The RFTSystems/ai‑systems collection brings together a set of interactive demos that showcase how new optimizers and symbolic kernels can be benchmarked in practice. From the DCLR Optimiser, which outperforms Adam and Lion on CIFAR‑10, to adaptive computing modules and kernel stability APIs, each Space is designed to be simple to try out while also highlighting deeper research goals around stability, coherence, and agent lineage. Whether you’re curious about optimizer performance, collapse torque overlays, or emergent system behaviour, this collection offers a hands‑on way to explore and contribute to the growing Codex of re
producible AI systems.https://hf.co/collections/RFTSystems/ai-systems