No matter how capable a model is, if it isn't coherent enough, I can't steer it where I want it to go. Neon solves this. Because of the way it talks to you, you immediately know when it's right and when it's wrong and since it's genuinely smart, it's usually right. @Google has really done something special here. My problem with other "really smart" models is that they make me feel dumb while I'm using them. It creates a huge disconnect where we aren't on the same page, and the longer the session goes, the worse that gap gets. Writing software has become cheap and fast, so if the end user doesn't feel in control, a model loses its value. That's exactly why Sol and Astra 6.2 suck. The biggest challenge ahead of us is the interface and coherence problem, and it looks like Google just solved it. @GoogleDeepMind@OfficialLoganK
@rodydavis@TokenGremlin 1. Computer use built in
2. Better mcp plugin support
3. Proper usage indicator for the session with tools showing prompt cache and tokens spent with tool calls breakdown
4. Allow use with lid down
5. App shots
6. the entire chrome dev tools in the inbuilt browser