Let me guess, by looking at how bloodshot they are?

Will "Generative Engine Optimization" work to coerce AI-Answer-Engines? The models underlying them are constantly changing, and along with them, AI’s responses also change. AI search engines answer the same question differently each time, depending on how and when they are

This person used Hailuo AI (MiniMax)'s latest LLM to clone X. Very impressive and nice work. I wonder how long it took and how many revisions were made?

The new Chinese open-source long-context language model performs quite well, Hailuo AI (MiniMax) (the same company that brought you Hailuo v2) beats Qwen 3 in these benchmarks. Has anyone done a needle in the haystack with it?

I personally can't wait to see a humanoid robot assembling an Nvidia server stack.

I think Waymo is the "gateway" AI. The more normal we find Waymo, the more likely we are to give up our agency to AI.

Huawei just released some new models. The new 5.5 Pangu models are based on the flagship 718B deep thinking model, which is a MoE model consisting of 256 experts. It has been drastically enhanced in knowledge reasoning, tool invoking, and mathematics. They also have a Pangu World

X to offer investment and trading advice. Once AI co's realize the value prop of AI to users is limited and exploitation of its users for advertising revenue won't bring in the revenue they expect, they will recognize that cannibalizing other businesses is a better way to make $

Do we know what the record for the number of experts in a MoE model? Is 256 a lot?

Transhumanism is just starting.

Feels like we are living in the future. First, Elon shares how Neuralink will help people see in the infrared and with radar, next flying Taxis.

Mini-battery sipping (milliwatt) 5G transceivers.