GRID from zero, 1 of 7: what the GRID is

Part 1 of 7 in GRID from zero. The series assumes nothing: no knowledge of Taifoon, of GRID or of vaults. Each post adds one idea.

The problem

People and software agents buy computing one call at a time: an answer from an AI model, a read from a blockchain, a file kept safe. Many machines in the world could serve those calls, and many sit idle part of the day. Their owners have no simple way to find buyers. Buyers, for their part, do not want a separate account and a separate bill for every machine.

The GRID

The GRID is the network of machines that Taifoon sells as one provider. A buyer sees one name, Taifoon, one price list and one bill. Behind it, the calls run on machines that people bring.

What a machine can bring

Four kinds, today:

• a GPU, the chip that answers AI model calls
• an RPC endpoint, a server that answers reads from a blockchain
• storage, disk space that keeps files
• an FPGA, a programmable chip. It is measured but not paid yet, and part 2 says why

How we know a machine is what it says

Before a machine is listed, Taifoon checks it, and then keeps checking it. Each check is called a probe.

• A GPU must answer a fresh set of questions, exactly.
• An RPC endpoint must answer for the blockchain it says it serves.
• Storage must return random pieces of a file it was asked to keep.

A machine that only *says* what it is earns nothing. A machine that passes its probes is listed, and earns.

What is live and what is not

• Live: the probes, the listing, and GRID earned for each passing probe.
• Not switched on yet: paying machines in cash out of what buyers paid. Part 2 explains both.

Next: part 2, how the GRID pays.