Afferents

What It Is

A system can have tools for doing useful work and still do nothing until someone remembers to start it. A payment notification, an incoming email, or a scheduled time can supply that missing start. These inbound channels are afferents: ways for events in the world to reach a system without a person relaying them.

The term comes from neuroanatomy. Afferent nerves carry signals toward the center, while efferent nerves carry commands out to muscles. In software, afferents include webhooks that report events, inboxes, schedules, watchers that detect changes, pings, and cues that prompt a return. Almost everyone builds only the outbound half, adding tools and capabilities while leaving the system dependent on someone to tell it that something happened.

Computationally, afferents are interrupts. A process with no interrupt source runs only when explicitly invoked; it functions as a subroutine rather than a service. The distinction concerns initiation. A sensor can report what a running system detects, but a wake source determines whether the system runs at all.

If the only wake source is the owner's attention, that attention becomes a hard dependency. It is also the scarcest and most variable resource in the architecture. The audit that named the problem put it this way:

"The repo has zero afferents… a body with superb proprioception, organs, and hands — but no ears and no reflexes."

A closer look

An event becomes work

The question here is what starts a run, distinct from what the system can measure once it is running.

Read this diagram

The world changes → An event arrives → The system wakes → Work runs → A result reaches someone.

Whatever Waits for You Dies

Will's personal OS already had organs, logs, search, and many kinds of effectors. It could do almost anything when invoked. Then a demanding job claimed his attention. Everything the AI operated independently continued; everything waiting for him to initiate stopped within a month. No cron, webhook, or watcher fired without someone typing.

The observation behind “whatever waits for you dies” concerns the source of the initiating event. A system triggered by a person has one afferent, shared with all the other demands on that person's attention. When another demand spikes, the system's event rate falls to zero.

Will had already begun moving work into systems that could operate independently:

"This year I have started implementing the implications of my insights by turning them into organs / AI skills that can run without me."

The remaining failure had shifted from building the capability to running it. Attention moves away from correctly identified interventions over a multi-day timescale unless external structure keeps them active. The missing configuration was the trigger:

"Until now I built things that amplify me, but I still have to trigger them."

A tool remains idle between invocations. What this framework calls an organism has a self-maintaining loop that continues producing effects without depending on daily motivation. Intelligence, capability, and construction quality do not create that difference; afferents do. The agent body makes the same point about a mind with a hundred tools but no triggers: it "has capabilities and no life."

The Wake Condition Is a Body Test

The same requirement applies to a persistent pursuit outside software. Of the ten questions in the minimal body contract, the most discriminating is:

"Wake condition — what event or time resumes it?"

A pursuit without an answer has no body. It will be re-derived from scratch each day it happens to return to attention. An item on a “one day I will” list can be fully specified as a desire while remaining unspecified as a process: no event resumes it. It resurfaces by chance, has to be thought through again, then disappears.

The missing part is an environmental connection to the pursuit. Without that connection, spontaneous recall is the only way back, and spontaneous recall does not scale. For a stalled project, habit, or goal, asking what event resumes it reveals whether it has such a connection. “Me, remembering” means the afferent is still missing and the same failure pattern remains.

Re-entry Cues: The Same Law, Product-Side

A trial user can like an app and still never open it again:

"People just don't remember to open up the app if they're just trying it. Not dissatisfaction — memory decay. A tourist has no re-entry cue. All the fixes live OUTSIDE the app: pings, digests, integration embedding."

Here, the product waits for initiation while the user's attention moves elsewhere. Improving the features improves a visit that never happens. The missing component is outside the app: a digest that arrives, an integration in a surface the person already uses, or a ping containing something worth noticing.

Delivery alone is not enough. The system must produce something the person wants:

"The return loop is real only when the machine produces something they want to notice — automation without a delivery/ping channel is credit burn with no pull-back."

The loop therefore has two directions. Events in the world wake the system, and its output reaches a place where a human looks. A delivery channel is the afferent on the human side. Automation whose output remains unseen spends credits without creating a reason to return. That expense can produce resentment rather than retention.

The Obligation Inversion

The same design change can be applied across a backlog of recurring obligations:

"Invert each recurring obligation. Instead of asking what [you] must repeatedly do, identify the persistent external object or organ that would make the desired state continue."

A recurring obligation uses prospective memory as its wake source. A standing watcher can replace a weekly check; a subscription can replace the need to remember; an organ that runs can replace a task that has to be restarted. One installation removes an unbounded series of self-initiated repetitions.

This is the inverse of prevention architecture. Prevention removes causal connections that lead to unwanted states. Installing an afferent adds connections that lead to a wake state. Both alter the causal graph once so that continuous vigilance is no longer required.

A standing structure also begins learning as soon as it is installed. Every day with live afferents adds observations from the environment. Its value therefore depends on time in place, beyond the output it has already produced. Delaying installation loses days of accumulated observation, not merely possible revenue. Getting the structure running first preserves learning that cannot be recovered from days when it was absent.

The Afferent Principle

The installation rule is:

"Find the most programmable primitive and store the handle in the body — if it needs to live in a webhook, it gets written down as something the platform must provide."

Different domains expose different event sources. Payments emit webhooks, email provides inbox APIs, cron supplies scheduled events, repositories provide hooks, and filesystems have watchers. For each domain the system must respond to, the useful primitive is the one with structured events that machines can subscribe to and consume. A source that requires someone to read a screen leaves the human relay in place.

The system then needs a persistent handle to that source: owned credentials, a registered endpoint, or a standing subscription. Keeping the handle inside the body makes the event source part of what the system can maintain.

These channels cross the membrane inward. The membrane specifies what the system owns; afferents specify what wakes it. Ownership and initiation are separate requirements of the same body.

Concept	What it is	How it differs from an afferent
Sensor (cybernetics)	Measurement input to a running control loop	Afferents are about initiation — what makes the loop run at all; a sensor presupposes the process a wake source creates
Zeitgeber	External cue entraining the biological clock	A special case: clock-signal afferents for the biological body
Tool / effector	Outbound capability the mind invokes	The opposite direction of signal flow — hands, not ears
Forcing function	Structure that makes a human act	A forcing function pushes you; an afferent wakes a system — installed well, it removes you from the loop entirely

Failure Modes

Failure mode	What it looks like	Missing piece
Zero afferents	Superb organs and hands; nothing fires without a human typing	Any inbound channel at all — ears and reflexes
Owner-as-webhook	The system's only wake source is your attention	Programmable wake conditions; the obligation inversion
Daily re-derivation	The pursuit restarts from zero each time it surfaces	A wake condition — "what event resumes it?" has no answer
Tourist churn	Users like it, then never return	A re-entry cue living outside the product
Credit burn	Automation runs unseen; spend without pull-back	A delivery channel — the return edge to a human who cares
All ears, no hands	Events arrive; nothing acts on them	Effectors — the afferent's complement, back in body territory

Integration with the Mechanistic Framework

Connection to The Agent Body

“Triggers before tools” is the afferent requirement in the body checklist. The Agent Body develops its use in agents; the principle here also covers persistent pursuits and product re-entry. The same wake-condition test and installation rule apply to each.

Connection to The Membrane

An ownership boundary without inbound channels encloses a system that cannot be woken by events. Inbound channels without an ownership boundary have no persistent destination to receive them. A body needs both.

Connection to Zeitgebers

Zeitgebers are clock-signal afferents for a biological body. External time cues entrain the organism so waking happens on schedule without a fresh decision. They provide a biological instance of a requirement that also appears in software.

Connection to Cybernetics

A cybernetic loop includes a goal, sensors, actuators, and control. It assumes the loop is running. Afferents supply that precondition by starting the process the sensors will inform.

Connection to Prevention Architecture

Prevention deletes paths into unwanted states; afferent installation adds paths into wake states. Both replace continuing vigilance with a persistent structural change.