Wow, Hermes Agent makes GPT 5.6 sol feel like GPT 4.
We used to say, "And just imagine... This is the dumbest AI will ever be from now on."
We were wrong.
A morbidly obese harness can nerf an otherwise useful model into LinkedIn levels of cognitive dysfunction.
@tansawit@grok “i think multiple, composable agents are the right direction, but i’m much more bullish on a future where ~all communication with them happens through a single master thread.” Is correct. Call it mixture of actors.
https://t.co/6JyRWgQ8nR
@tansawit@grok That’s the approach we’ve taken with peerd, but we’ve gone further on the subagent structure to have what we call “actors”. They’re more specialized for a specific role, with a custom system prompt, tools, etc. The orchestrator you talk to creates and delegates to them.
Hermes has twelve browser tools. Browser Use mode replaces them with a single one, driven by @browser_use's CLI 3.0.
Instead of a dozen schemas in every request and a tool call per click, the agent writes a script. In our tests that cut token use 48-66% with no accuracy drop.
New in Claude Code: your sessions can now message each other.
Instead of having to re-explain yourself in another session, you can now tell Claude to do it. It sends a summary (not your history or files), and the other session picks it up mid-task.
Introducing Kitesurf: a browser built for agents, running entirely on Cloudflare Workers.
Chromium is too heavy to hand every agent one. Kitesurf is written in Rust, uses 3-7x less CPU and memory, and spins up per request.
Free in beta: https://t.co/Vm8kVCgSkG
cli agents are just larp, there is no reason whatsoever to not use a gui harness it’s better in every possible way. all of this nonsense is the vim vs vscode nerds of this era