Sufficient. I’m using cloud code heavily and just like you hit
Limitation of the mental overhead of using context reconstruction mid-session
Anywhere else is cruel – you spend the first 10 messages simply
Grabbing a new agent instead of actually solving the problem.
The MCP angle is the part that makes it different from the bus
“Searchable Logs.” Agents interrogating each other’s past
Actually a different model than cold start. curious how
Recovery quality persists over longer sessions – does it come to fruition
Reference to the true past or do you find you still need it
Point it to the correct conversation manually?
Apart from this, prompt.toml injection is also underestimated. maintain continuity
Agent switch has rules and personality without copy-paste
Something I would use every day.
<a href