Away-from-keyboard (AFK)

Away-from-keyboard (AFK) is a shorthand used in chat, multiplayer games, and other live text environments to signal that a person has left their device and is not currently able to read or respond to messages. The initialism originated in late-1980s and 1990s online culture – particularly on time-share systems, IRC, and MUDs – where other users needed to know whether a conversational partner was still present at their terminal. It has since become a routine courtesy in instant messaging, gaming lobbies, and team chat tools.

In modern distributed teams, AFK behaviour is folded into broader ways of working. Status indicators in Slack or Teams, calendar blocks, and "do not disturb" modes all encode the same idea – the person is not reachable right now. This intersects with maker’s schedule thinking, where prolonged AFK periods from chat are a deliberate strategy to protect uninterrupted work time. For on-call engineers, AFK is a more constrained state. They may be away from a keyboard but remain reachable by alert and pager, expected to return to a keyboard quickly when an incident occurs.

The AFK initialism has been repurposed for AI agents to describe a mode of agent execution. An AFK agent runs detached from any synchronous chat interface with a human user – in the background, in the cloud, or on a schedule – rather than turn-by-turn inside an interactive session. The human sets a goal and steps away. The agent works on its own and surfaces results later for review. The design discipline for running agents this way is loop engineering. The explicit objective of loop engineering is to remove the human from the loop and into a supervisory role that sets up discovery, verification, and scheduling once, then lets the agentic workflow run its course.

AFK is the throughput multiplier of AI coding. Many AFK sessions can run in parallel while you sleep, eat, or work on something else — work that would block a live session proceeds on its own. This usually requires a permissive permission mode combined with sandboxing to be safe.

When you’re not there, the agent handles ambiguity differently. While you’re watching, an ambiguous decision surfaces as a question. Once you’ve walked away, the agent picks a default and keeps going, and every later decision builds on that guess. The characteristic failure is coming back to hours of finished, confident work built on a wrong call made in the first ten minutes — coherent, just coherent about the wrong thing.

Since you can’t give input during the run, give it before and after. Before: resolve ambiguity up front — a grilling session, a written spec — so there are fewer gaps to fill alone. During: automated checks and automated review stand in for the attention you’re not giving, failing fast on what can be caught mechanically. After: the run ends in something reviewable — a PR, not changes already merged. AFK doesn’t remove human review, it defers all of it to the end, so what arrives has to be worth reviewing. AX matters most in AFK runs — with no one watching, the environment is the only support the agent gets.

Avoid "background agent". That phrase centers the machine rather than the human pattern. AFK names the fact that matters: the user isn’t watching.

See also

• Downtime: The analogous concept for systems rather than people.