@jpthor Add more hooks, watchers, and sentinels.
Deploy them to deal with problems that surface persistently.
Create an agent to count your errors, categorize them, and enumerate them for you.
Create another to count the errors of the other agents.
Compile the data.
Then, fix it.
@jconorgrogan It's a language aggregator. Feed it the same phrases over time and it will say anything. This is not evidence of anything other than your ability to train an LLM.