the great stupidity that Signal requires a phone number. In my country, you can’t sign up to a mobile phone plan without revealing your full identity. If the fascist government realizes that I’m part of the rebellion, they can make my phone number disappear in unfortunate circumstances.
So, yeah, I’d at least want to self-host the communication platform. I’d probably use an existing open-source solution, but would try to audit at least part of it…
Hat gestern auch eine Kollegin gemeint, dass man bei einem der Streaming-Anbieter eigentlich ein total gutes Angebot bekommt, weil man 100€ im Jahr bezahlt und dann aber so viele Accounts machen kann, wie man will, also die Kosten mit allen Freunden teilen kann oder irgendwie so.
Und in meinem Kopf direkt so, häh, aber selbst wenn du jetzt neun Freunde hast, die ihr Geld beisteuern, dann bezahlst du ja immernoch 10€ im Jahr???
Reorganizing the menu takes maybe two collective working days until all discussions have been had. Impacts all users.
OLED black theme, probably takes about as long, with accessibility testing, and getting the settings entry translated into all the languages. Impacts the 1% of users, who bother looking through the settings and use a dark theme on an OLED screen.
Integration of theming add-ons takes more than a month, quite possibly more than two. And this is taking into account that they do already have a rough theming extension API (for desktop) in place. This is also a task that you can’t give to the people who reorganize menus. Again, impacts maybe 1% of users.
I get that you don’t feel like reorganizing the menu was necessary, but it’s really not like you’d save the world in that same time…
Today, a colleague couldn’t do docker login for an internal registry. Constantly got an error which just said “unauthorized”.
The password couldn’t be the problem, because you actually generate a token on the registry webpage, so we tried all the different ways to spell his username (uppercase, lowercase, e-mail address) and tried different URLs for specifying the registry, tried toggling the VPN, a reboot etc., even though we knew what should work, because the login worked for me.
Eventually, we gave up and figured there must be some permission problem in the registry. Ten minutes later, he tells me that it works, without doing anything different. Now I’m wondering, if the IT saw our desperate login attempts and quickly fixed the problem. 🫠
There’s no real mobile app for it, is kind of my personal main reason why I didn’t pick it up…
I would be surprised, if you couldn’t also sell an account with a bunch of stars on the black market to bot farm operators…
Damn, I wonder what was going on in his head. Was he trying to get to Saltlake City? Or did he just go for a walk and got lost, then tried to hitchhike back home in the wrong direction?
It won’t rise much beyond that, since you only get one update per package. Whether it’s upgrading Firefox from version 120 to 121 or to version 130, it doesn’t change much in terms of download size, nor the number of updates.
At least, I assume, Arch doesn’t do differential updates. On some of the slower-moving distributions, they only make you download the actual changes to the files within the packages. In that case, jumping to 121 vs. 130 would make more of a difference.
If you do want lots of package updates, you need lots of packages. The texlive-full package is always a fun one in that regard…
A few months ago, I got myself a keyboard which specifically says “Meta” on the meta-key, because you know, I didn’t want a Windows logo and I decided, I didn’t care about Facebook trying to claim that word. I definitely don’t care for Musk trying to claim that letter either, but damn, the combination of the two might do it that I’ll have to pick a different shortcut…
I’ve seen doormats that look more alive than this lawn…
I mean, different regions have different plants which they call “grass”, not to mention different climates. It is genuinely possible that even grass in the wild goes brown in that region…
There is somewhat of a development towards modularity going on. There is now the agent-based approach. As I understand, you run multiple sessions/discussions with the model and give each session a smaller, simpler job. So, you might tell the first agent/session that its job is converting file types, that it should convert from XML to JSON, that it should only output the JSON and so forth. And then the second agent you tell that its job is extracting data from JSON, that it should read the value of the field “tomato” and output only the value. And then you write some shoddy script, which pumps your XML into the first agent, then takes the response and pumps it into the second agent and the response of that is what you’re looking for.
And then you can individually optimize these agents to have as much of a success rate as possible, by repeatedly tweaking parameters and running them with all kinds of input data and measuring whether your changes made it better or worse.
Having said that, for the modularity of what we do in open-source, you need effectively 100% reliability. If you combine ten systems with 99% reliability each, you end up with 0.9910 = 0.904, i.e. 90% reliability for the whole system. It is brutal. And to my knowledge, these agents don’t typically hit 99% either.
We’ll have to see how it works out. You will probably need fewer agents than you need libraries. And of course, you could combine these agents or LLM models in general with traditionally developed software. If 95% reliability is fine for you, then the LLM only needs to be 95% reliable, so it could take over the input-output, for example.