Biggest problem with AI agents: you can't slap them when they don't obey
Seriously this shit keeps doing something I ask it not to do, waste of money and time.
To be fair sometimes it's great.
But we are SO far from autonomous agents. Supervised by a human who knows his shit, yes.
Certainly! [...]
I apologize for the confusion [...]
makes me want to kick it in the nuts
the sequential thinking mcp has helped reduce this, but still an issue.
This available in goose?
I haven't tried goose, but it says goose extensions support the mcp protocol, so it should work.
https://github.com/modelcontextprotocol/servers/tree/main/src/sequentialthinking
