A SOCIAL RESEARCH STUDIO

Intelligence at work.
Time to live.

equilibri

A thesis on how agentic work redistributes authority, income and time — and the institutions a viable hybrid life requires.

Listen to the audio introduction
TRUDY HALL / CURATED RESEARCHRead the argument
THE POSITION

The gains from agentic work need a social settlement.

When software can pursue a goal across several steps, delegation becomes a question of authority: who defines success, permits action, verifies the result and absorbs the cost of failure? equilibri argues that workers should help set these terms before faster execution becomes a higher performance baseline. The dividend should include stronger income and time they can direct — toward leisure, relationships, care and meaningful work. The compensation question

Hybrid life requires an ethical settlement: who may delegate, who can refuse, and who bears the consequences? UNESCO places dignity, privacy, fairness and ecological well-being at the centre of AI ethics. equilibri translates those principles into worker voice, shared gains and protected time away. An arrangement succeeds when people retain meaningful choices and affected communities can challenge its costs. UNESCO · Recommendation on AI Ethics

THE ARGUMENT

From task efficiency
to temporal autonomy.

Seven propositions examine where control moves when execution is delegated: into objectives, permissions, review queues, pay arrangements and the infrastructures of daily life. Each asks what workers and institutions can change before new defaults become entrenched.

01 / THE TIME DIVIDEND

Efficiency is a distributional question.

A faster task creates a claim on capacity. An employer may allocate it to more output; a worker may seek higher income, learning or leisure. Before measuring a gain, decide what counts as completion and whose time enters the calculation. Prompting, checking, correction and downstream rework belong in the account. Concurrent agents can shorten elapsed time while leaving human attention fully occupied.

A 2026 preprint surveying 5,512 Korean workers reports time savings and increased on-the-job leisure among recent generative-AI users. Berkeley’s eight-month ethnography at one technology company describes an opposing pattern: broader responsibilities and work expanding into pauses. These studies use different methods and cannot establish a universal agentic-work effect. Together they make allocation, rather than speed alone, the central research question. Suh & Oh · preprint, 2026 Ye & Ranganathan · Berkeley Haas, 2026

The proposed response is a negotiated review cycle: establish a quality-adjusted baseline, record the full cost of delivery, then agree how verified gains become pay, reduced hours or protected intervals. Revisit the baseline when the tool or task changes. Treat a productive week as evidence for a discussion, rather than an automatic increase in next week’s quota.

equilibri asks: who can claim the released capacity, and what prevents a temporary efficiency gain from becoming a permanent expansion of workload?

Evidence & limits / time

Observed time savings / what they establish

A randomized field experiment across 66 firms and 7,137 knowledge workers provided access to AI within familiar workplace applications. In the experiment’s second half, treated workers who used the tool spent about two fewer hours per week on email and worked less outside regular hours. The researchers did not detect changes in task quantity or composition. This supports task-specific time savings; it does not demonstrate a general transition to leisure or test a coworking model. Dillon et al. · American Economic Review: Insights, forthcoming

Intraday leisure / an emerging signal

In Suh and Oh’s survey, recent users reported a 3.8% reduction in active working time on average; the estimated workforce-wide reduction, including non-users, was 1.4%. The reported share of on-the-job leisure rose by 1.3 percentage points among users. These are survey estimates from Korea, not observed time logs or randomized treatment effects. The paper also finds uneven experiences of task fulfillment. Less active work cannot by itself establish a better job. Suh & Oh · methods and results, 2026

Work intensification / the countervailing mechanism

Berkeley’s in-progress ethnography describes task expansion, erosion of stopping points and overlapping AI workflows. Separately, an NBER working paper associates occupational AI exposure with longer workdays and less leisure in U.S. time-diary data. Exposure measures are not direct observations of a person using an agent; these findings should not be read as a universal causal estimate. Together they motivate testing workload and bargaining arrangements alongside tool capability. Jiang et al. · NBER working paper, 2025

Measurement / concurrent work changes the denominator

METR’s February 2026 update judged its follow-up developer experiment an unreliable estimate of current speedup because of selection effects and difficulties measuring time across concurrent agents. The implication here is methodological: compare complete workflows, distinguish human effort from elapsed runtime and report task selection. A benchmark or a self-reported speedup cannot alone establish the amount of time available for leisure. METR · methods update, 2026

Agent execution / time that may still require supervision

Anthropic’s February 2026 analysis reports a median Claude Code turn of roughly 45 seconds, while the extreme tail had exceeded 45 minutes. Long unattended intervals exist in some workflows; they are not typical of every task. Tool runtime is neither verified labor saved nor guaranteed leisure. Anthropic · observational product research, 2026

Institutional change / a different kind of intervention

A six-month four-day-week study followed 2,896 employees in 141 organizations, with 12 comparison companies. Reduced working time with maintained pay was associated with improved reported well-being. Organizations participated voluntarily; this was not a randomized trial. It examines deliberate work reorganization, not AI adoption. Fan et al. · Nature Human Behaviour, 2025

Human–AI performance / the quality constraint

A meta-analysis of 106 experiments found that human–AI combinations improved on humans alone on average but underperformed the better of the human or AI alone. Results varied by task. Any time dividend must account for review, correction and coordination, measured at an agreed quality threshold. Vaccaro et al. · Nature Human Behaviour, 2024

02 / COMPENSATION

How AI can raise human income.

AI can raise earnings by extending the scope of human expertise, reducing delivery costs and making new services viable. Turning that potential into compensation requires a mechanism: career progression, a negotiated share of verified surplus, employee ownership or a project fee that rewards a trusted result. Faster execution and higher wages are separate outcomes that job design must connect.

David Autor argues that AI could enable more workers to perform consequential tasks previously reserved for narrower groups of experts. The strategic question is which capabilities a worker gains authority to exercise. Training without access to more responsible work may change competence without changing pay. Autor · MIT / NBER, 2024

PwC’s 2026 analysis associates faster wage growth with roles where AI increases the importance of human expertise and responsibility. It does not establish an individual causal return to tool adoption. The opportunity is to build compensation around problem framing, domain judgment, verification and accountable delivery — including the labor that makes delegation reliable. PwC · Global AI Jobs Barometer, 2026

equilibri proposes negotiating money and time together: protect dependable base income, account for oversight costs, and give workers a defined share of gains. For independent workers, specify scope, acceptance criteria and revision limits so efficiency can improve margins without creating an unlimited service obligation.

Four pathways to higher income / how the opportunity works

1 / Make existing expertise more valuable

Domain knowledge becomes more useful when a worker can apply it across a broader set of tasks. AI assistance could help experienced people undertake analysis, preparation or decision support that previously required a larger specialist team. With appropriate training, oversight and professional standards, this can support progression into more responsible work. Autor’s framework locates the opportunity in extending human expertise; higher pay also requires employers and customers to value the resulting contribution. Autor · MIT / NBER, 2024

2 / Create a surplus that can fund better pay

In a customer-support study, AI assistance increased issues resolved per hour by 15% on average. A gain like this can create capacity to improve service, expand sales or reduce delivery costs. After accounting for tools, review and implementation, a sustained surplus can fund higher base pay and shared bonuses. The study measures productivity, not wage increases; the link to compensation is an economic mechanism that an organization must put into practice. Brynjolfsson, Li & Raymond · Quarterly Journal of Economics, 2025

3 / Expand the market for human work

AI can help businesses develop products and serve demand that was previously too costly to meet. In their account of their research, Babina and Fedyk describe AI investment associated with firm growth, employment and product innovation. This opens a route to more opportunities for workers as businesses expand. The evidence largely concerns earlier AI adoption, and growth within adopting firms is not a measure of economy-wide wage gains. Babina & Fedyk · authors’ research synthesis, 2025

4 / Give independent workers a larger productive reach

For an independent professional, agents could make previously unaffordable commissions viable. A scoped project fee can reward efficient delivery, while reusable methods and evaluation criteria can make expertise easier to apply across clients. The proposed strategy is to build a portable practice: maintain exportable working records, test alternative tools and price verification into the offer. This is a business-model possibility, not an established earnings effect; demand, confidentiality, quality and competitive pricing remain constraints.

What the evidence shows / signals & boundaries

A market premium for complementary skills

PwC’s 2026 Global AI Jobs Barometer reports a 62% average wage premium for AI skills and 42% faster salary growth for roles it classifies as “professionalised” by AI than for “democratised” roles. The latter comparison concerns growth rates, not a 42% salary increase. These patterns across jobs and firms suggest that AI fluency combined with human expertise can be valuable. They do not establish that taking a course or adopting a tool causes an individual raise; selection and employer differences also matter. PwC · industry analysis, June 2026

Work can change before average earnings do

Humlum and Vestergaard’s Danish study finds task reorganization and some movement into higher-paying occupations, while estimating no detectable average effects on earnings or recorded hours. Its estimates rule out effects larger than 2% over the two years following ChatGPT’s launch. This points to the importance of turning new capabilities into career progression and compensation arrangements. It does not establish what long-run agentic work will pay. Still Waters, Rapid Currents · NBER working paper, revised 2026

Design for a broad opportunity

Hui, Reshef and Zhou find short-run employment and earnings losses among highly exposed freelancers on a large online platform. That counterevidence matters: access to a powerful tool is only one part of an income strategy. Access to customers, differentiated expertise and bargaining power matter too. Most available studies concern AI assistance or earlier AI systems; a mature autonomous-agent economy remains an emerging scenario. Organization Science · peer-reviewed study, 2024

Turn the boom into shared prosperity / equilibri’s proposal

Build compensation into adoption

Agree how verified gains will support base-pay progression, broad-based bonuses and paid training before deploying a new workflow. Establish baseline quality, workload and costs; include the workers whose knowledge improves the system and those who keep the organization running. Worker representation and transparent reviews give this agreement practical force. The OECD supports social dialogue in shaping workplace AI; this particular arrangement is equilibri’s proposal. OECD · Employment Outlook, 2023

Offer a return in both money and time

A verified surplus can be allocated between higher income, investment and shorter hours. Illustratively, $1,000 for 40 hours is $25 per hour; $1,000 for 32 hours is $31.25 — a 25% higher hourly rate with the same weekly income and eight hours returned. This is arithmetic, not an AI forecast. Income-preserving four-day-week research supports well-being benefits; each arrangement still needs sustainable funding and realistic workloads. Fan et al. · Nature Human Behaviour, 2025

Share in the value of ownership

Broad-based profit-sharing and employee ownership can add another source of returns alongside wages. Earlier research examines these arrangements; it is not a test of an AI dividend. equilibri proposes dependable base pay plus transparent participation in gains, with safeguards against concentrating a worker’s income and savings in a single employer. Freeman, Blasi & Kruse · NBER, 2010

Strengthen purchasing power and productive capacity together

Higher real incomes can support household security and spending. Productivity gains can finance compensation growth while containing labor costs per unit of output; competition can also pass savings to customers. Research by Auclert and Rognlie supports a demand channel from income distribution, while finding a small short-run output effect in their model. A stronger economy needs both broadly shared purchasing power and the capacity to meet demand. Inequality and Aggregate Demand · research paper, 2020

equilibri therefore pairs better pay with affordable housing, care and energy, accessible training and support through job transitions. Higher nominal wages alone cannot resolve shortages of essentials. IMF analysis identifies social protection, portable entitlements and stronger taxation of capital income as ways to broaden AI’s gains. These provisions can help people participate in new opportunities and extend prosperity to work that is less amenable to automation. IMF · fiscal-policy analysis, 2024

The measure of success is rising median disposable income, greater security and discretionary time, alongside sustainable businesses and responsible resource use. The AI boom offers a chance to increase the economic value of human capability — and to make that value visible in everyday life.

03 / AGENCY

Authority must be designed into delegation.

Human oversight is an ethical commitment only when someone can understand, question and stop a decision. Distinguish permission to inspect, draft, recommend and act. People affected by consequential outputs need an explanation and a route to human review. Assign accountability to the organization deploying the system, rather than leaving the final reviewer to absorb failures they could not prevent.

NIST’s AI Risk Management Framework warns that human–AI configurations can amplify bias; a human signature alone is no safeguard. It calls for clear responsibilities and input from affected people. equilibri proposes revocable permissions, documented reasons for consequential decisions, and paid, protected authority to challenge them. These are governance commitments, not evidence that oversight always works. NIST · AI RMF, Appendix C, 2023

Agency also depends on retaining the competence to disagree. A survey of 319 knowledge workers associated greater confidence in generative AI with less reported critical thinking; it identified a shift toward verification, integration and task stewardship. It did not demonstrate lasting skill loss. The question for job design is how workers continue to acquire the expertise that makes review meaningful. Lee et al. · CHI, 2025

equilibri proposes paid practice, independent checks and mentoring: preserve the competence to disagree. Trust should follow demonstrated reliability in a specific task; fluency, friendliness and a colleague-like interface cannot establish responsibility or care.

Evidence & limits / autonomy

Stanford · whose work is being redesigned?

Stanford’s SALT Lab and Digital Economy Lab built WORKBank from 1,500 workers across 104 U.S. occupations, alongside expert assessments of AI capability. The study maps what workers want delegated against what agents can do, finding varied preferences for human involvement. Capability and consent are different questions. This is a survey-based preprint, revised in 2026, rather than a trial of workplace outcomes. Shao et al. · WORKBank, 2025–2026

King’s College London · choice matters alongside location

Surveys and interviews with UK hybrid workers examined return-to-office policies, work-location choice and well-being. Unclear or inconsistently applied policies could undermine the experience of choice even when employees retained flexibility. Office time itself was not linked to better well-being. These findings concern employee experience, not a randomized comparison of offices and coworking spaces. Clinton et al. · Human Resource Management, 2026

Further evidence

A 2025 study of 16,422 university-educated, full-time workers across 40 countries found that home working had broadly stabilized after its earlier decline. The pattern varied considerably by region. Aksoy et al. · PNAS, 2025

At Trip.com in China, a randomized trial gave some employees two home-working days a week. Across 1,612 participants, that hybrid arrangement cut resignations by a third without detectable harm to performance. Bloom et al. · Nature, 2024

The international survey describes graduate employees, not the entire global workforce. The Trip.com trial tested a specific hybrid schedule at one company; it did not test coworking spaces.

04 / SOCIAL INFRASTRUCTURE

Shared space should sustain shared judgment.

When execution becomes easier to obtain, a shared workplace should justify itself through what remains difficult: defining a worthwhile problem, exposing a mistaken assumption and transferring knowledge that has never been written down. Proximity alone does not provide these functions. They need time, facilitation and freedom to question someone more senior.

Research on scientific collaboration, creative ideation and weak ties suggests that different forms of contact support different exchanges. Physical proximity can matter for some conceptual work; digital networks can extend opportunity beyond a local circle. These findings support designing for particular interactions, rather than treating attendance as a general proxy for collaboration.

A forward-looking workspace could offer paid mentoring, voluntary case discussions and short peer reviews using non-confidential material. The apprenticeship question is especially important: if routine drafting is delegated, which tasks allow newer workers to develop judgment? A proposal worth testing is to have learners frame a problem independently, compare approaches and defend a decision before consulting an agent.

For equilibri, social infrastructure should support both professional learning and life beyond work. Shared meals, music and conversation need no productivity justification; participation and remote access should remain matters of choice.

Evidence & limits / encounter

Oxford · connection and invention are different achievements

Research co-authored by Oxford’s Carl Benedikt Frey analyzed 20 million papers and four million patent applications. Geographically dispersed teams were less likely to produce disruptive work; their collaboration was more concentrated in technical tasks than in developing concepts together. The analysis is observational and uses historical scientific and patent records. It supports examining the conditions for shared thinking, rather than treating proximity as a universal advantage. Lin, Frey & Wu · Nature, 2023

Further evidence

Laboratory and field experiments found that pairs on video generated fewer creative ideas than pairs meeting in person, while finding no disadvantage in idea selection. The result suggests assigning settings by cognitive task: divergent exploration and evaluative selection may benefit from different arrangements. Brucks & Levav · Nature, 2022

Large randomized LinkedIn experiments found that weaker connections could improve job mobility, with limits and differences by industry. A local community should therefore maintain connections beyond its own membership; a cohesive internal network and access to unfamiliar opportunities are distinct design objectives. Rajkumar et al. · Science, 2022

Three field experiments in an online labor market found that referrals conveyed information about performance and persistence beyond workers’ observable characteristics. Referred workers also performed particularly well alongside their referrers. These findings concern informed recommendations; they do not establish a coworking advantage or justify allocating opportunities exclusively through personal networks. Pallais & Sands · Journal of Political Economy, 2016

The design invitation is simple: leave room for introductions and unfinished ideas, while keeping remote participation available. This is the studio’s translation of the findings; these studies did not test a work playground.

05 / ATTENTION

Review capacity sets the limit of delegation.

A worker managing several agents can become a bottleneck for approvals, corrections and exceptions. The relevant limit is how much consequential work they can assess well. Launching additional tasks without reserving review time risks exchanging a production queue for a verification queue — with leisure reduced to waiting for the next interruption.

The proposed response is to budget human attention before increasing concurrency. Batch routine reviews, cap unfinished work and specify which exceptions justify interruption. If an agent encounters uncertainty during a protected interval, it should pause or pass to an explicitly assigned, paid duty role. A colleague silently absorbing the alerts is not a time dividend.

Rest remains valuable independently of this arrangement. A meta-analysis found small benefits of micro-breaks for vigor and fatigue, without a significant overall performance gain. That evidence concerns brief recovery; it does not establish that monitoring a running agent constitutes leisure. Albulescu et al. · PLOS ONE, 2022

The operational question is whether someone can leave without monitoring, checking or anticipating a response. Protected time requires a functioning handover or a system that can safely wait.

Evidence & limits / attention

MIT Media Lab · the story around a tool changes its use

Researchers at MIT Media Lab and Arizona State University varied how a chatbot’s motives were described before people used it. Those introductions changed perceived empathy, trustworthiness and effectiveness, as well as the interaction. The experiment involved a mental-health companion scenario, not a workplace deployment. Its workplace relevance is a design inference: calling a system an expert or a teammate can shape expectations before anyone checks its output. Pataranutaporn et al. · Nature Machine Intelligence, 2023

Further evidence

A review combining results across studies found that short breaks helped people feel more energetic and less fatigued, with small average effects. Overall performance did not improve significantly. Feeling less depleted is already a worthwhile result. Albulescu et al. · PLOS ONE, 2022

A two-week experiment blocked mobile internet on participants’ smartphones while preserving calls, texts and internet access on other devices. Sustained attention and self-reported well-being improved. The intervention distinguished reachability from continuous internet access, a useful distinction when designing availability arrangements. Castelo et al. · PNAS Nexus, 2025

The phone intervention was temporary and adherence was a limitation. It was not a workplace trial and does not establish the effects of a short phone-free lunch. It supports examining intentional breaks from constant connectivity.

06 / DISTRIBUTION

The gains should reach beyond delegable work.

The capacity to automate varies across occupations; the claim to a better working life should not depend entirely on that variation. Hosting, cleaning, care and maintenance sustain the settings in which knowledge workers gain flexibility. A settlement that rewards only directly measurable automation can overlook these contributions and transfer pressure to workers with less bargaining power.

equilibri proposes discussing gains at the level of the organization: how much supports compensation, staffing, training and time across roles? In a shared workspace, reduced-rate membership needs a funded allocation and paid hosts need predictable coverage. Neither inclusion nor leisure should depend on discretionary goodwill or invisible overtime.

Coworking research identifies potential value in community and discretion alongside problems with sound, privacy and ergonomics. A membership cannot change an employer’s workload expectations. The studio’s remit should therefore be explicit: provide facilities, peer learning and a forum for negotiating practices, while acknowledging the limits of its authority over members’ jobs.

Evaluate distribution as well as averages: who gains income and discretionary time, who performs the new review labor, and whose access is constrained by cost, care responsibilities or disability?

Evidence & limits / belonging

Michigan Ross · coworking is already a social arrangement

Research with several hundred members of U.S. coworking communities identified meaningful work, job control and community as explanations for reported thriving. Members valued both opportunities to connect and the option to work privately. This was not randomized assignment, so self-selection matters. It nevertheless documents a model already in use: independent work supported by a recurring community, rather than everyone sharing an employer. Spreitzer, Garrett & Bacevice · university research summary, 2016

Oxford Internet Institute · the workplace includes its rules

Fairwork’s AI for Fair Work report offers ten principles developed through consultation. It treats AI as a workplace governance issue, bringing working conditions into a discussion often centered on technical performance. This is a normative framework, not an experiment. For a shared workspace, the implication is to specify how tools, data and accountability will be handled alongside the physical design. Oxford Fairwork · AI for Fair Work

Further evidence

A review of eleven studies identified potential coworking benefits for networking, collaboration and productivity alongside privacy, ergonomic and acoustic problems. Evaluation should examine these competing effects together, including who avoids a space because its sensory or privacy conditions are unsuitable. Kinsman et al. · Work, 2024

A separate review connects quality of working life to hosting, community culture, services and the physical setting. This supports treating the operating model as part of the spatial proposition: the same floor plan can accommodate substantially different practices. Ciccarelli · CIDADES, 2023

These reviews bring together varied, often small observational and qualitative studies. The eleven-study review searched literature in 2020 and first appeared online in 2023. It offers useful design lessons, not proof that every coworking model improves health or employment.

07 / ECOLOGICAL SUFFICIENCY

Efficiency needs an absolute resource account.

Agentic workflows can multiply searches, model calls and retries in pursuit of a single accepted result. The proposed unit of assessment is the completed service at an agreed quality threshold, including failed attempts and human correction. Compare that arrangement with a credible alternative, then track total demand: a cheaper task can be performed more often, offsetting the benefit of doing it efficiently.

The IEA reports that global data-centre electricity demand grew 17% in 2025, while electricity use per AI task was declining. These observations illustrate why efficiency at the task level can coexist with increasing aggregate demand. The figures cover data centres and should not be treated as the footprint of AI alone. IEA · Key Questions on Energy and AI, 2026

Ecological sufficiency asks which services support a good life and how they can be provided within material limits. The IPCC’s demand-side assessment places well-being and service provision within mitigation, rather than assuming ever-greater resource throughput is the objective. For equilibri, that informs a preference for adaptive reuse, shared facilities, nearby leisure and computation proportionate to a task. These are design choices to evaluate, not a demonstrated carbon saving. IPCC · AR6 Working Group III, Chapter 5, 2022

Set resource budgets before execution: choose a sufficient tool, limit speculative parallel runs, reuse valid results and stop when acceptance criteria are met. Assess commuting, building use and consumption separately. More autonomy is a social gain; a lower-impact arrangement needs its own evidence.

Evidence & limits / ecology

Time use / rebound matters

Research using Swedish household consumption and time-use data models how shorter working hours could affect energy use and emissions. Its results depend on income changes and the activities that replace work. Those assumptions cannot be transferred directly to an AI-enabled arrangement that maintains pay. Additional leisure might support walking, reading or shared meals; it might also support more travel and consumption. Nässén & Larsson · Environment and Planning C, 2015

Assessment / compare whole arrangements

Compare the proposed day with a defined baseline: travel to the workspace, the life of its fit-out, building energy, and the incremental digital services used. Track absolute demand alongside efficiency per task. Where cloud providers do not disclose enough data, report that uncertainty. No environmental benefit should be inferred from an attractive material palette or from fewer human work hours alone.

02 / THE ATLAS

The future of work
has local roots.

Can a worker convert thirty released minutes into anything they value? Distance to care, public space and other people determines whether temporal flexibility is usable. The local question is how much life becomes accessible without another journey or purchase.

A comparative reading of places, rather than a league table of workplace cultures. The inquiry is situated in Trudy Hall’s experience of living in Lower Manhattan, San Francisco and Europe. These experiences frame questions, not a comparative field study. The references examine how urban form, institutions and shared spaces support life beyond the interface.

EUROPE / THE CONDITIONS OF CHOICE

Freedom depends on what surrounds it.

A laptop can make work portable. Transport, care, housing and access to a nearby place to work shape how useful that freedom becomes. Europe’s COST network on new working spaces examined coworking, inclusion and peripheral regions, widening the conversation beyond major cities. COST · CA18214

Eurofound places autonomy, ergonomics and social dialogue within the design of hybrid work. Its reporting also identifies a gap between demand for flexibility and the remote opportunities available. Eurofound · hybrid work, 2025 European trends · 2025

For equilibri, the design question is whether flexible work can fit around care, mobility and household budgets. Evaluate access by total time and cost of participation, including the journey, rather than the desk price alone.

SAN FRANCISCO / BAY AREA / THE PACE OF INVENTION

Fast tools still need a human setting.

San Francisco offers a close view of technological invention. The wider region’s Silicon Valley Index documents concentrated AI investment alongside housing and income pressures. A productive economy and a livable working day are related ambitions; one does not secure the other. City of San Francisco Silicon Valley Index · 2026

The Commons places recurring gatherings and member-led events alongside coworking. In its account, people share a setting for intellectual and social life as well as a place to get work done. The Commons · operator source

For equilibri, the question is how a culture of rapid experimentation can also sustain durable relationships and independent judgment. Recurring peer sessions could examine failures and trade-offs without turning every gathering into a pitch or a performance.

MANY RHYTHMS OF WORK

A useful model has to adapt. The 2025 Global Survey of Working Arrangements found substantial regional differences, with more home working in English-speaking countries and less in East Asia. The survey concerns university-educated employees; it does not describe every worker or occupation. A shared workspace should respond to local routines rather than assume a single global working day. Global Survey of Working Arrangements · 2025

THREE SPATIAL IDEAS

A change of setting, within the day.

LISBON / PORTUGAL

Bring the garden inside.

Second Home Mercado makes plants and natural light central to its workspace. For equilibri, the invitation is to design around living greenery from the outset.

Second Home
ULUWATU / INDONESIA

Make stepping outside easy.

BWork combines workspaces and meeting rooms with a garden and outdoor lounges. The idea to carry forward is a simple transition between concentration and a different pace.

BWork
CAPE TOWN / SOUTH AFRICA

Give the in-between a purpose.

Workshop17’s Kloof Street precinct connects buildings through a courtyard and offers a terrace. Shared outdoor areas suggest another setting for the conversation that continues after a meeting.

Workshop17

Operator descriptions inform these design readings; they are not measured outcomes or a record of personal visits.

03 / THE PRINCIPLES

A protocol for
the hybrid day.

A proposed working agreement for employees, independent practitioners and teams. Put these terms in place before delegation, then revise them against actual quality, income, workload and recovery.

Specify the outcome.

Define the outcome, evidence, budget and deadline with affected workers. Name the accountable decision-maker and route for appeal. Agree how gains and review costs will be shared.

Delegate authority precisely.

Separate drafting from consequential action. Minimize personal data, restrict access and set expiry and stop conditions. Make permission revocable; test reversible tasks before widening authority.

Budget the review.

Reserve paid time and independent evidence for review. Give reviewers authority to reject outputs without penalty. Cap concurrent tasks; outside agreed availability, pause or fund explicit coverage.

Reassess the arrangement.

Compare quality, income, workload and resource use with the baseline. Examine unequal impacts and hear challenges from affected people. Revise or retire workflows that shift harm onto others.

ONE DAY / AN ILLUSTRATIVE RHYTHM

Work and leisure, with room between them.

A research brief illustrates the sequence. These are proposed phases, with an agreed off-duty interval and time reserved for review. The agent may run or pause during that interval; leisure is not conditional on continuous machine activity.

  1. Frame & delegateDefine the research question, permitted sources and evidence standard. Authorize a draft; reserve publication for human approval.
  2. Step into lifeLet routine questions wait until the review window. Step away under the agreed coverage plan, without monitoring notifications.
  3. Return & judgeCheck decisive claims against original sources, examine omissions and test an alternative interpretation. Accept, revise or reject.
  4. Finish & releaseRecord the accepted result, review time and unresolved limits. Close the task; negotiate extra scope instead of silently extending the day.

The architecture needs to respect these choices. In two workplace studies, removing spatial boundaries reduced face-to-face interaction. An open floor plan alone is no recipe for connection. Bernstein & Turban · 2018

A membership should offer useful facilities and human hospitality. It should also fit into a wider neighborhood of libraries, parks and public places where belonging does not depend on buying a desk.

04 / THE SPATIAL PROPOSITION

equilibri, as a place.

A neighborhood work playground would test whether shared facilities, peer learning and explicit work agreements make agent-assisted life more viable. Its research value lies in separating the effects of the space from those of the practices adopted within it.

THE SPACE

Design for distinct modes of attention.

Adapt an existing building near public transport, with daylight and step-free access. A quiet laptop room would provide ergonomic desks, task lighting, power and reliable internet; enclosed rooms would contain calls and meetings.

A shared kitchen table, reading corner and bean bags would offer a change of pace. Generous planting and carefully chosen furniture would give the interior its character, with a shaded, accessible patio for lunch, conversation or optional outdoor work.

THE DAILY EXPERIENCE

Make peer learning part of the infrastructure.

A paid host would protect quiet hours and facilitate voluntary introductions and shared meals. Separately funded peer sessions could examine delegation briefs, compare verification methods and rehearse recovery from tool failure using non-confidential examples. Experienced practitioners would be paid for mentoring. The aim is to make adaptation a supported practice, with clear scope and payment for any commissioned work.

Personal productivity would stay private. Research consent would be optional and revocable, independent of membership. Collect minimal data, define deletion periods and publish aggregates; never sell member data or attention.

MATERIALS / USE LESS, KEEP LONGER

Concrete

Retain and repair existing slabs where practical. Avoid a new pour for appearance alone; compare environmental declarations for essential new material.

Plywood

Plywood walls, desks, shelving and shared tables would give the interior a warm, coherent character. Use reclaimed or verified FSC-certified panels, low-emitting finishes and removable fixings so walls and furniture can be repaired or reconfigured.

Cork

Specify replaceable flooring and pinboards where suitable. Check binders, emissions and durability; design acoustic control separately.

Reclaimed timber

Give sound salvaged wood another life as tables, shelving or benches. Inspect its condition and previous treatments, and use repairable joints so it can be adapted again.

Hemp

Hemp-based acoustic panels would soften sound around shared work areas, adding texture alongside the plywood. Select products with disclosed binders, tested acoustic and fire performance, and a clear plan for reuse or disposal.

Salvaged brick

Reuse suitable local brick for patio paving or landscape edges. Check durability, drainage and an even, accessible surface; let the available material shape the design.

Reuse comes first. Carbon claims would require a whole-fit-out assessment, including transport and replacement. EPA · material reuse EPA · salvage and reuse SAPPA · hemp acoustic panels EPA · embodied carbon FSC · certified products EPA · indoor air

MEMBERSHIP & REVENUE

A place in your working rhythm.

Flexible and dedicated-desk memberships would fund everyday hosting and community activities. Day passes, team packages and meeting-room bookings would fit around capacity. Reduced-rate places would need a defined revenue allocation or committed partner funding.

Prices and access rules would be transparent, with membership based on a shared code of conduct.

COSTS & VIABILITY

A foundation for the long term.

Startup costs would include the deposit, surveys, essential fit-out, furniture and contingency. Running costs would cover rent, hosting, cleaning, utilities, insurance, maintenance and reserves — including plant care and patio upkeep.

Recurring membership contribution, plus conservative room and day-pass contribution, must cover fixed costs. Contribution means revenue after direct service costs; startup recovery and cash runway need separate planning.

Test the model before taking a lease

Demand and daily life.

Run a consent-based pilot in an existing venue. Separate two questions: does the delegation agreement improve the working day, and does the shared setting add value? Compare similar workflows with and without the agreement; where feasible, compare its use in different settings. Track accepted quality, review and correction time, on-call waiting and protected leisure. Record tool changes and task selection so novelty or easier assignments are not mistaken for durable gains.

Proceed only with demonstrated demand, adequate reserves and a workable lease. Complete local access, safety and fit-out reviews; select plants for actual light and climate, with clear routes, drainage and ongoing care.

Opportunity and impact.

Predefine measures of perceived time control, fatigue, work quality, social connection and access. Track total earnings, benefits, actual hours and income volatility alongside effective hourly pay; shorter hours must not conceal lost income or unpaid review work. Include hosts and support workers as participants in the evaluation. Use a comparison group where feasible and report selection effects, attrition and differences across occupations. Treat interviews as evidence of experience rather than proof of causal impact.

Assess travel, building energy, material reuse and incremental computation against a defined baseline. The proposition would need revision if leisure shrinks, real income falls, review costs erase the gain, or burdens shift to other workers. Keep research participation separate from membership, minimize data collection and publish aggregate findings, including null and adverse results.

A research and design proposition. No venue opening, pilot results or validated financial forecast are claimed.