Where we’d love to take OpenOrbit next

[forthfate]

OpenOrbit already helps us evaluate AI and product behavior, keep the evidence, and review improvement ideas. But we’ve been thinking a lot about what would make it genuinely useful in day-to-day work — not just for spotting a problem, but for helping teams make changes with confidence.

Here are a few big directions we’re excited about.

Making improvement ideas easier to act on

A good suggestion is only the start. We want to help teams try an improvement somewhere safe, compare it with what came before, and then decide whether to keep it, revert it, or turn it into a PR — with the evidence right there.

Making recurring evaluations easier to trust

Evaluations are most helpful when they keep running in the background, not only when someone remembers to start them. Better retries, scheduling, safe stopping and cleanup, and stronger support for remote agents would make that feel much more dependable.

Making it easier to understand what happened

A pass or fail label rarely tells the whole story. We want richer logs and metrics, safer handling of retained data, and clearer approval records so it is easier to understand what changed, why it mattered, and how a decision was made.

Trying more ideas without creating more risk

Running one experiment at a time can be slow, but running many changes in the same place can get messy. We’d like to explore isolated Docker environments so teams can evaluate several ideas in parallel without touching their working repository or getting in each other’s way.

None of this is a fixed promise or a finished plan. It’s the direction we’re excited to explore, and we’d love to shape it with the community.

What would make OpenOrbit more useful for your own AI or product-improvement work? We’d love to hear your thoughts.