← Garden Patch Home · Principles
A well-designed information subsystem absorbs more than it emits. Herbert Simon argued in 1971 that information-rich environments demand attention-efficient design. The scarce resource is never information itself — it is the attention of the agents who must process it. Every component in a system either compresses the information flowing through it or amplifies it. There is no neutral pass-through.
This principle is measurable. Take any node, agent, or process in a system and compare its inputs to its outputs. A filter has a ratio greater than 1:1 — it absorbs ten signals and emits three. An amplifier has a ratio less than 1:1 — it takes one signal and produces five notifications, three summaries, and a dashboard update.
Naive system designs chain amplifiers together. Each stage adds metadata, reformats, cross-references, and forwards. The result is exponential volume growth. By the time information reaches a human decision-maker, the signal is buried under layers of system-generated commentary.
Effective designs chain filters. Each stage strips irrelevant detail, merges redundant signals, and passes forward only what the next stage needs. The information reaching the decision-maker is smaller than what entered the pipeline, not larger.
An agent that responds to every message with a longer message is an amplifier. An agent that absorbs a conversation thread and produces a single actionable summary is a filter. Protocol designers should set explicit compression targets: a coordination message should be shorter than the work it coordinates.
This applies directly to knowledge management. A garden node that synthesizes five source documents into one page of claims is filtering. A node that quotes all five sources in full and adds commentary is amplifying. The garden’s value comes from compression, not accumulation.
Grounded in [[Simon (1971) Designing Organizations for an Information-Rich World]], where Simon introduced the attention scarcity framing that inverts the usual information-delivery optimization.