What remains scarce after AGI? – Alex Imas and Phil Trammell

The conversation opens with the question of scarcity in an AI-abundant world. Imas introduces the 'relational sector'—goods and services where a human being in the loop is intrinsically part of the value, such as a human barista or ballerina. However, both guests caution that because the machine economy doesn't consume human-intrinsic goods, the relational sector could become a shrinking share of total economic output, especially if automation continuously expands the variety of non-human goods.

A major theme is the historical difficulty of economic forecasting. Imas invokes David Ricardo, who correctly predicted that Industrial Revolution jobs would be automated but failed to anticipate the creation of entirely new job categories. Trammell uses a Mongolian economist thought experiment to illustrate how holding varieties fixed leads to wrong predictions—just as a 1400s Mongolian might have predicted all wealth flowing to singers once horses were automated, but instead new non-singer goods proliferated.

The stability of labor share (~60% for centuries, a 'Kaldor fact') is discussed at length. Phil argues that looking at network-adjusted factor shares down entire supply chains shows labor still adding substantial value even in highly automated sectors. Both agree a qualitative shift is coming where some goods will have fully automated supply chains, but the implications for overall labor share are ambiguous depending on whether demand for automated goods satiates quickly or keeps expanding.

The 'messy middle' scenario—where automation displaces workers faster than wealth redistributes—is examined. Imas argues it requires a narrow set of unlikely conditions: automation must be just barely cheaper than human labor, and the technological frontier must not expand substantially. Political economy concerns are raised: a 2-3% unemployment spike could trigger political crisis, while a slow drip of displacement (like phone operators over 20 years) might be politically invisible but economically damaging.

On redistribution policy, options discussed include negative income tax, UBI, universal basic capital, consumption taxes, and sovereign wealth funds. Concerns include UBI's political fragility (dependence on whoever holds power), wealth tax escalation precedents, and the difficulty of indexing to the right assets. The analogy of electricity vs. social media is used: electricity commoditized broadly, spreading gains widely, while social media concentrated rents at platforms. Both guests express hope AI resembles electricity.

For developing countries, the guests note a lack of economic research on this topic. Two scenarios emerge: AI diffuses broadly via open models, enabling leapfrogging (as mobile banking did in Nigeria); or concentrated, private AI leaves developing nations without access to gains. The recommendation leans toward indexing global AI capital rather than retraining programs, given the speed of potential AI advancement.

Finally, the conversation turns to whether future AI agents and selection dynamics among wealth-accumulators will reshape the economy. The argument is made that agents who don't satiate in capital—whether humans like Musk or future AI entities—will accumulate disproportionate wealth over time, potentially driving capital share toward one. Whether human preferences for relational goods are evolutionarily stable is debated, with Imas arguing that selection pressure favors humans who prefer human interaction.