When a system is drawn in energy circuit language, the subsidies the interface conceals become lines on the diagram. The flows the user does not see become arrows. The storages the economy does not price become tanks. The full cost of the system becomes legible in a way that prose descriptions struggle to achieve.
The builder-AI circuit has two primary input channels. The first is the builder's cognitive energy — attention, judgment, creativity, domain expertise — the high-transformity products of a lifetime of education and experience, metabolized from glucose sustained by agricultural, medical, and social infrastructure. The transformity per unit is extraordinarily high. The second is computational energy — electricity processed through data center infrastructure, drawn from grids, generated from fossil fuels or renewables. The transformity per unit is lower than cognitive energy, but the quantity is vastly larger. The asymmetry in quantity compensates for the asymmetry in quality.
The interaction node — where the two inputs converge — is the conversation itself. The builder's cognitive energy provides direction, evaluation, and high-transformity judgment. The computational energy provides execution, pattern-matching, and the vast associative reach of a trained model. The interaction produces artifacts: code, text, design, analysis — embodying contributions from both in proportions that vary with each exchange.
The failure modes revealed by the circuit are more instructive than the successes. Cognitive withdrawal: the builder reduces her contribution, the circuit continues producing artifacts, but the transformity declines because high-quality input has been withdrawn. Feedback loop degradation: the builder stops learning from the collaboration, the circuit produces output without transforming the builder, her expertise stagnates. Thermal runaway: the circuit operates beyond the builder's capacity for cognitive renewal, waste heat accumulates faster than it can be dissipated, burnout follows. These are not moral failures. They are circuit pathologies predicted by the diagram.
Odum began developing the energy circuit language in the 1950s alongside his doctoral work. Ecological and General Systems (1994) contains the most complete systematic presentation. The language's universality — the same symbols applying across biological, technological, and social systems — was deliberate, reflecting Odum's conviction that all organized systems share fundamental thermodynamic architecture.
The application to human-AI collaboration is developed here. Earlier applications covered ecosystems, industrial economies, and national defense; the extension to the builder-machine circuit follows the language's universalist ambition.
Symbols make structure visible. Circles, tanks, blocks, arrows, and heat sinks compose into diagrams that expose what prose obscures.
Universal across substrates. The same notation describes wetlands and workflows, economies and organisms.
Feedback loops as structural features. Healthy circuits include feedback from output back to input, renewing the sources that sustain them.
Heat sinks are inevitable. Every transformation dissipates energy as unavoidable thermodynamic tax; the diagram makes this explicit.
Failure modes are predictable. Cognitive withdrawal, feedback degradation, thermal runaway — each corresponds to a specific structural pathology in the circuit.
Whether the language achieves the universal applicability Odum claimed remains debated in systems ecology. Critics argue the notation is better suited to some systems than others; defenders point to its successful application across domains as evidence of adequacy. The application to AI-human collaboration is new enough that its empirical track record remains to be established.