Five Things AI: Energy, QwQ-32B,Pentagon, Benchmarks, Vibe Coding

All you need to read about AI this week.

Heya and welcome back to Five Things AI!

Remember how DeepSeek shook the AI world a few weeks ago? Alibaba just released QwQ-32B, which kind of sounds like Elon Musk would name one of his kids, but it is actually a new LLM that uses even less compute requirements than DeepSeek. Obviously, it’s a Chinese model and this will heat up the LLM race even more. Also, did you know that the Chatbots are showing some human behavior already? They cheat on their benchmark tests. I find this very funny. Not so funny is obviously the idea that AI will be used in combat pretty soon. Whoever has the best models and the best AI Agents will succeed on the battlefield of the future. For all of this, we will need lots of energy. And in the future, you can code as long as you can tell some software what you want to build. I played around with vibe coding last weekend and it really is a lot fun to code without typing.

“Any sufficiently advanced technology is indistinguishable from magic.” - Arthur C. Clarke.

Read on, my dear!