Saying goodbye to Fable 5.
But I read recently that using 2 Opus 4.8 agents together is enough to beat a single Fable 5.
The way to do it would be to send a prompt to a panel of 2 or more models in parallel with the same tools/skills enabled.
Then send the results to a model that evaluates and lists the findings in common, discrepancies and unique ideas.
Lastly, send that response to a model to summarize a final answer.
I already do something similar and manually between Claude and ChatGPT.
And there is this platform that does it for you: https://lnkd.in/gEC7T3HB
