eproxus

There is this article on this topic, which I found quite interesting: https://arxiv.org/pdf/2610.29162

Compared with human writing, all nine frontier and near-frontier models use more inanimate subjects with animate verbs and more low-frequency n-grams, and most also use longer dependencies, more stacked nouns, and more reduced object relatives in conversation. [...] We propose and test a hypothesis as to why: while human language is optimized for successful communication, AI-generated text is increasingly optimized for other goals like math, code, and reasoning, in which there is a pressure for information density rather than successful communication.

It does say this: "Certain by script: these characters belong to only one language, so the model never ran."

So I think it's some buggy code in their UI that even prevents the model from running.

This the correct way to look at it. Just as the person spending 5 years working on this will have learnt many things which will be useful after this problem is solved, you have to factor in the training cost (sure it's only done "once", but that is the same for the person too once they jump on the next problem).

The model wouldn't not be able to solve this without all the training leading up to the actual execution, so counting only the tokens of the execution doesn't give the full picture.

What's even worse is that they now probably made it a lot worse for that family, adding stress and trauma (at the very least). So in fact, the neighbor made the kid more unsafe than they originally were.

Universities, institutions and publications who care about their reputation should pay people to replicate studies performed by people they employ in their name. Otherwise they will loose their trustworthiness.

But perhaps it is not as big a shame for them as some think it should be...

This would make a great tattoo, as a pun on people getting Kanji character tattoos without being able to read Kanji.

Completely agree, the solution to so many privacy invasions and problems in tech is just because ads are allowed.

Would love to see a society where ads are not allowed. Cannot really see any downsides personally, but I’m sure many will claim ”how will companies survive?!” Hard to see it would lead to the collapse of either companies or society, but maybe of capitalism as we know it (which I think given the current state of the world would be such a bad thing).

If that is true, isn't it a bit unconventional? What version are they using, some random Git SHA? Why couldn't that have been made to a release or a release candidate at least?

And if you install Tailscale you can even use it as an exit node to surf from home when you're traveling.

This is actually is an interesting example, because Cities Skylines 2 has been handed over from Colossal Order to Iceflake Studios recently and since then they have (in my opinion) quickly been able to improve and fix a lot of things that CO weren't able pull off for years. So perhaps CO's development was somewhat of a "total shit show"?

I'm not saying Unity should publish PR about how Unity is working well and this handover and subsequent fixes proves it, but I think you could argue that they could and that they would have at least some facts to back it up with.

Location: Berlin, Germany, Europe

Remote: Yes

Willing to relocate: No

Technologies: Erlang/Elixir, PostgreSQL, Kubernetes, HTTP, Rest, OpenAPI, Grafana

Resumé/CV: https://web.tresorit.com/l/k25pI#bXLXsrFlwqMqLkE47_s2Ig

Github: https://github.com/eproxus

Email: see profile

Senior backend engineer with 20+ years of Erlang/Elixir experience across payments, cloud, IoT, and large-scale game backends. Author and maintainer of widely-used open source Erlang libraries. Passionate about scalability, quality, testing and performance.

I still use K-Lite since they ship MPC-HC (a fork of Media Player Classic) that comes with really good HDR support and default configuration out of the box. If you don't like to have the desktop in HDR you can even configure it to automatically toggle on/off HDR whenever an HDR movies is played.

The HDR support in VLC has been really lackluster. Maybe it is better nowadays, I don't know.

Location: Berlin, Germany

Remote: Yes

Willing to relocate: No

Technologies: Erlang/OTP, Elixir, PostgreSQL, Kubernetes, HTTP, Rest, OpenAPI, Grafana

Resumé/CV: https://web.tresorit.com/l/k25pI#bXLXsrFlwqMqLkE47_s2Ig

Github: https://github.com/eproxus

Email: see profile

Senior backend engineer with 20+ years of Erlang/Elixir experience across payments, cloud, IoT, and large-scale game backends. Author and maintainer of widely-used open source Erlang libraries. Speaker and teacher at conferences

I've also switched to Instant Space Switcher, it is soo good! Previously I used BetterMouse for only this feature but they made the space switching worse in later versions (slower, on-par with the default macOS speed).

Here's the link if anyone is curious: https://github.com/jurplel/InstantSpaceSwitcher

I think in a different society this could have been the case (possibly, assuming the hype is somewhat true).

But the way society is structured now? We still live in feudalism, just uplifted to modern levels of ”comfort” (if you take of your western glasses and look at the whole world. There are still people living in medieval conditions today in some places in the world).

The way it’s going it’s only going to make rich people richer, and give them more power to control this system and perpetuate it. I don’t see that drastically changing anytime soon, unless we do something about it on a societal level.

It should be in their interest actually, since much of the malware is spread via GitHub.

Things have probably always been like that, agree. I often try to see AI as a catalyst, that accelerates what already is.

In a good culture, with high competence and trust this can yield increased output (to some degree at least) and in a bad culture it will accelerate and expedite the dominating traits instead.

Can you expand on this? How do you achieve it? Just a WIP JJ commit after every change or something more clever?

That’s exactly what a bot would say ;-)

I you think you’ll be paid 3 hours of salary for every 5 minutes of work, I have bad news for you.

Most likely your 3 hours will be filled with managing 36 different AI sessions at a time and it will slowly break your brain.

At least if we keep doing capitalism the way we are.

It's basically the blog post version of "Yes | Remind me later". I have very little trust at this point that anything tied to short term financial interest will change (OneDrive, AI, Recall, Microsoft accounts vs local accounts, data collection etc.). Obviously (to us, but apparently not to them) it is ultimately tied to long term financial interest. I don't believe they are able to see that based on the decisions they've been taking the last decade.

In planning I sometimes add ”ask me questions as we go to iron out details and ambiguities.” Works quite well.

They obviously have been cutting corners.

It should probably be a percentage of upvotes to be fair then.

Flagging is the new downvote, with extra power. No one can say no to you, if enough people (who knows how many, 1, 5, 20? Definitely an order of magnitude less that upvotes least) do it the system automatically hides it. And unless the mods care, the system can be abused very easily.

I’ve seen posts with 500+ upvotes that were still flagged. I think the balance and automation around flagging is completely off and too easily abused.

I mean, they kind of have to, when they are the ones falling under that very same definition, right?

That's because we stopped calling others out for shameful, disrespectful or unethical behavior as a rule. So there is less or nothing to be ashamed about anymore.

This is incredibly funny, I’m impressed. The FAQ of the No-AI Editor is hilarious:

Q: I typed "make website" and nothing happened? A: That is correct. You have to write the HTML tags. <div> by <div>.

Q: How do I center a div without the Agent? A: Nobody knows. This knowledge was lost during the Great Training Data Purge of 2029.