I worry that GPT 5.6 will be heavily restricted and have the same feature to fallback to another model like Claude fable 5 does all too often. That fallback shenanigans mess up actual benchmarks and I don't like it.
The biggest loss to them is the right to repair stuff. They will be still making it exceptionally difficult to repair their stuff, and might even dip into exotic materials to make cheaper parts fail more often, but this is a bigger loss to them in the long run.
Unfortunately, I hate that they got away with such a low AF fine.
I ran it through paddle paddle OCR and it flawlessly did it.
Google's OCR through my phone's Google lens had also worked at getting a very good extraction but not 100% correct. Definitely would spend less time fixing it than hand copying.
IDK what the author was using but I feel like he could have shared how his OCR attempt went, but I am thinking he tried some naive OCR tools.
Plus, it is not the bottom I fear, it's the precedent from letting companies slide down the slope.
Regulation may try to stop it but history has shown some have slid to the point of no return or past a point where people can care enough to build out of.
Prevention is better than retroactively fixing stuff.
I've been seeing LLMs act lazy from the very beginning. They got a little better but smaller models really only want to have a single task given to them. Mythos at least does work. RIP
I am finding Chinese models are introducing more guidelines against cyber. Especially Kimi k2.7 code seems to have extra training against cyber security capabilities. Last one, k2.6 was a lot stronger at cyber but obviously the Kimi team improved over time, so this is not the best they can do but no one will be able to get the best anymore.
I expect future Chinese models to introduce even more of this type of bogus "safety" training.
Looks like if you are a white hat, then you will be fighting an uphill battle. Black hats will be fine, they will not care, they can just run a heretic model or specialty trained model.
I believe it is because GLM 5.2 has extra anti-cyber training instilled in it. Similar to Kimi k2.7 code.
Deepseek v4 pro being in preview with less "safety" training makes it stronger for that reason. Thinking will be different and in the end, it will actually try to be useful. Just expect future Chinese LLMs to further push out "safety" guided LLMs. The future is bleak for open weight models. Prepare to have "guidelines" enforced unceremoniously to all.
AI consistently places animated objects behind a blur object which causes the browser to constantly repaint. Google's ai mode introduced one, some other websites clearly vibe codes included them too.
At first it confused me why my GPU usage spiked and fans started blowing harder, but now I see it is a common mistake that AI makes but no one tests properly. It is possible a human can make this mistake but I sheldom experienced this ever in my life until now.
I run 240hz monitors, and that meant the browser was trying to do 240 repaints per second. Blocking it with unlock origin is the only way. Ridiculous
When ctf organizers attempt to make a challenge "harder", I find they push the challenge into a more "guessy" state. Instead of proving skill, you basically need to guess some obscure or random step in the puzzle that the challenge is meant to give you. It is one of the most common problems with any puzzle based challenge system.
Yeah, but we have AI now, we don't need our blog posts to over explain or state what it all means to general audiences.
The author name-drops a bunch of CTF events hosted by a variety of independent organizations and name-drops well-known teams.
To help everyone, this Capture The Flag is specifically Cybersecurity adjacent, there is a Wikipedia article on it as the top Google search result for me when searching "CTF". This is why the acronym is used, because searching for the full will get you to the wrong "sport" vs the cybersecurity one.
I don't want to explain what a CTF is. look at the Wikipedia article. It is there for a good reason.
Just look at deepseek V4, this preview model uses only 8 GB for 1M token KV cache(the context). It's insanely efficient already. It's just that most models that are coming out are barely catching up with technical breakthroughs.
Deepseek are pioneers.
Unfortunately V4 is not trained for most real world usage, it is mainly for world general knowledge.
I have a 1660ti and the cachyos + aur/llama.cpp-cuda package is working fine for me.
With about 5.3 GB of usable memory, I find that the 35B model is by far the most capable one that performs just as fast as the 4B model that fits entirely on my GPU.
I did try the 9B model and was surprisingly capable. However 35B still better in some of my own anecdotal test cases.
Very happy with the improvement. However, I notice that qwen 3.5 is about half the speed of qwen 3