The human conditions
of an AI age.

Machines can perform more of the work of intelligence. The question that concerns equilibri is what people should receive in return. More time, greater agency and a repaired ecology are possibilities to pursue. They are not properties of the technology.

The same systems that may help us understand a changing planet depend on energy, water, minerals, land and human labor. Their capabilities and their costs belong in the same account. [01] [03] [04]

Our position is conditional. AI may become one of the most powerful instruments available for mapping the scale of its own ecological consequences. Analytical capacity does not cancel physical cost. The livable balance has not yet been achieved.

This studio investigates the conditions that balance might require, across attention, leisure, relationships, income and the places where life happens. The purpose is orientation among possible futures, with space for adoption, criticism and refusal.

Stanford’s 2026 AI Index describes strong advances alongside persistent failures on other tasks. A benchmark result is evidence about a test, not permission to hand over every decision. The OECD’s September 2026 study of agentic organizations offers early practitioner accounts rather than proof of a settled operating model. [01] [02]

An agent can pursue a goal through several steps, sometimes acting across software. This changes the stakes of delegation. A mistaken draft can be rejected; an action already taken may need to be reversed.

If some cognitive capabilities become cheaper and more available, judgment, trust, care and attention may become more consequential constraints. This is an equilibri hypothesis about what institutions should value, not a claim that human intelligence has become obsolete.

Scientific discovery and creative assistance can coexist with concentrated power and unreliable outputs. We need to ask which capability, for which task, under whose authority, and against what alternative.

In a study of 5,172 customer-support agents, AI assistance increased issues resolved per hour by 15% on average, with different effects across workers. That is a task-productivity finding, not a finding about shorter days. [06]

A separate six-month study of 2,896 employees across 141 organizations found improved well-being after work reorganization and a four-day week at maintained pay, compared with 12 control companies. It was not an AI intervention, and participating organizations were not a random sample of all employers. [07]

A faster task can become another assignment, a staffing reduction, higher income, a shorter day or time for learning and care. Contracts, ownership and bargaining power shape that choice. Time remains occupied if someone must keep watching for an alert.

A CHI study of 319 knowledge workers found that greater confidence in generative AI was associated with less self-reported critical-thinking effort. Participants described more emphasis on verification and integrating responses. The survey does not establish that AI causes lasting cognitive decline. [09]

Human–AI interaction research gives practical guidance on expectations, correction and control. Those controls matter most when a system is wrong, not just when it is convenient. [10]

Cognitive offloading can make room for thought; it can also remove occasions to practise. Our design question is where assistance helps someone learn, and where it conceals the reasoning they need to challenge an answer. Judgment, questioning, synthesis, taste and ethical interpretation deserve deliberate practice.

These are studio design commitments, not a claim that an interface alone can secure them.

Humlum and Vestergaard’s Danish study finds changes within jobs without detectable average effects on earnings or recorded hours during its early observation window. The ILO estimates occupational exposure to generative AI, not actual job losses. Exposure, adoption, productivity and income are different measures. [15] [16]

The OECD examines collective bargaining as a way to shape transitions, while its market analysis identifies concentration risks along AI supply chains. Oxford’s Fairwork framework brings pay and voice in platform labor into view. A distribution question must include workers behind the interface. [17] [18] [22]

AI could support higher incomes when it expands valued work and workers can claim part of that value. Gains may also go to owners, model providers, customers through lower prices, or the public through institutions. Technical performance cannot choose the settlement.

For a freelancer, faster delivery can increase effective hourly earnings under a fixed fee and reduce the invoice under hourly billing. Demand, quality and bargaining power remain decisive. Paid hosts, cleaners, evaluators and care workers must be inside any account of shared prosperity.

These are research designs for partners to investigate. No trial, venue or demonstrated outcome is claimed.

Across the pilots, predefine outcomes and comparison methods. Study schedule control, income, concentration, movement, social connection and access to nature alongside output. Protect consent and report harm, no improvement and uneven effects. Keep these outcomes separate rather than compressing a life into a score.