We build Dandela: continuous-learning AI infrastructure for collaborative projects — governed recursive self-improvement (RSI) with human attribution.
Serious research and innovation need more than isolated AI outputs. Projects must preserve context, adapt to new evidence and deliver reliable results — without losing human governance or attribution for the people creating value.
Founded by MIT CSAIL scientists, we combine frontier models, neurosymbolic methods and bio-inspired system design. Dandela improves both the project and how the work gets done, with human governance and verification guiding improvements in performance, cost, reliability and discovery.
Dandela operates as a system of systems. Persistent project intelligence keeps work connected over time. Meta-RSI, our central coordination layer, supports shared strategy and continuous learning across projects.
One technology, two ways to build.
Ledger-based attribution keeps accepted contributions connected to the people behind them. Project agreements define revenue sharing and co-ownership where granted. For selected projects, we provide compute and support, with eligible contributors receiving the majority of defined project receipts.
Verified capabilities become reusable building blocks for larger compositions. Where permissions and technical conditions allow, useful work can support further discovery across projects, domains and model generations.
Pronounced “dice” — exploration and programmable adaptation.
We’re hiring — get in touch!