It is 11:47 p.m. somewhere in Madrid. A woman is sitting on the edge of a borrowed bed, a child asleep against her hip. Her phone shows a message from a caseworker she met for fifteen minutes that afternoon: “Tomorrow, 9 a.m., calle López de Hoyos. Bring everything.” She doesn't know what everything means. She doesn't know the metro stop. She doesn't know whether the photocopy of her birth certificate — the one she carried across three borders — will be accepted. She is alone with a phone and a question.

I built Acogida.es so that, at 11:47 p.m., she has somewhere to go before tomorrow at 9.

Today the site launches. It is free, multilingual, and asks for no login. Your data stays in your browser. The rest of this is not a product post. It is the letter I owe to anyone wondering where this came from. Acogida is not a year-long sprint. It is the corollary of a longer trajectory — a question I have been carrying, in different cameras and different code, for the better part of twenty years.

The woman above is a composite. Every detail is taken from real conversations. I will not put a real person's bedroom into a launch post.


The first border I ever crossed was internal

My parents packed up their lives in the south of Spain and got on a train north to Catalonia. They did not cross a state line. They crossed a country, an accent, a language, and — for years — a quiet line that reminded them they were from somewhere else inside their own state. I was born on the receiving end of that train.

So when I say welcome, I am not speaking from theory. I am speaking from my mother's kitchen table. From mispronounced surnames. From the silent, daily effort my parents made to not be in the way while they were building something. My mother's southern accent got softer year by year, like a coin rubbed thin in a pocket. Some part of me has spent my adult life trying to give that coin back to her in another form.

Spain has not yet finished telling itself the story of its own internal migration. And it is the same story it now refuses to recognise in the people arriving from Bogotá, Caracas, Bamako, Damascus, Karachi.

Different decade. Different distance. The same train.

A camera on the Mediterranean shore

I was in my late teens when I joined the Taller Bigas Luna — the small, stubborn film lab Bigas built in Barcelona to play with digital and documentary, with whatever rules he had not yet broken. Bigas had a phrase he kept coming back to: cine vital. Vital movies. Films that come from blood, kitchen smells, mistakes, weather and bodies — not from a moodboard. He taught us, in the way teenagers learn: by watching him refuse anything that did not have a heartbeat in it. That filter never left me, and it is the same filter I have tried to apply to software ever since.

There I shot my first documentary: a fieldwork piece along the Mediterranean ribera, interviewing people who had arrived by boat, by plane, by silence. What stays with me are not the shots. It is the pauses. A man who took forty minutes to tell me his mother's name. I kept the camera rolling. I had no idea what to do with the footage. I had no idea what to do with him.

Then I shot a short fiction film about a love story between two people separated by a piece of paper. A visa. A stamp. Three months. That was my first serious intuition that:

The border is not in the water. The border is the paperwork.

I carried that intuition for twenty years like a bruise that would not fade. Every time I read about another shipwreck, every time a friend's cita previa got pushed six months out, every time the news anchor said oleada, the bruise pressed back.

A small NGO, a lot of Spanish bureaucracy

When I started university, I helped run a small NGO that worked with newly arrived migrants on the only thing that actually mattered to them in their first months: integration. Not posters. Not declarations. Forms. Appointments. Photocopies. The empadronamiento. The school enrolment. The recognition of a foreign degree. Where the metro station was. Which civil servant was kind. Which one to avoid on a Tuesday.

I was twenty. I had no idea what I was doing. The people I sat next to had crossed continents and were teaching me, in halting Spanish, how Spanish bureaucracy works. We were a handful of volunteers and a phone. We helped maybe a few dozen people, badly. But I learned a thing I have never been able to unlearn: welcome is not the door. Welcome is the hallway after. A bed for one night is shelter. A job, a school place, a recognised qualification, a way to say good morning to your neighbour — that is welcome.

Sat4survive: a miniFrontex, but humanitarian

A few years later, in the middle of the Syrian migratory crisis, I built one of the first deep-learning open-source projects I ever shipped. I called it sat4survive.

It was, in plain language, a small humanitarian counterpart to Frontex's eyes — the same kind of computer vision pointed at the same horizon, but for the opposite question. Frontex was, and is, optimised to control the border. I wanted a system optimised to find people in time. The pipeline combined Sentinel satellite imagery with convolutional neural networks trained to flag the visual signature of an adrift cayuco in open water — a small wooden boat, often invisible to anything that is not actively looking. The idea was simple and a bit naive: if a satellite can see a wave and a hull, a satellite can see a person. We just have to decide what we are looking for.

It was rough. It was a multi-week project that I kept coming back to. I open-sourced it at github.com/apolmig/sat4survive, when sharing weights and notebooks for humanitarian computer vision was much less common than it is today. It did not save anyone by itself. But the conviction it embodied has not aged: civilian, transparent, public-good AI on the same satellites and the same models that states use to police us is not a hobby. It is a posture.

That posture is the through-line of everything that comes next.

Refugeeslive: one of the first social-impact open-source LLMs

There came a point where filming, and even satellites, were not enough. The bottleneck had moved. By the late 2010s the pinch point was no longer seeing people in distress. It was answering them once they arrived — in their own language, with the right form, on the right day, before the office closed.

At Saturdays.AI — the AI-for-social-good community I co-founded — we worked on something that still feels urgent: refugeeslive, an early experiment in tuning a language model to answer, in the speaker's own language, the questions refugees actually ask. How do I file for asylum? What does this appointment letter mean? What do I do if I am being thrown out tonight?

It was rough. It was a weeks-long project, not a weekend hack. And it was, to my knowledge, one of the first social-impact open-source LLM efforts ever released — trained, documented, and pushed publicly at a time when most of the open community was still arguing about whether language models would ever be useful at all. It worked just often enough to make me unable to forget it.

It was not the only one. The Saturdays.AI community archive holds dozens of weeks-long prototypes — early gender-violence detection, accessibility tooling, mental-health triage, language rights, AI for the Spanish public health system — built by volunteers in multi-week sprints, alongside their day jobs, with a refusal to pretend the world was fine.

I left those projects with one question that would not go back in the box:

If AI is going to rewrite the world, who is sitting at the table when it gets rewritten?

I had been the man behind the camera, recording the pauses. Then the man behind the satellite, scanning for hulls. Then the man behind the model, training it to answer in someone's first language. None of those was, by itself, enough. The pauses did not need another witness. They needed someone to put a phone number, a date, a metro stop, a printable letter into the hands of the person doing the pausing.


A long trajectory, not a year-long sprint

Different decades. Different distances. The same train.

Acogida.es is what happens when a teenager who was rolling a camera in front of pauses, a student who spent his evenings helping a small NGO move people through Spanish bureaucracy, a developer who pointed satellites at the Mediterranean to find boats before headlines did, and an AI volunteer who trained an early open-source LLM to answer asylum questions in someone's first language — finally stops moving and sits down to build the tool he has been trying to build, in pieces, for twenty years.

Different stops. Different routes. The same question, refusing to go away: how do I make myself useful here, with the capabilities I happen to have at this moment?

This is that answer, for this moment.


Why now, in three numbers

I am not going to drown this letter in statistics. Three numbers are enough.

1 in 5 People living in Spain with migrant origin (9.96M, 20.28% of the population, 1 Jan 2025)
~90% Of net new employment filled by foreign-born workers (Jan 2024 – Mar 2025)
500K+ Undocumented people who may benefit from the 2026 regularisation. Deadline: 30 June 2026.

One in five. As of 1 January 2025, 9.96 million people living in Spain — 20.28% of the population — have migrant origin (INE; Servicio Jesuita a Migrantes, 2026 report). Spain is already a country built by arrival. Pretending otherwise is not nostalgia, it is denial.

Nine in ten new jobs. Between January 2024 and March 2025, foreign-born workers filled roughly 90% of net new employment in Spain (Real Instituto Elcano). The country's central bank and its independent fiscal authority both say openly that without sustained migration, the pension system does not survive the next two decades (Banco de España; AIReF; CaixaBank Research; BBVA Research). Every retired teacher in this country, increasingly, is paid by someone who arrived.

One door, closing on 30 June. On 15 April 2026 the Council of Ministers published a Royal Decree (BOE) opening an extraordinary regularisation that may benefit more than 500,000 undocumented people with at least five months of prior residence. Deadline to apply: 30 June 2026. No extension. Eighteen days earlier, on 12 June 2026, the EU Pact on Migration and Asylum becomes operational, with a contested list of “safe countries of origin” (Council of the EU, December 2025) that includes Colombia — the same country that tops Spain's irregularity gap.

In other words: a once-in-a-decade chance to bring half a million people out of the shadows, opening at exactly the moment European procedure is getting harder for the next half million.

THE 76-DAY WINDOW From Royal Decree publication to application deadline 58 days 18 days Today 26 Apr 2026 15 Apr 2026 Royal Decree opens regularisation 12 Jun 2026 EU Pact operational “safe countries” list active 30 Jun 2026 Application deadline no extension
Two windows overlap: a national door opening, an EU procedure tightening. Eighteen days between them.

This is the moment in which a tool like Acogida.es either becomes useful or arrives too late. I did not have the option of arriving too late.

What I actually built (the short version)

A free, multilingual website that walks a person, step by step, through the things Spain quietly assumes they already know: how to file for the regularisation, how to find a cita previa, what to put in a carta de motivación, how to enrol a child in school, what the padrón is, what to do at 11:47 p.m. when somebody is being thrown out of a flat. It speaks Spanish, English, Arabic, French, Ukrainian and Romanian, with right-to-left layout where it should be. It calls 112 and 016 before it calls anything else.

It runs on the user's own phone. It does not ask them to register. Their data does not leave the browser unless they choose to share it.

A moral position, not a feature

The people who most need this tool are also the most exposed — to surveillance, to data leaks, to a future government that may not be friendly. A welcome that puts the person at risk is not a welcome. It is a trap with better lighting.

The source is open and a longer engineering post is coming. This letter is about the why, and the why is not technical.


Welcome is the hallway, not the threshold

If welcome were only the door, this tool could be a single page with a phone number on it. It is not. The hard, slow, undramatic part of welcome is the hallway after the door — and that is the part Acogida is actually built for.

So the tool does not stop at the regularisation form. It walks with the person through the unglamorous machinery of becoming someone who can contribute:

  • Work. Which permit goes with which job, what an employer must sign, when the NIE becomes a working NIE, how to write a CV that actually gets read in Spain, where to ask for help if the job offer that brought you here turns out not to exist.
  • School. How to enrol a child mid-term, what the local concertada / pública distinction actually means, what to do if the school refuses a child without a padrón (it cannot, and the tool will tell you so in your language).
  • Qualifications. The homologación and equivalencia tracks for foreign degrees, the difference between them, and the paths that exist when the paperwork from your country of origin is genuinely lost.
  • Health, housing, language, civic life. Where the public health card comes from, how the rental market really works, where free Spanish (or Catalan, or Basque, or Galician) classes are, how to register to vote in the elections you are entitled to vote in.

This is the part that takes years, not minutes. It is also the part where a country either earns the word civilised or quietly fails the test. A migrant who can work, study, learn the local idioms, raise children in the school down the street, contribute taxes, disagree with their neighbours like everyone else, is no longer “a migrant” in the political sense. They are a neighbour. That conversion — from category to person — is the actual product of integration. Everything else is theatre.

The Mediterranean has never been a wall

It is worth saying this plainly, because the public conversation often pretends otherwise.

Spain has been mixing for three thousand years. Phoenicians, Romans, Visigoths, Berbers, Jews, Castilians, Andalusians, Catalans, Galicians, Asturians, Basques, Latin Americans, North Africans, sub-Saharans, Eastern Europeans, Filipinos, Chinese, Pakistanis, Ukrainians. Every richness this country has — its food, its language, its music, its science, its football, its architecture, its very face — is the result of mixing. The Mediterranean has never been a wall. It has been a corridor. To pretend otherwise is to ask Spain to forget what it is.

And the mix has not stopped working. The same data that the political conversation tries to weaponise tells a quieter, more honest story: the country is older than its workers, the workers are increasingly arrivals, the arrivals are increasingly building the lives the country needs them to build. We are not, on net, helping them. They are, on net, helping us. A society that pretends the relationship is one-directional is a society that has stopped reading its own balance sheet.

The point is not to be hospitable as a favour. The point is to be honest about who we already are, and to make the next chapter of that mixing easier on everyone in it — arrivals, neighbours, employers, schools, parishes, town halls, all of us.

Acogida.es is, in the end, infrastructure for the next chapter.


The Christian case, without apology

Part of what made me actually open the editor and start writing this code is a Christian conviction. I am not going to bury that.

“When a stranger sojourns with you in your land, you shall not do him wrong. The stranger who sojourns with you shall be to you as the native among you, and you shall love him as yourself.”

— Leviticus 19:33–34

“I was a stranger and you welcomed me.”

— Matthew 25:35

“Do not neglect to show hospitality to strangers, for thereby some have entertained angels unawares.”

— Hebrews 13:2

There is no honest reading of the gospel in which the migrant is a problem to be solved. Hospitality is not the decorative trim of the faith — it is the final exam. And that exam is not passed with feelings. It is passed with structures: just laws, neighbours who learn names, tools that work at 11:47 p.m. when the office is closed and the caseworker is asleep.

If you do not share the faith, you do not need to. The conviction translates without losing anything important: a society earns the word civilised in how it treats the person standing at its door with a child on her hip and no one to ask.

Acogida.es is, for me, a very modest way of showing up to that exam.

Walls, hugs, and what I would rather build

The migration debate in Europe is broken — when it reduces people to costs and risk vectors, and also when good intentions replace concrete tools.

There are three doors out of this:

1
Walls

Politically cheap. Humanly expensive. We already know how this story ends.

2
The rhetorical hug

Warm words that never become a form, a phone number, a deadline, an outcome. Comfortable for the speaker. Useless for the listener.

3
Civic infrastructure

Clear processes. Accessible digital tools. Data in the hands of the people they describe. Real access to advice. Integration that can be measured, not just declared.

Acogida.es is one small contribution to door three. It does not replace caseworkers, lawyers, NGOs, or the public administration. It amplifies them. A volunteer at CEAR, ACCEM, the Red Cross, the Jesuit Migrant Service, a parish, a town hall — they can drop a single WhatsApp link into a conversation and let a person who is alone at 11:47 p.m. wake up tomorrow with a plan instead of a panic.

That, to me, is applied democracy: pushing the cost of accurate, official information toward zero, in the language of the person who needs it most.

What I owe you, honestly

A launch letter that does not name its limits is not honest.

The AI can be wrong. Eight of our supported languages still fall back to Spanish, and that is not good enough. The legal landscape moves faster than any tool. The 2026 regularisation is a once-in-a-decade door, and I will not catch every edge case before 30 June. The “safe countries” list will harm individual asylum claims that this tool, by itself, cannot fix.

What I commit to: open code, clearly labelled fallbacks, a public changelog, a way for people to tell me what is broken — and turning the tool off if it ever stops being honest about what it does not know.

The ask

If you work at an NGO, parish, union, professional association, town hall or social-service team: try the tool, break it, send me what is missing. It is built to work with you, not around you. One link in a WhatsApp group tonight is one person with a plan tomorrow.

If you are a migrant or refugee: open Acogida.es. It is free. It does not ask you to register. Your data does not leave the browser. If you find something wrong, write to me.

If you build technology: read the code, file an issue, send a pull request. We especially need fluent contributors in Urdu, Bengali, Hindi, Persian and Turkish — those languages still fall back to Spanish, and that is not good enough.

If you make policy or law: protect the framework that makes this possible. The 30 June 2026 deadline is a historic opportunity. Missing it would be a decision, not an accident.

If you pray: pray for the people arriving. Then get up and make room for them.


It is now 11:48 p.m.

Somewhere in Madrid, a woman has stopped scrolling. The screen on her phone shows a checklist in her own language, three documents to put in a folder, a metro stop, a bus line, the name of the office, and a number to call if anyone tries to send her away. She does not know me. She does not need to. Tomorrow, at 9, she will not be alone with a question.

That is all this is. That is all this has to be.

Try it · Share it · Break it

Acogida.es is free, multilingual, login-less, and runs entirely in your browser. If a single WhatsApp share tonight gives one person a plan tomorrow, this letter has done its job.

Sources

Bibliographic references below. Direct URLs are included only where they point at the exact document or page that supports the claim. The named publishers and reports are themselves the canonical sources; finding the latest edition takes one search on the publisher's site.

Demographics & Labour Market

  1. Servicio Jesuita a Migrantes (2026). Informe sobre Población de Origen Inmigrado en España, 2025.
  2. Instituto Nacional de Estadística (INE). Estadística Continua de Población — 1 de enero de 2026.
  3. Real Instituto Elcano. Inmigración y mercado de trabajo en España.

Pensions & Fiscal Sustainability

  1. AIReF & Banco de España. Long-term sustainability analyses of the Spanish pension system.
  2. CaixaBank Research. Reforming the pension system.
  3. BBVA Research. Immigration: temporary relief for pensions.

EU & National Legal Framework

  1. European Commission. Pact on Migration and Asylum. Operational from 12 June 2026.
  2. Council of the EU (December 2025). EU list of safe countries of origin.
  3. Boletín Oficial del Estado (BOE), 15 April 2026 — Royal Decree, extraordinary regularisation. Practitioner overviews at Balcells Group and Remote from Spain.

Author's Prior Work

  1. Guerrero, M. sat4survive. Open-source CNN + Sentinel pipeline for early detection of adrift cayucos in the Mediterranean.
  2. Saturdays.AI — AI-for-social-good programme, and the refugeeslive prototype (one of the first social-impact open-source LLM efforts released).