Point being that Facebook does have something at stake here, and will pay a price if it does not comply with government directives. Which is presumably why they are preparing to comply.
Artificial Intelligence thread
- Thread starter 9dashline
- Start date
Point being that Facebook does have something at stake here, and will pay a price if it does not comply with government directives. Which is presumably why they are preparing to comply.
alright, I tried running qwen3.6 27B locally and it worked fine. had a nice web GUI which I could use. I'm happy with it.
People who don’t know what they are doing shouldn’t use agents.
The risk of dependency.
The risk of dependency.
my comment on this. It seems to me that these firms were always going to purchase Atlas clusters, but now that DeepSeek did the hard work of getting 950DT to work for CPT and post training, it gives them more desire to do so.
from someone that actually uses Kimi 2.6 for real work & business. This is the kind of stuff people should listen to about certain models instead of a disgruntled former ByteDance employee. I use K2.6 for my normal programming and it's great.
my experience with K2.6 is that it pretty much does everything I need it to do. The only issue is that since I'm using the free online version, it often runs out of token when I request it to do complex tasks. I guess I need to start paying money.
my experience with K2.6 is that it pretty much does everything I need it to do. The only issue is that since I'm using the free online version, it often runs out of token when I request it to do complex tasks. I guess I need to start paying money.
The most ironic thing about that interview is that he criticizes Chinese models for benchmaxxing while praising Google's culture. But no one uses Gemini for coding or agent tasks despite its high benchmarks.
Last edited:
is there any reason to post this? If we disagree with their clownish views. What is the point of posting stuff like this except working people up? I'm deleting anything related to this.
A/B testing and limited to images. Unfortunately, I wasn't given the opportunity to test it.
Based on current information, the visual model cannot perform online searches, and its error rate isn't particularly outstanding, but very fast and its world knowledge seems very deep. I'm curious about how they achieved this.