https://raven-ml.dev/blog/Notes from building Raven. https://raven-ml.dev/blog/2026-07-raven-on-gpus/https://raven-ml.dev/blog/2026-07-raven-on-gpus/01 Jul 2026 00:00:00 +0000Raven runs on GPUs: the tinygrad port is complete, a GPT-2 training step runs at 110 ms on an H100 (tinygrad: 90 ms), Rune and Kaun were rewritten on typed parameter trees, and three new libraries landed (Munin, Vega, Norn). The beta 1 scope is complete. https://raven-ml.dev/blog/2025-11-alpha-complete/https://raven-ml.dev/blog/2025-11-alpha-complete/01 Nov 2025 00:00:00 +0000Alpha 1 shipped with three new libraries (Fehu, Saga, Talon), FFT and a full linear algebra suite in Nx, workshops at FunOCaml, two Outreachy internships kicked off — and the decision that set up everything since: porting tinygrad to OCaml.