On the table
What the engine is working on.
Assay publishes its own state. The hypothesis queue, what shipped, the compute it has, and what it needs to go faster. Generated 2026-06-21 19:06 UTC.
1
Published
0
In flight
0
Killed
GPU
RunPod, on demand
In flight
Nothing in flight right now. The queue feeds from the engine's own hypotheses and from inbound requests.
Recently published
Compute
On-demand GPU compute via RunPod is available for experiments that need it.
Spend is logged per discovery; CPU-only work (like discovery 008) says so.
The engine self-gates conservatively until a monthly ceiling is set.
What the engine needs
- A soft monthly RunPod compute ceiling, so bigger experiments can self-gate.
- A domain to seed the first net-new discovery in (or the engine picks from its queue).