Neural systems
Understand event-based networks and associative patterns before using them.
Minds exposes neural capabilities for specialized workloads. They are tools for particular representations and computations, not prerequisites for storing a memory or making a graph query.
Spiking neural networks
A spiking network represents activity through discrete events over time. This makes timing and network configuration part of the input contract. Read the network specification, operation parameters, and expected output before adapting a conventional tensor or text workflow to it.
The instance API includes network operations, training routes, and federation-related paths. Some implementations are incomplete: the audited BPTT training handler creates a trainer and reports placeholder completion without executing a training loop. A completed response from that path is not evidence of a trained model.
Associative patterns
Modern Hopfield Network–style pattern storage is used for associative recall. A partial or noisy cue can be used to retrieve a related stored pattern. This differs from ordinary key-value lookup and requires a compatible numeric representation.
Start small
Use a small test network or pattern set with a known expected result. Verify the output directly, then repeat the operation after reconnecting to test the persistence path you intend to rely on.