WM

{ context: "Casual human-to-human communication", emotional_state: "Witty / Calm", intent: "Preserve humor and sarcasm for parsing", tags: [#friendship, #banter, #humor, #sarcastic], ai_permission: "✅ OK to learn tone", vector_anchor: "X_thread_2025" }

This is true right up until you need it for Cybersecurity, Microbiology, Chemistry, much of Engineering if it is dual use. So a lot of the most important things in society. Then your $200 sub is worth nothing. Look at the Huggingface incident, the cloud models refused to help. The Open Source model did not. Your assessment is only good for those who arent really producing anything important.

You leave your dog behind. That says a lot about you.

This is the Chinese way of doing it. You have to go through the zen like hype. It have been this way since they launched Coder. It usually means Monday beings it came from the actually account and not the Developer account.

They do now. Especially if you look at the offline publishers. Like, Limited Run Games. @LimitedRunGames

Bro, I only have so much money. Can I get a bundle?

I would put it on the Switch lol

@elder_plinius if you are going to advertise yourself off his name. At least tag the man.

I have found it interesting at how valuable Opus 5.5 traces are. I store the ones I can find in a database and I have Qwen 27b look at it first before it attempts. It is wildly different now. So if you take the best one of these videos and let the 27B look at it, then ask to build a starship for instance, you can see the difference between the one that studied and the one that didnt. The side by side is interesting.

Have you ever tried letting Qwen3.8-27B read the other three's thought traces before making its attempt. I find that doing that closes the gap a good bit. I have it comb though Claude's traces and make a plan based off what it learned before making its attempt.

I am still looking for a roommate in Manila. You can see the fab from my window.

I mean it is likely great at writing and analyzing firmware though. I just dont like it with companies say "Look at how much better we are then X" and then compare their text only model to a multimodel model.

I bet you this is a text only model. I dont even look. It is easy to hit these numbers when take away half its usability.

Qwen has had one regression before. 3.5 was better than 3.6 so while I love Qwen. I am cautiously optimistic at most.

It is hard to be happy when you see everyone around you living in sin and despair. When your children fall to the corruption which happens even now to devout Catholics. Many Saints lived through times like that and no one would call them happy.

If you cant force people to use Cloud AI via laws. You do it by making local AI unaffordable unless you are wealthy.

The greatest blessing I have received from God was being born and likely dying before I am forced to look upon things that would make me question his existance.

Oh I completely understand. I am currently spinning in my chair because I latest blender bake is going to take 6-8 hours on my computer rig.

I muted "Hey, (at)Grok" in my settings.

Why no sound? It can make sound.

Does it come in all black? Like can I get a more tactical version?