Gemini flash 2.0 experimental is the first model I feel that really generalizes over different modalities. It also feels like real reasoning even when it gets it wrong. It seems to think outside the box in some cases.

I showed Gemini a video and it recognized it was a spoof video. I asked how it knows and it gave very good reasons

@garymarcus.bsky.social has 25 predictions for AI. I have two for him. He will sit and cry about everyone else's work while he does not lift a finger to build anything. Those who cannot do cry. When AGI is reached this is how he will be remembered!

This kind of question is not only hard for many humans it is for LLMs as well. The new Gemini version gets it 5 out of 5 times now.

Gemini 1206 is a much better model all the way around, even vision is much better

If you are adding MCP tools to Claude and get the cant attach error. You may have a jacked up file system like me. Just add the direct path in the Claude desktop config. Example: C:/Users/shawn/node_modules/@modelcontextprotocol/server-brave-search/dist/index.js

Many of my network friends are scared over AI. The thing is software has been gunning year after year to turn routing and switching into a one button solution anyways. Software has been slowly been working its way in many areas of work. AI may just speed it up.

Gemini figuring out songs by putting images together