Ash · extension
Train machine-learning models like you build scenes.
Ash is the ML training workstation of the BigBrain line — the flagship AI tool. Design training pipelines visually, run them GPU-first on your own hardware, watch losses and samples live, and keep every experiment in a ledger you can trust. It runs as an extension on the godspeed engine.
Buy godspeed + ash
Two license keys arrive by email — one for the engine, one for ash. Stacking more modules? Every combination is cheaper together.
What it does
The diagram is the program
Build training graphs on a blueprint-style canvas; ash compiles the graph itself through our native toolchain and runs it. There is no separate config file and no script that can drift from the picture — no glue code, no Python environment to babysit.
GPU-first training
Training runs exclusively on the GPU — NVIDIA and Apple, picked automatically — through the same GPU stack that renders the engine. Runs are detached, so closing the editor never kills a job, and reopening re-attaches to it.
Wrong wires refuse to connect
Every port is typed, so a mismatch is caught when you draw it — named on the edge, both types shown — and again by the compiler. You find out before the run, not at epoch three.
A run screen worth watching
Losses, samples, and diagnostics live as they happen — training is a first-class screen in the workstation, not a scrolling log.
Trainer and runtime, cleanly split
The trainer is the product you buy; the runtime primitives are an installable substrate other software can detect and use — our other products do exactly that.
Pairs with boltzmann
Add boltzmann and model development gets version control: experiments, model versions, and tickets in one flow.
ASH- key. It is prealpha and moving fast: the graph editor,
the run screen, the node library and the experiment ledger work now; a headless
ash run command line, image previews in the run view, and collapsing a
subgraph into a reusable node are the roadmap. Subscribing now locks the founders price
forever. Cancel or refund freely within the 14-day window.
Deeper dive: the ash documentation.