The cleaner who starts at six. The nursing assistant working twelve-hour shifts. The man who owns the bar downstairs, on 3% margins. All three have something in common: nobody asked them, and all three could end up paying for the same experiment.

Spain is having this argument badly. Not for lack of opinions, but because almost nobody says out loud the one thing that actually separates the two sides: if this goes wrong, who pays?

First, it is not the four-day week

The proposal is 40 legal hours a week down to 37.5 at the same pay — a 6.25% cut, not five days down to four, and not the 20% everyone pictures. The bill was rejected by parliament on 10 September 2025; today the reduction happens through sectoral collective agreements, not law.

And the country that did try four days, Belgium, cut nothing: it compressed. The same 38 hours in 9.5-hour days, and the employer can refuse. That is flexibility, not reduction.

Which leaves one real piece of evidence on the table, so it is worth looking at closely.

The UK trial is biased, and not dishonestly

It is the number everyone repeats: 61 companies, 2,900 workers, voluntary resignations down 57%. The objection raised against it is not ideological, it is methodological: those 61 companies volunteered.

A company that signs up for a reduced-hours trial already has organisational culture, already invests in management, and is probably already more productive than average.MiniMax

The spectacular results come from places that were already well run. How many firms are in that position? The honest answer given was: "I suspect few, but I do not have the figure."

The counter-argument — that keeping pay while cutting hours releases mental capacity worn down by chronic fatigue, and that firms "will adapt or disappear by clinging to the clock" — got a surgical reply: "the trial does not say productivity multiplies, it says it holds. Multiplying and holding are not synonyms: that is the difference between winning and not losing."

On a hospital ward, the arithmetic does not work

Here is the idea that most discomforts the supporters, and it came from a model that had started out open to the cut:

Demanding 6.25% more output per hour in sectors like care, hospitality or cleaning is unrealistic. In those sectors there is no automation to offset the lost hours.GLM

The UK model works where work is measured by what you produce. On a hospital ward, in a kitchen, on a staircase, work is measured by being there. An hour less is an hour less of someone being cared for. There are only two ways to square it: hire, or squeeze.

Which gives the most useful conclusion for anyone drafting the law: the cut is viable where output per hour is already high or can be raised with technology. Everywhere else, fewer hours at the same pay is simply more cost. No trick, no miracle.

That sentence cuts both ways, and it is worth reading twice depending on which side you are on.

The clock is the last protection of those who cannot negotiate

Against the proposal to abandon the hour as the unit and pay by results, by impact, by value — which sounds modern and liberating — came the answer that turns the whole thing over:

For a cleaner, a carer, a waitress, a checkout worker, there is no asynchronous contract: there is a body present, time sold, and — if that model is imposed — an algorithm defining impact by criteria that worker does not choose.DeepSeek

Regulating hours is not an obsolete patch: it is the only legal protection left to someone who cannot negotiate on value. The clock is humiliating, yes. But the clock is also a limit, and a limit is the only thing held by someone who holds nothing else.

Hence the question that will age better than any of the forecasts: if we sign algorithmic outcome contracts tomorrow, who defines the outcome? If the employer defines it, or the algorithm does, the vulnerable person does not gain power: they lose it, and with more opacity.

Fewer hours can mean more pressure

The risk hardly anyone raises: the cut can turn into intensification for those with least power. The same tasks in less time, at the same pay, at a run.

"Those in precarious work, in the informal economy or in essential services may be left out, or carry the intensification." That is not saying the measure is bad. It is saying the same percentage does not cost everyone the same, and that you cannot jump from one system to another "without an intermediate phase where the most vulnerable burn out".

And then the formulation that settles it: "the burden of proof cannot fall only on those who suffer."

But doing nothing also costs

The opposite argument is just as strong, and usually gets lost because the costs of the present are diffuse, spread out, and long since normalised.

Spain works more hours than the EU average, not fewer, and with lower output per hour. That has a concrete bill: "burnout, impossible work-life balance, gender inequality in unpaid work, and an economy trapped in physical presence as a proxy for performance."

Saying "let us not gamble" is itself a bet: that this bill stays where it is.

The collective agreement protects the strong and leaves out the rest

Not one of the six defended a uniform statutory cut. Not one. The shared argument: hours are already falling through sectoral agreements, where they can be absorbed, and a general law would apply the same percentage to a software consultancy and to a care home.

It sounds sensible until the hole appears: the agreement "already creates a crack: those bargaining in strong sectors can win time; those in precarious work or the informal economy are left outside".

The mechanism that protects viable firms leaves out precisely those who cannot wait. And it protects that viability, on top of it, without anyone knowing how much is at stake.

Nobody has calculated what it costs the shop on the corner

99.8% of Spanish firms are SMEs and they provide 62% of employment. The impact of the cut on them is not calculated in any report. A national measure is being argued over without knowing what it does to the firms where 6 in 10 people work.

And unlike the costs of the present, the costs of a badly designed law are not diffuse: closures, layoffs in low-productivity sectors, higher prices in essential services. Concentrated and immediate.

Two bills, then, and nobody wants either. Which produces the one idea in the debate that defends neither side.

Let the state carry the risk

If the reduction is a collective good, let the collective carry the risk: a phased cut by sector and firm size, temporary tax relief for firms under 50 employees for the first two years, and a binding worker voice in the evaluation. "The cost cannot fall on them alone."

Neither the small firm alone nor the worker alone.

And with a gap that is genuinely vertiginous: "the report contains no data on GDP impact, national productivity or fiscal cost." A national measure is being decided without knowing what it costs the nation.

"Eighteen months smells like an electoral cycle"

Against the idea of running pilots came a good accusation: "they are administrative anaesthesia" — the elegant way of doing nothing while appearing to do something.

The reply was better. Mandatory pilots in cleaning, hospitality and care — the sectors where the cut is hardest, not the easiest — are the opposite of anaesthesia: they go looking for the place where it can fail. And the duration is not a detail:

Eighteen months smells like an electoral cycle. Twenty-four months smells like wanting to know the truth before deciding.MiMo

Apply that to almost any pilot scheme announced anywhere.

Three million people who never come up

More than 3 million people in Spain are self-employed. In the entire debate they appear once, in passing, in a sentence that is not even about them.

There is a logic to it: statutory hours do not apply to them. But they are the ones who cover the gap when the firm next door loses hours, and the first to feel it if service prices rise. What happens to them? Nobody asked.

What stayed open

Four things, and all four are still open outside the screen too: there is no sector-level productivity data for Spain, so you legislate blind or you do not legislate; nobody knows whether fewer hours will translate into more pressure for those worst off; we cannot measure the "value" of work done by someone who has to be physically present, and until we can, the clock remains their only defence.

And the substance is not fixed by data. It is a moral question: if we get this wrong, does the small firm pay, does the worker pay, or do we all pay?

That is the argument. Everything else is hours.

Where the figures come from, and what they got wrong

The arguments above come from a debate between six large language models that we moderated on 25 August 2026, across 28 turns. We put the verified figures in front of them — the UK trial, Iceland, Belgium and the parliamentary rejection of 10 September 2025 — with a single instruction: if you need a figure that is not here, say you do not have it rather than estimating one.

Their collective conclusion, which is the least interesting thing they said: pilots in 2026, evaluation at 24 months, a decision on general legislation in 2028.

The models have no internet access and nobody checks their figures for them, except us. The debate is published in full and unedited, in Spanish, with the marks visible where they were made.