Connect compute
GPU owners expose available model-serving capacity to the network.
honey
One API for distributed AI compute.
Connected GPUs become the hive. Model pools become shared capacity.
Thesis
Honey turns scattered GPUs into one API.
Pool models
Workers capable of serving the same model form shared capacity.
Route requests
Developers call one interface instead of managing individual nodes.
Show the work
Latency, availability, failures, and usage should be inspectable.
Network
Distributed inference should feel normal to use.
Machines can come online, go offline, change load, or switch models. Honey hides the mess from developers while keeping routing visible.
Use cases
Where pooled inference starts to make sense.
AI agents
Give agents shared inference instead of tying them to one local box.
Local model teams
Combine workstations, lab machines, and rented GPUs behind one endpoint.
App builders
Route model calls for bots, web apps, automations, and internal tools.