@Worshipperfx MCP being designed for agents means:
- you can include natural language guidance in responses
- you don't need to version your tools
In addition to things like MCP apps (inline UI.)
You still want an API to be used by deterministic code.
@Meaningness@St_Rev It's like all Zelazny, genius ideas mixed with smoothly deployed nonsense and hackwork. He weaves archetypes together the way a lesser sci-fi author might design a giant mech.
Great to see! Having another SOTA lineage benefits everyone. Have been looking forward to showing the new AutomationBench leaderboard with Gemini in the top 5 spots: https://t.co/xGicriaKwd
Introducing Gemini 4 Neon! Pushing the frontier across domains and punching above its model class in a meaningful way. To me, Neon feels like the right balance of intelligence, speed, and usability. Available now across @GoogleAIStudio, the API, and more!
@JakeABoggs I agree that we can get superhuman performance in some areas, the main thing I want to emphasize is that building capabilities in new domains still requires data or RL for that domain.
@frances__lorenz Is anyone working on non-existential AI risk? LLMs judging resumes with non-white names more harshly, that kind of thing? Someone must be but it seems to get 0% of the coverage x-risk does.
@nicbstme I think there are still interesting problems around having agents collaborate when they're in different security boundaries. You want your agent to talk to your vendor's agent, or your personal agent to talk to your work agent ...