They can recombine. They can’t create new knowledge or reason in new ways.
RESEARCH
-
Multimodal image input model version in development
By
–
I have a version that does that for the image input models, I hope to write that up soon
-
Qwen3 4B Thinking Model Achieves Fantastic Pelican Benchmark Results
By
–
I've even seen some of them wildly over-think on my pelican benchmark – got a fantastic result from Qwen3 4B Thinking yesterday
-
LLMs Becoming Too Agentic By Default
By
–
I'm noticing that due to (I think?) a lot of benchmarkmaxxing on long horizon tasks, LLMs are becoming a little too agentic by default, a little beyond my average use case. For example in coding, the models now tend to reason for a fairly long time, they have an inclination to
-
LLM Security Challenges: Protection Against Emerging Attack Classes
By
–
Right, that's why the stuff is such a massive nightmare: so many of the things that we want to use LLMs for can't be done securely in the absence of a completely reliable protection against this class of attacks – which so far does not exist
-

GPT-5 Achieves 70 Score on Offline IQ Test
By
–
GPT-5 scored 70 on the offline IQ test. via Reddit, h/t AloneCoffee4538
-

Google Brain and DeepMind as Training Grounds for AI Leaders
By
–
Almost everyone great / relevant in AI was at google brain or deepmind at one point (or still are).
-

GPT-5 Controversy: What Went Wrong and What’s Next
By
–
GPT-5 caused quite a stir—and sparked a very controversial discussion in the community. It is questionable whether the new model has lived up to expectations or not. I sat down and traced the events. How did the controversy arise, what went wrong, and can GPT-5 now be
-
Rethinking Agent Architecture: New Design Paradigms Emerge
By
–
If you're building agents, now is the time to rethink the architecture. Checkout the Paper here:
-
SLMs Improve Rapidly Migration From LLMs Easier Than Expected
By
–
A few takeaways: 1. SLMs are improving rapidly, even at the same parameter size 2. Migrating from LLMs to SLMs is easier than expected