In a podcast on *Every (AI & I)*, Trevin argued that developers are coding with AI agents incorrectly by relying on a single model, presenting Compound Engineering's latest release as a solution that orchestrates multiple models as a panel. The release adds deterministic bridges between harnesses, allowing a user in Claude Code to invoke CE plan use Fable for planning by a different model and delegate implementation to another via CE work use Codex. Trevin predicted the most popular multi-harness patterns will be "using the point-of-view skill for multi-model debate, then planning with Fable, and finally implementing with Codex," adding "I think this will quickly become the most commonly used skill for a large percentage of people." The release's POV Oracle skill constructs a neutral prompt sent to multiple models simultaneously, with the orchestrating agent running up to three rounds of debate. Trevin shared that "more than half the time, Claude will say, 'I completely missed something, and this other model pointed this out, and now I've changed my mind.'" Kieran Flanagan, the host and co-creator, stated: "The point is that we don't want to say Claude Code is the best so you can only use Claude Code because it is not the best."