tjungblut

Great, instead of backpropagation, we now have slopagation.

yep, it seems this was trained on large amount of podcasts with ad jingles or phone call queues with elevator music. I was also pretty disappointed to run the TTS last week.

Sorry Pavan, I'm happier with Fedora Atomic and Bazzite now.

I wonder if we can do a prompt injection from the comments

I'll probably get DV to oblivion for this, but I have to constantly wonder where those parents come from that need to forbid their children to roam freely on the internet.

Didn't they grow up in an age unrestricted web either? By now we must have two generations of unhinged children grown up with unsafe World of Warcraft, MSN, Whatsapp and ICQ. Oh and the p0rn... I mean, seriously, do you guys have nothing else to do than to moderate your kids Minecraft servers?

too bad that ubitricity was sold to Shell

If you are curios, like me, how the actual reinforcement learning happens. It uses verl [1] underneath. The paper "HybridFlow: A Flexible and Efficient RLHF Framework" [2] explains it really well.

[1] https://github.com/volcengine/verl [2] https://arxiv.org/abs/2409.19256v2

tldr, they now do a zero (?) copy of raw bytes instead of marshaling and unmarshaling json.

thank you for referring, feel free to ask any questions you may have

not entirely, but it's maybe 1-3% over a week depending on the outside temperature

Get a flipper, you can simply download and emulate all of them

The watch is a simple processing loop that receives and sends on a bi-directional GRPC channel. Leases have a similar loop for keep-alive messages, everything else is quite literally delegated.

I get that it's difficult to translate this 1:1 into Rust without channels and select primitives, but saying it's complex is wild. Try the server-side code for leases ;)

could you be more specific on what's so hard to follow? it's quite literally just the implementation of the GRPC interface [1].

[1] https://github.com/etcd-io/etcd/blob/main/client/v3/kv.go#L3...

Inverting a binary tree became implementing SVD with arrays only.

Your generational approach to segment numbering is fine, if you prepend enough zeros to format the files properly then you're also able to sort them correctly. etcd uses the same trick.

Besides what Phil mentioned below, I can't write more than one record to the WAL. You're closing the file after every write, the second time you write the error `seek data/rebuf.tmp: file already closed` is returned.

I also think your rotation will delete the wrong segment when you have more than ten segments - imagine you're writing rebuf-1 to rebuf-10 - what's the "oldest file" to delete now? Besides, should you really delete those files?

and why would the German red cross own it and sell it on Ebay? Ebay wasn't even a thing back when the DM was still around.

This seems like a small CRUD app to replace apiserver/etcd with a nodejs app and mongodb.

Two gaming Hatters from Austria? Of course it had to be Sanja and János. I will check your game later on deck, but congratulations on launching :)

Yep, that was for permanent loss.

I had a dentist explain to me that my lip could go numb with a 50% chance after my wisdom teeth removal with that exact explanation just two minutes before the surgery. I was infuriated, turns out I was on the lucky side of her coin.

Gameboy was THE mobile first platform when pokemon first came out. I see no issues here.

Odd, on my S22 it's just showing a black screen.

When parsing the proc files, many people forget that process names can have spaces in them and that causes some very funny outputs. The Python script in this posts handles those correctly, kudos.

Just tried to sign up, got an immediate rejection.

I don't immediately find it, but couple of years back there was a "meta-feature" which was the size of the MNIST image. I think that scored about 90'ish % accurate results on its own - without even looking at the image.

How to you think trucks would cross the channel to deliver goods? Swap the steering column in the tunnel?

I vividly remember cracking the admin PWs with ophcrack or the school WEP wifi with aircrack-ng.

Indeed. For me the most infuriating part is the amount of low-quality questions that could easily be answered by AI before submitting the question. They definitely should embrace AI and generate an answer. If you think that question isn't properly answered, include why the AI proposed solution is wrong and let humans answer it.

As for the correctness of the solution: that's why the voting system and tickmarks are there. Wrong solutions would ideally be downvoted and never marked correct, I don't really see how AI is making a big difference here. Moderators today aren't running the answered code either.