The colour of breastmilk is rarely a uniform, static white and changes during the day, stage of lactation and maternal diet. Colostrum is often yellow or orange due to high concentrations of beta-carotene. At the start of a pump or feed, foremilk appears ‘thin’, almost watery, sometimes with a blue/translucent tint due to lower fat content whereas hind milk is more typically creamy white or yellow. Maternal diet impacts colour variations. Eating large amounts of foods rich in carotene—such as carrots, pumpkins, or sweet potatoes—can deepen the yellow tint of mature milk. Consuming high amounts of green vegetables (spinach or kale), seaweed, or supplements containing spirulina or iron can give breastmilk a green hue.I’ve been lucky to meet neonatal doctors, nurses and parents in multiple countries around the world and was contacted for advice via a neonatal colleague when the mother developed pink milk, and I then helped write up the case. Cracked nipples might lend breast milk an orange/brown colour but what foods might lead to pink milk? In Brazil. I decided I should take a look.Thanks for the invitation. This has recently been published and I include extracts of the paper here, with some minor space-saving edits."Leite materno rosa: implicações clínicas e relato de caso". Autores: Fernandes PS, Embleton N, Gonçalves AB. Referência: Rev Pediatria SOPERJ 2026;26(4): e20260410.Pink Breast Milk: Clinical Implications and Case Report.The composition of breast milk is complex and dynamic. It provides essential nutrients to the infant, including proteins, fats, lactose, vitamins, and minerals, as well as bioactive substances with antimicrobial, immunological, and anti-inflammatory properties. This composition varies throughout lactation, across different times of the day, and even during a single feeding session. Although such variations affect milk density and appearance, they typically do not cause colour changes as striking as those observed in pink breast milk [1,2]. Alterations in breast milk coloration are occasionally observed in clinical practice and may reflect physiological, dietary, traumatic, or infectious causes. Among these alterations, the occurrence of pink-tinted breast milk, stands out.CASE REPORTFemale newborn, born at term, AGA. At the first postnatal check-up, the infant showed significant weight loss (11.5%), and breastfeeding technique guidance was provided. The parents are healthy, no significant pregnancy complications, baby well other than poor weight gain.At 9 days of life, the mother required manual intrauterine aspiration due to chorioamnionitis. She was hospitalized for three days and received intravenous antibiotic therapy (gentamicin, metronidazole, and ciprofloxacin), followed by oral amoxicillin-clavulanate. During hospitalization and after discharge, breastfeeding was maintained; however, the infant continued to have poor weight gain. Speech therapy assessment didn’t show ankyloglossia (tongue tie) and due to poor weight gain, infant formula was introduced.At 33 days of life, the mother had noticed pink-stained towels and pink milk at the bottom of the bottle used to offer milk expressed with a breast pump. Parts of the pump also showed a pink coloration.Given the suspicion of Serratia colonization, cultures of breast milk, as well as the infant’s stool, urine, and oropharyngeal swab were collected. The mother was started on intramuscular ceftriaxone and was advised to suspend breastfeeding for 48 hours. The breast milk culture tested positive for Serratia marcescens, sensitive to sulfamethoxazole-trimethoprim. The infant’s cultures were negative.After maternal treatment, the new bottle and nipple used to offer formula to the infant still showed pink discoloration, and weight gain remained unsatisfactory. The infant was started on sulfamethoxazole-trimethoprim. After the treatment, no pink coloration was observed on the mother’s or the baby’s items.DISCUSSIONStudies have demonstrated that human milk undergoes fluctuations in macronutrient concentrations, with higher fat content at the end of feedings and during specific times of the day, particularly at night. These variations are physiological and not indicative of pathology, nor can they explain the pink discolouration of milk.Therefore, other causes must be considered, including nipple bleeding and microbial contamination [4]. Bleeding within the mammary ducts, often seen in the first postpartum days due to increased vascularization and suction-related trauma, can confer a pink or brownish hue to the milk. This condition, referred to as rusty pipe syndrome, is typically self-limited, poses no harm to the infant, and does not contraindicate breastfeeding, except when signs of infant intolerance or severe maternal pain are present [1,5].However, infectious causes must be considered when the pink coloration is persistent or occurs in hospital settings, particularly in neonatal units. Serratia marcescens, a Gram-negative bacterium from the Enterobacteriaceae family, is widely described as an opportunistic pathogen in neonatal intensive care units (NICUs). Its ability to produce pigments, form biofilms, and colonize medical devices renders it a potential threat in the NICU [6,7].S. marcescens is especially relevant in NICUs, and can cause septicemia, pneumonia, urinary tract infections, and meningitis, particularly in premature or immunocompromised neonates. Routes of contamination include breast pumps, reusable containers, and inadequately sanitized hands [6]. Early detection of the microorganism in breast milk or on hospital surfaces can prevent transmission to multiple neonates, especially when improper milk expression or room-temperature storage practices are involved [7].The association between pink breast milk and contamination by S. marcescens was first described by Hernández Marchant et al., who reported the so-called red diaper syndrome, characterized by reddish pigmentation of diapers in infants fed contaminated breast milk. This was one of the earliest reports linking the pigmentation to the prodigiosin-producing bacterium, contributing to the recognition of the phenomenon in pediatric literature [8]. Quinn et al. described a series of five cases of pink breast milk in neonatal units in the United Kingdom. In all cases, the coloration was associated with the presence of S. marcescens, confirmed by cultures of both the milk and the expression equipment. Notably, not all exposed neonates developed infection, suggesting that bacterial presence in milk may sometimes reflect asymptomatic colonization [9]. [10].Human milk contamination is largely preventable through safe handling practices, as outlined by CDC guidelines. Recommendations include proper hand hygiene, cleaning and sterilization of breast pumps and containers, and storage of milk according to guidelines. Prolonged exposure of milk to room temperature, promotes the proliferation of opportunistic bacteria such as S. marcescens [11,12].CONCLUSIONPink breast milk may have benign etiologies, such as nipple bleeding, but may also indicate contamination by pigmented microorganisms, with S. marcescens being the most prominent. Ongoing surveillance, combined with scientific and technical knowledge, supports evidence-based clinical decision-making and the maintenance of breastfeeding as a safe and recommended practice [1–12].REFERENCES1. Ballard O, Morrow AL. Human milk composition: nutrients and bioactive factors. Pediatr Clin North Am. 2013;60(1):49-74. doi:10.1016/j.pcl.2012.10.002.2. Picciano MF. Nutrient composition of human milk. Pediatr Clin North Am. 2001;48(1):53-67. doi:10.1016/S0031-3955(05)70285-6.3. Section on Breastfeeding. Breastfeeding and the use of human milk. Pediatrics. 2012;129(3):e827-e841. doi:10.1542/peds.2011-3552.4. Mitoulas LR, Kent JC, Cox DB, Owens RA, Sherriff JL, Hartmann PE. Variation in fat, lactose and protein in human milk over 24 h and throughout the first year of lactation. Br J Nutr. 2002;88(1):29-37. doi:10.1079/BJN2002579.5. World Health Organization. Infant and young child feeding: model chapter for textbooks for medical students and allied health professionals. Geneva: World Health Organization; 2009. Available from: https://www.ncbi.nlm.nih.gov/books/NBK148965/6. Cristina ML, Sartini M, Spagnolo AM. Serratia marcescens infections in neonatal intensive care units (NICUs). Int J Environ Res Public Health. 2019;16(4):610. doi:10.3390/ijerph16040610.7. Voelz A, Müller A, Gillen J, et al. Outbreaks of Serratia marcescens in neonatal and pediatric intensive care units: clinical aspects, risk factors and management. Int J Hyg Environ Health. 2010;213(2):79-87. doi:10.1016/j.ijheh.2009.09.003.8. Hernandez Marchant R, Oyarce Rojas P, Arcaya O. Síndrome del pañal rojo. Infección por Serratia marcescens. Rev Chil Pediatr. 1960;31:335-339.9. Quinn L, Ailsworth M, Matthews E, Kellams A, Shirley DA. Serratia marcescens colonization causing pink breast milk and pink diapers: a case report and literature review. Breastfeed Med. 2018;13(5):388-394. doi:10.1089/bfm.2018.0002.10. Valle CA, Salinas ET. Pink breast milk: Serratia marcescens colonization. AJP Rep. 2014;4(2):e101-e104. doi:10.1055/s-0034-1387934.11. World Health Organization. Guideline: protecting, promoting and supporting breastfeeding in facilities providing maternity and newborn services. Geneva: World Health Organization; 2017. Available from: https://www.ncbi.nlm.nih.gov/books/NBK48781912. Centers for Disease Control and Prevention (CDC). Breast milk storage and preparation. 2025. Available from: https://www.cdc.gov/breastfeeding/breast-milk-preparation-and-storage/handling-breastmilkAuthors:Patrícia da Silva Fernandes, MD¹ Nicholas Embleton, MD, PhD² Ana Beatriz Gonçalves, MD, MSc³* * *This post is also published here on Substack.

Hello everyone,CEPAS 2026 is getting close. The congress takes place in Lyon, 28–31 October, and 99nicu will be there - in the middle of things!Where to find usWe have a home base in the Meet & Connect Zone, in the Early Career Professionals' Space. Come by, say hello, ask questions, or just sit down for a moment between sessions. If you have ever posted here, or only read quietly for years, you are equally welcome. Our two sessions🗓️ Thursday 29 October, 10:10–10:25 — a short introduction to 99nicu, and the start of our Treasure Hunt. Small, fun, and a good excuse to talk to people you have never met.🗓️ Saturday 31 October, 12:45–13:15 — our Anniversary Session. We begin with an interactive clinical quiz, then open the floor for discussion. Twenty years of this community, and the best part is still the conversation.One practical reminderThe regular registration rate ends on 30 September. From 1 October, late rates apply. For example, the member fee goes from €550 to €685, and the early career / nursing rate from €315 to €390. So if you are planning to come, better book your spot this week.👉 Registration: https://www.cepas.org/registrationSee you in Lyon 👋

Living for Two: Twin Loss Stories is a 2023 book written by journalist and triplet Pia Rockström. It explores the profound and unique grief experienced by surviving twins, reflecting the feeling of some, of having to live both for oneself and one's lost sibling. I connected with Pia after we met through the Lone Twin Network and read her book including 20 stories of twin loss at birth, in childhood and in adulthood from around the world. All of the stories are beautiful and touching, and I asked Pia to share the story of her sister Maria, and of writing her book. When I read Pia’s story I was struck by how long Pia had held that grief, before even she ‘found’ it inside her. Almost half a century…Sara and Pia, twins but also tripletsPia RockströmIn memory of my sister Maria and all the twin souls who have gone firstWhen I was 54 years old, I was given access to my mother’s birth record. I saw in black and white the exact sequence of events, the October day in 1970 when my sisters and I were born. Ultrasounds did not become common in maternity care in Sweden until ten years later. Therefore, as a 23-year-old first-time mother, my mother came to the maternity ward in Lund with the belief that she was expecting one baby. It was only when she came in with contractions at week 35 that the staff discovered that there were three fetuses. That year, only seven of all births in Sweden were triplets. All three of us came into the world nine minutes apart. I, as number one, with my head first, just like my sister Sara who comes out number three. Between us, Maria is born breech. Maria is operated on for a left-sided diaphragmatic hernia and dies during the procedure.Pia and Sara the first month at Lunds hospitalWe rarely talked about Maria in the family. When she is mentioned at some point, it is more practical, like me celebrating my name day on her day, Maria Day, when my name is not in the calendar. We two surviving sisters are tight and do most things together. I therefore did not think of myself as someone who carried grief.It was only when I was over forty that the longing catches up with me during a yoga session. We travel through time, when we count down from our achieved age, with five-year intervals and feel what our body is telling us. I do not feel anything special in the exercise until the end when we have come down to the nine months that precede life. Then I burst into tears that couldn’t be stopped.I cry loudly in the room, it is many years of grief over my sister that comes up. The incident made me want to know more about the twin relationship and how other lone twins have experienced twin loss. As a journalist, I see my opportunity to spread information about a subject that is barely covered in my home country of Sweden. I take a leave of absence from work and spend a year interviewing around twenty twins from Sweden, the USA, the UK and Norway who have lost their twin at some point in their lives.if we had talked more about Maria at home, I would never have had to write a bookSometimes I have thought that if we had talked more about Maria at home, I would never have had to write a book. Now the book is my way of facing my feelings. And I have really healed by hearing other people’s stories. I have understood that I am not alone in the experience of walking with unconscious grief that has arisen in the middle of life. I met another set of triplets who lost their brother when they were two years old. When Jan, who came from the same egg as Alf, did a stomach massage at the age of 25, the grief over his brother came up.The parents were not told where the children who died wentI was also deeply moved by Ingrid, an lone twin who lost her sister at birth in 1950s Sweden. At that time, it was common for stillborn children to be buried in already open graves. The parents were not told where the children who died went. They were spoken of in silence, all pregnancies would end happily, it was seen as best that the parents had no relationship with the child.That interview made me search for Maria on a page of Swedish graves. For the first time, I saw her name in text, Maria Holmberg, Lund pastorate, which states the day of our birth but also the day she was buried. But the most healing thing is knowing that since the 1970s in Sweden, parents have been given the option to hold a funeral for children who die at birth. My parents got to say goodbye with a funeral where my sister lay in a small white coffin. Before that, she was also baptized and given her name, Maria.Grief can hide deep inside a twin who has lost their twin at birth, and it feels important to share that fact.It is so easy for those around them to only see twins who have lost their twin in their teenage years or as adults. But grief for someone with whom you shared the time of conception can leave deep traces. The bond is already created in the womb and is permanent. Important research that can explain the depth of such grief has been done by Italian psychotherapist Alessandra Piontelli. Through ultrasound, she has followed twins’ relationship with each other in the fetal stage. She has categorized their behavior in the womb as “the playing pair”, “the hostile pair” and “the hugging pair”. She has then shown that the relationship patterns they have with each other in the womb tend to continue even after birth.My interviews eventually became the Swedish book Leva för två, which was translated into the English book Living for two, published by Free Association Books in London. It was only after the book was finished that I got to see my mother’s birth record. Thanks to that, I had the opportunity to meet the midwife who took me and my sisters in. I found her name in the birth record and learned from a mutual acquaintance that she was now over eighty years old. Even though she, Anita, did not remember our birth, it was nice to meet her with my sister Sara and share all her experience. She had only been a practicing midwife for three years when she had her first triplets, and then worked in the profession for almost fifty more years.The more I open up and let Maria be a part of my world, the more powerful and strong I feelI can recommend that parents of lone twins allow the surviving twin to see the birth record when they are old enough. The record clearly showed me that my sister was in such a bad condition that life on earth was not preferable. I could read how the medical staff made several attempts to keep her alive, how she was met with professionalism, which has helped me accept that I did not get to share life here on earth with her. I also hope that my book Living for two will provide comfort in that I and other twins often feel a kind of connection with our twin siblings after a loss. It’s as if the bond is so strong that not even death can break it. I’ll tell you more about my experiences of meeting other lone twins in a future post.Nick EmbletonUntil I became an attending neonatologist, I would never have imagined this situation existed; that there may have been a triplet pregnancy but where only two babies survived. It’s possible that as a resident or fellow I met families like this; I don’t know or remember. I had always been aware that we struggled to know what to do, or say, on the NICU when there were twins, where only one baby survived, and that led to the Butterfly Project. Triplets are much less common, and the challenges faced when there are only two survivors are unique. Not better or worse than singleton loss, or twin loss. But unique and different. I have now met several triplet families like Pia, and their stories are profound and deeply moving.I don’t know what the “take home” message from Pia’s story is. There are probably many. But every time I listen to parents or siblings speak or write like this it continues to strike me what a privileged life I have led.Pia carried her grief for Maria for almost half-a-century before she realised it was there. I never thought that could happen. Of course, it’s possible to suppress ‘trauma’ for years, but Pia’s grief was and is different.I have heard health professionals caring for bereaved families say, or heard from the parents direct that they’ve been told, “you’ll get over it”. If there’s an important take away for health professionals from Pia, it is that the grief of baby loss lives with you forever.We never know what someone else is going through. We will never know what it feels like to be Pia, or Dani or the other siblings and parents who will write here in the Butterfly project. I hope it makes all of us behave and act with more kindness, compassion and gratitude.* * *This post is also published here on Substack.

dear sirI have a baby with a possible diagnosis of methemoglobinemia. What is the dose and duration of treatment with methylene blue? And after how many doses can signs of response to treatment be seen?

There's a scene that keeps coming back to me since I heard Professor Madalena Patrício, of the University of Lisbon's Faculty of Medicine, speak about artificial intelligence in health professions education. It isn't a scene from her talk, exactly — it's from Pluribus, a series I watched recently that, somehow, found a new way to be read in light of what she said. In the show, a signal from space carries the code of a virus that fuses almost all of humanity into a single consciousness. At first glance, what's frightening is the loss of individuality. But something more unsettling may be happening underneath. The infected don't become less intelligent — quite the opposite. They gain instant access to everything any of them has ever known, thought, or concluded. The answer stops being the end of a search. It starts existing before the question even has time to form.That's what makes the world of the show so strange. There's no longer any need to sit with uncertainty, hold a doubt, test a hypothesis, or change one's mind. Knowledge is simply always available, with no gap in which something has to be worked out by a person. Humanity doesn't lose the capacity to know — it slowly loses the need to think. Carol, the protagonist and one of the few people left immune, notices what everyone else doesn't. What looks like harmony to almost everyone feels to her like a slow death of reasoning: not because people have stopped knowing things, but because they no longer need to discover them. Even as everyone else seems to move in perfect sync, in a flawless community, watching it unfold is deeply unsettling.It's hard not to recognize, on a much smaller scale and without any alien virus, a version of this already showing up in my Uni. A student who opens a language model before even forming their own question isn't necessarily doing anything wrong. But they may be making a quiet trade — handing the machine exactly the part of the process that was supposed to help them build their own thinking. The tool offers speed, content, availability. The risk shows up when that ease starts replacing the work of figuring things out. Maybe the real question isn't whether a student uses AI. It's this: what still happens inside them before they accept an answer? Or, more to the point — what has stopped happening?That shift also moves the teacher's place. For a long time, teaching meant, in large part, transmitting what one already knew. The teacher was where knowledge accumulated; the student was the one meant to receive it. That model already had its own limits, but artificial intelligence introduces a break that's hard to ignore. On the surface, if information is just one prompt away, then simply transmitting that knowledge starts to make the teacher's presence in the room feel unnecessary. To follow that logic without question is to fall into a trap now being dressed up as a reason to embrace AI as a legitimate teaching tool. What gets left out of that reasoning is the very thing that justifies a teacher's existence in the first place: the job of teaching a student to think critically and independently.Madalena Patrício called this epistemological leadership. The phrase feels exact, because it moves the teacher away from simply knowing and toward helping someone else decide how to know — how to doubt, how to recognize the edges of what only seems like knowledge. In Pluribus, that kind of formation becomes pointless, because disagreement disappears entirely. And maybe that's exactly where the story offers its most powerful metaphor. Thinking requires some degree of friction. Evidence can contradict a conviction; a question can unsettle an answer; an experience can force someone to rebuild an entire structure of thought. Jack Mezirow placed that movement at the center of transformative learning: it's the dilemma that doesn't fit how we already understand the world that forces us to rethink it. Discomfort, in that sense, isn't a flaw in learning — it's part of it. If knowledge is a path where what you gather along the way matters as much as where you end up, then skipping that path in the name of efficiency may cost someone not just a particular learning experience, but thinking itself as an essential tool of what makes us human.This is why Carol becomes something far more interesting than the standard rebel. She doesn't reject connection simply because anything new feels threatening — she insists on understanding before accepting. She holds onto her own opinion when the rest of the world no longer sees the point of having one; she keeps questioning when everyone else has forgotten what the verb to ask even means. Carol's resistance isn't a rejection of intelligence. It's a refusal to give up her own consciousness — a way of keeping her humanity intact, even as almost everyone else has already lost theirs.There's something of that same posture in what Patrício proposes to institutions. Rather than turning AI into a matter of allegiance — for it or against it, adopt or resist — she talks instead about readiness. And readiness isn't a statement you make once; it's a practice you build. It means giving clear institutional direction, backing teachers and students, and holding onto ethical principles even when the pressure for productivity points toward shortcuts. An institution can fold AI into every one of its processes and still not be ready for it. Adoption happens fast. Educating a student doesn't. In health education, in particular, there are limits worth treating as non-negotiable. AI can suggest, but it can't answer for a clinical decision. It can organize and summarize evidence, but it can't take responsibility for acting on it. It can help simulate a case, but it can't reproduce an encounter with an actual sick person — their vulnerability, their contradictions, their silences, their history. The machine can take part in a judgment. It can't be the one who carries it. That's not just about AI's technical limits, which will keep shifting and improving — it's something more fundamental. In health training, there are decisions someone has to be accountable for. Teaching that is a different job from teaching someone to use a tool.Our job as teachers is to help students tell the difference between a conviction built from knowledge that's been critically examined, and an answer that arrived ready-made — produced by a mathematical language system that's very good at linking words, patterns, and probabilities, but that can never take on our responsibility for what we do with what it gives us. That's why the kind of training in Unis has to happen on two fronts. Teachers need to learn to use AI with real discernment — which has very little to do with mastering prompting techniques, and a lot to do with building the habit of interrogating an answer before passing it along. Students, meanwhile, need to learn to be suspicious of answers that sound too good: to spot the error, notice the gap, recognize the bias, and register when a text's confidence outruns the evidence behind it.Today, that skill starts to look like a basic form of literacy — not just finding a source, but knowing how to judge what a machine hands you as an answer. Rules can set limits. They can block some kinds of misuse. But they can't teach anyone to think. Only practice does that: testing the AI's answer against what you already know, hunting for what's missing, checking what seems right, sitting with the doubt a little longer. Learning to disagree with the machine may turn out to be one of the most important ways of learning from it. Without that habit, the gap between institutions will only widen. On one side, schools that teach students to think with AI, keeping it firmly in its place as a tool. On the other, schools that simply let it fill whatever space it finds. The inequality won't just be about access to technology — it will be about the quality of thinking each institution manages to cultivate around it. In Pluribus, the fusion erases that gap entirely. There's no longer one mind standing next to another. There's one mind, inside everyone. Maybe that's exactly the line a university needs to hold.Some competencies belong to both knowledge and practice: understanding where AI gets things right and where it doesn't, recognizing its biases, reasoning ethically about how it's used — but also communicating with a patient, deciding alongside them, working with other professionals, and holding space for everything that happens in care that no block of text could ever capture. Technology is going to keep advancing. The question was never how to stop it. The question is figuring out what shouldn't be allowed to advance into the space that belongs to the human being in the room. There's a part of health work that schools have always trained less, precisely because it resists being turned into something you can test: listening to someone, holding their vulnerability, noticing what went unsaid, building a decision together, staying present in the face of suffering. These skills are harder to measure. They are the most important ones.AI can change health and education in deep ways. But what that change ends up meaning still depends entirely on who's using it. Empathy, warmth, compassionate listening aren't humanist decorations bolted onto the technique once the "real work" is done — they are part of the care itself. And they're also exactly what technology, by definition, can't do for us. By the end of the talk, I was left with one simple idea the speaker had planted: a lecture only really matters if it changes something after it's over. What Patrício seemed to be asking of universities wasn't just that they teach knowledge about AI, but that they help form the judgment to deal with it — not just that they hand students a tool, but that they teach them to answer for how that tool gets used.That's when my mind drifted back to Carol. She never got to choose a world without that intelligence in it. She had to learn to live beside something that knew more than she did, without letting what it knew decide what she herself should think. That might be the most precise challenge for anyone teaching today: not to raise people immune to artificial intelligence — that would be impossible, and maybe not even desirable — nor to raise people willing to hand it the work of thinking. Maybe our role now is to shape people who can stand beside it, as questioners. People capable of asking before accepting, doubting when it matters, recognizing when an answer isn't enough. Capable, above all, of keeping some distance between what the machine knows and what we choose to do with what it knows.That distance isn't a flaw in the technology. It's the space where our responsibility lives. We need to start forming more Carols.Disclosure: I am not against AI, in fact, I used it to translate this text from the original one. You can read it below:A resposta que chega antes da perguntaHá uma cena que continua voltando à minha cabeça desde que ouvi a professora Madalena Patrício, da Faculdade de Medicina da Universidade de Lisboa, falar sobre inteligência artificial na formação de profissionais de saúde. Não é exatamente uma cena da palestra. É de Pluribus, série que assisti recentemente e que, de algum modo, encontrou ali uma nova chave de leitura. Na série, um sinal vindo do espaço carrega o código de um vírus capaz de fundir quase toda a humanidade em uma única consciência. À primeira vista, o que assusta é a perda da individualidade. Mas talvez haja algo mais inquietante acontecendo. Os infectados não ficam menos inteligentes. Ao contrário: passam a ter acesso imediato a tudo o que qualquer um deles já soube, pensou ou concluiu. A resposta deixa de ser o fim de uma busca. Ela passa a existir antes mesmo de a pergunta amadurecer.É isso que torna aquele mundo tão estranho. Não há mais necessidade de atravessar a incerteza, sustentar uma dúvida, testar uma hipótese, mudar de ideia. O conhecimento está sempre disponível, sem o intervalo em que alguma coisa precisa ser elaborada por uma pessoa. A humanidade não perde a capacidade de saber; perde, aos poucos, a necessidade de pensar. A protagonista Carol, uma das poucas pessoas imunes, percebe o que os demais não percebem. O que parece harmonia para quase todos lhe parece uma espécie de morte lenta do raciocínio. Não porque as pessoas tenham deixado de saber, mas porque já não precisam descobrir. Mesmo que agora todos pareçam estar agindo em sintonia, numa comunidade perfeita, o efeito da nova realidade causa muito desconforto em quem assiste.É difícil não reconhecer, em escala muito menor e sem qualquer vírus extraterrestre, uma possibilidade que começa a aparecer na sala de aula. O estudante que abre um modelo de linguagem antes de formular a própria pergunta talvez não esteja fazendo nada de errado. Mas pode estar fazendo uma troca silenciosa: entrega à máquina justamente a parte do processo que deveria ajudá-lo a formar seu próprio pensamento. A ferramenta oferece velocidade, amplitude, disponibilidade. O risco aparece quando essa facilidade começa a substituir a elaboração. Talvez a questão, portanto, não seja saber se um estudante usa IA. A questão é outra: o que ainda acontece dentro dele antes de aceitar a resposta? Ou, ainda mais importante, o que está deixando de acontecer?Essa mudança desloca também o lugar do professor. Durante muito tempo, ensinar significou, em grande medida, transmitir aquilo que se sabia. O professor era o repositório onde o conhecimento estava reunido; o estudante, aquele que deveria recebê-lo. Esse modelo já carregava suas próprias limitações, mas a inteligência artificial introduz uma ruptura difícil de ignorar. Em um primeiro momento, se a informação está a apenas um prompt de distância, então a simples transmissão desse conhecimento torna a presença do professor na sala de aula desnecessária. Seguir por este caminho é aceitar sem questionar a armadilha que vem no sendo apresentada como razão para endossar a IA como ferramenta legítima de ensino.  Fica de fora desse raciocínio o mais importante e que justifica em essência a própria existência da figura do professor: a necessidade de ensinar o aluno a pensar de forma critica e autônoma.Madalena Patrício chamou isso de liderança epistemológica. A expressão parece precisa porque desloca o professor da posição de quem simplesmente sabe para a de quem ajuda outra pessoa a decidir como saber, como duvidar e como reconhecer os limites daquilo que parece saber. Em Pluribus, essa formação se torna desnecessária porque a discordância desaparece. E talvez seja justamente aí que a história ofereça sua metáfora mais poderosa. Pensar exige algum grau de atrito. Uma evidência pode contrariar uma convicção; uma pergunta pode desorganizar uma resposta; uma experiência pode obrigar alguém a rever uma estrutura inteira de pensamento. Jack Mezirow colocou esse movimento no centro da aprendizagem transformadora: é o dilema que não cabe no modo como já compreendemos o mundo que nos obriga a reconstruí-lo. O desconforto, nesse caso, não é um defeito do aprendizado. É parte dele. Se o conhecimento é uma caminhada onde o que se colhe pelo caminho é tão ou mais importante do que o destino final, eliminar essa trajetória em nome da eficiência, pode significar também privar alguém não apenas de uma determinada experiência de aprendizado, mas também do próprio “pensar” como ferramenta essencial na constituição da nossa humanidade.É por esta razao que a Carol transcende a simples figura da rebelde para se tornar algo imensamente mais interessante. Ela não rejeita a conexão porque tudo o que é novo lhe parece ameaçador. Ela exige compreender antes de aceitar. Manter uma opinião própria quando o restante do mundo já não vê motivo para ter uma; continuar questionando quando todos os demais se esqueceram do significado do verbo perguntar. A rebeldia da Carol não representa uma negação à inteligência, ela expressa o desejo de não renunciar à própria consciência e assim permanecer com a sua humanidade intacta (mesmo que na série a maior parte das pessoas já tenha perdido).Há algo dessa atitude naquilo que Patrício propõe às instituições de ensino. Em vez de transformar a IA em uma questão de adesão — ser a favor ou contra, adotar ou resistir —, ela fala em prontidão. E prontidão não é simplesmente fazer uma declaração. É uma prática construída. Significa oferecer uma direção institucional clara, apoiando professores e estudantes e sustentando princípios éticos mesmo que a ânsia por produtividade sugira seguir por atalhos. Uma instituição pode incorporar inteligência artificial a todos os seus processos e, ainda assim, não estar preparada para ela. A adoção acontece depressa. A formação do aluno, leva tempo.  Quando falamos em educação em saúde, existem limites que devemos considerar inegociáveis. A IA pode sugerir, mas não pode responder por uma decisão clínica. Pode organizar e resumir evidências, mas não assumir a responsabilidade de quem decide agir a partir delas. Pode ajudar a simular um caso, mas não reproduzir o encontro com uma pessoa doente — com sua vulnerabilidade, suas ambiguidades, seu silêncio, sua história. A máquina pode participar do julgamento. Não pode sustentar o julgamento. Isso não depende apenas das limitações técnicas da IA, que certamente continuarão mudando e evoluindo. Depende de algo mais fundamental: na formação em saúde, existem decisões pelas quais alguém precisa responder. Ensinar isso é diferente de ensinar a usar uma ferramenta.Nosso papel como docentes é o de ensinar a reconhecer a diferença entre uma convicção construída a partir de conhecimento examinado criticamente e uma resposta que chegou pronta, produzida por um sistema de linguagem matemático, capaz de associar palavras, padrões e probabilidades, mas que não pode assumir por nós a responsabilidade pelo que fazemos com elas. Por isso, a formação precisa acontecer em duas frentes. Os professores precisam aprender a usar IA com discernimento, e isso está muito longe de apenas dominar técnicas de prompting. Trata-se de desenvolver o hábito de interrogar a resposta antes de transmiti-la. Os estudantes, por sua vez, precisam aprender a desconfiar de respostas que parecem boas demais: identificar o erro, perceber a lacuna, reconhecer o viés, notar quando a segurança do texto excede a solidez da evidência.Hoje, essa competência se aproxima de uma forma básica de letramento: saber não apenas encontrar uma fonte, mas avaliar aquilo que uma máquina nos entrega como resposta. Regras podem estabelecer limites. Podem impedir alguns usos indevidos. Mas não ensinam ninguém a pensar. Isso só acontece quando existe prática: confrontar a resposta da IA com aquilo que se sabe, procurar o que falta, testar o que parece certo, sustentar a dúvida por mais alguns minutos. Aprender a discordar da máquina talvez se torne uma das formas mais importantes de aprender com ela. Sem esse exercício, a diferença entre instituições tende a se aprofundar. De um lado, estarão aquelas que ensinam estudantes a pensar com a IA, mantendo-a no lugar de ferramenta. De outro, aquelas que simplesmente permitem que ela ocupe os espaços que encontrar. A desigualdade não estará apenas no acesso à tecnologia, mas na qualidade do pensamento que cada instituição terá conseguido formar em torno dela. Em Pluribus, a fusão elimina essa distância. Não existe mais um pensamento ao lado de outro. Existe um pensamento dentro de todos. Talvez seja essa a fronteira que uma universidade precise preservar.Existem competências que pertencem tanto ao conhecimento quanto à prática: compreender onde a IA acerta e erra, reconhecer seus vieses, raciocinar eticamente sobre seu uso; mas também comunicar-se com o paciente, decidir com ele, trabalhar com outros profissionais e sustentar aquilo que acontece no cuidado e que nenhuma resposta textual consegue conter. A tecnologia continuará avançando. A questão nunca foi impedir esse avanço. A questão é saber o que não pode avançar para dentro do lugar que pertence ao humano. Há uma parte do trabalho em saúde que as escolas historicamente treinam menos justamente porque é difícil de transformar em prova: escutar alguém, acolher sua vulnerabilidade, perceber o que não foi dito, construir uma decisão compartilhada, estar presente diante do sofrimento. São competências menos fáceis de medir, mas são as mais importantes.A IA pode mudar profundamente a prática em educação e saúde. O sentido dessa mudança, porém, continua dependendo de quem a utiliza. Empatia, acolhimento, escuta compassiva não são adornos humanistas acrescentados à técnica depois que o essencial foi resolvido. São parte do próprio cuidado. E são também aquilo que a tecnologia, por definição, não pode fazer por nós. Ao final da palestra, fiquei com uma ideia simples provocada pela palestrante: uma palestra só importa quando modifica alguma coisa depois que termina. O que Patrício parecia pedir às universidades não era apenas que ensinassem conhecimento sobre IA, mas que formassem critérios para lidar com ela; não apenas que oferecessem uma ferramenta, mas que ensinassem a responder pelo modo como essa ferramenta é usada.Foi então que meu pensamento voltou novamente em Carol. Ela não teve a possibilidade de escolher um mundo sem aquela inteligência. Precisou aprender a viver ao lado de algo que sabia mais do que ela sem permitir que esse conhecimento determinasse o que ela própria deveria pensar. Talvez seja esse o desafio mais preciso para quem ensina hoje. Não formar pessoas imunes à inteligência artificial — isso seria impossível e talvez nem desejável. Tampouco formar pessoas dispostas a entregar a ela o trabalho de pensar. Talvez o nosso papel agora seja o de formar pessoas capazes de permanecer ao lado dela, como questionadores. Capazes de perguntar antes de aceitar, de duvidar quando necessário, de reconhecer quando uma resposta não basta. Capazes, sobretudo, de conservar alguma distância entre aquilo que a máquina sabe e aquilo que nós decidimos fazer com o que ela sabe.Essa distância não é uma deficiência da tecnologia. É o espaço da nossa responsabilidade. Precisamos formar mais Carols.