Playable theory · The Instinctual Learning Agent Harness
The Instinct Engine
A language model sits at the centre of this machine, frozen: it will not learn a thing. Everything around it can. You play that part, the harness, over four days at a clock-repair shop’s help desk, and teach it to turn slow thinking into fast instinct.
The harness · around a frozen oracle
What the harness has learned
A concept game for The Instinctual Learning Agent Harness (Micheal Lanham, Packt, in development). The shop, its customers and its policies are invented. Times and costs are illustrative game units chosen to show the shape of the trade, not measurements of any model. Jev is TypeSafe’s decision model; here it is simulated. Nothing on this page calls a model or the network.