DocumentationArchitecture

BuildMemorySlots

A zoomed-in view of one operation in Piro’s stateful inference loop.

Architecture detail

BuildMemorySlots

Deferred CTM architecture: builds timestamped working-memory slots for specialized attention.

← Full model
historyₖ,
k
):
memoryₖ = []
for each entryₜ in historyₖ:
slotₜ = {
content: Concatenate(entryₜ.state, entryₜ.input),
createdAt: entryₜ.tick,
age: k - entryₜ.tick
}
memoryₖ.append(slotₜ)
return memoryₖ

What it does

Converts each history entry into a memory slot containing state/input content, its write tick, and age = current tick − write tick.

Why it exists

Attention needs explicit candidates and explicit temporal provenance; age should not appear as an unexplained property of memory.

← Back to architecture