I built superResearch because I kept running into the same problem with coding agents: they could generate changes, but they were hard to trust, hard to measure, and hard to use as part of a real improvement loop. I wanted something different: a native desktop app that can work on a local codebase, run evaluation commands, extract a metric, preserve diffs and traces, and keep or discard changes based on actual results. The idea evolved a lot while building it. It started as “can I make agents edit code locally?” and became “can I turn a software project into a measurable optimization loop?” That shift changed everything. The product became less about one-off code generation and more about planning, applying, evaluating, reviewing, and iterating in a way that a human can actually inspect and control. superResearch is local-first, Git-backed, and designed to be reviewable instead of magical. You point it at a workspace, define what success looks like, and let it search for improvements while keeping the process visible. It’s still early and experimental, but the direction is clear: make autonomous software improvement more measurable, more inspectable, and more useful in real workflows. I’d love feedback from people working on developer tools, AI coding, research automation, or optimization systems. What part feels most useful, and what feels missing?

superResearch

Autonomously improving systems.

Autonomously improving systems.