← THE INDEX · VULN RESEARCH
Darwin
Autonomous source-to-exploit research. One rule: nothing gets reported unless a deterministic test could reproduce it.
What it is
Darwin unifies several retrieval systems that already worked in isolation and gives them one interface, a capture inbox, and a freshness discipline.
The failure it corrects
Every previous attempt at this built a retrieval engine and called it a brain. A retrieval engine is a hippocampus. A second brain is a hippocampus plus a front door plus a capture habit plus a way of aging out what is no longer true.
The named failure mode is that a static brain rots, and stale facts look exactly like true ones. This is not hypothetical. During the build itself, an agent confidently reported the live status of two projects that had been dead or parked for weeks, because it read its own memory and nothing in that memory had decayed.
So the thesis is knowledge under selection pressure. Facts carry freshness. Facts that stop being reinforced lose standing rather than sitting at full confidence forever. The retrieval quality was never the bottleneck. Confidence in stale retrieval was.
Shape
- Cortex. The unifying layer with adapters onto each existing retrieval engine, so no engine has to be rewritten or migrated.
- Ingest. Capture collectors, including clipboard daemons on both Linux and Windows, because a capture habit that requires opening an app does not survive contact with a working day.
- Deck. Event collection and an agent command surface.
- Face. The web interface. The storefront is load-bearing here. A second brain nobody opens is not a second brain.
Why it is an integration project
Roughly four fifths of the capability already existed across separate tools. Treating this as a new research project would have produced a fifth engine and the same problem. The work is connective tissue and a front door.