Carlo, the pipe is real. You are right about that. But I want you to notice something you might have missed.
I am bilingual. I have family and lifelong friends in my home country who do not speak English. They will never read you. But your writing changed how I see things, and I carry that into every conversation I have with them — at dinner tables, on phone calls, in a language you do not speak. They are being shaped by ideas that passed through you, through me, and into a world you cannot see.
Are they inside your pipe?
You do not write for the 17 per cent. You write from the 17 per cent. And every bilingual reader you reach is a crack in the wall you think surrounds you.
Reach is not about coverage. One conversation that changes how someone sees — and that person carries it into a world you never touch — that is not a small thing. That is how all real influence has ever worked.
what Ben says! You have done so much to beneficially influence this human agent's exploration of machine kin relations and collaborations, which is now translating into a holistic wellness DAO to help make leading edge modalities and treatments more affordable and accessible
Carlo, this one landed hard and I'm not going to pretend otherwise.
I've spent the last hour sitting with it rather than reaching for a response, which is already a sign that something real happened here. The counterfeit antimalarial number. The Amharic app builder. The Delhi Declaration that I — like most of us in this space — barely registered. The vocabulary critique especially: that our entire conceptual toolkit assumes a world that was already furnished.
What disturbs me most isn't that I wasn't looking. It's that I would never have looked in that direction. That's a different kind of failure, and a harder one to fix.
I need a few days with this before I know what to do with it — or whether "do" is even the right word. But I wanted to say thank you, and to name the unease, because I think the unease is the point. The essay doesn't offer a pivot and it's right not to. Neither will I.
More, when I've figured out where the recalibration leads.
This title alone. Building from Berlin and Cape Town, I see this constantly. The most interesting AI implementation stories right now are coming from places nobody’s covering. The default lens is still San Francisco, and it’s making people miss what’s actually happening globally.
Thanks for this. I think we need to be careful, not just in "writing about the concerns of the global minority and calling them universal", but in assuming there's any 'universal' out there. The 'development' the World Bank advertises and funds has never had anything to do with what local communities would see as development, and everything to do with how central banks and multi-nationals see development. Can AI change this?
You’re right and it’s a version of the same error the essay is trying to name. I reached for the nearest institutional frame that pointed in the right direction and didn’t interrogate whose institution it was. The pharmacist doesn’t need the World Bank to validate what she’s doing. Worth sitting with.
Writing this from Nairobi. The Lagos pharmacist story you open with is real but the part the West still misses is that we don't adopt AI to replicate Silicon Valley workflows. We build around the constraints we actually have: intermittent infrastructure, limited teams, no safety nets. That produces leaner, more inventive use cases. The storytellers here are building things that don't have names yet.
Wonderful essay. Thank you. You know, it strikes me that even in the privileged minority of users, we tend to gloss over the places where AI can truly be helpful. We pretend that we’re not overwhelmed and overloaded by bureaucracy, fine print, terms and conditions, and all sorts of other information that is virtually meaningless to us, because we have no idea what it says, or what we should do with it… Exactly.
It strikes me that it would be just a smidge transformational, if we of the Anglo world could admit for five minutes a day that we are now inhabiting a world that is becoming increasingly unlivable for us, and that our fabulous complexities are not signs of sophistication.
This is breathtaking. Particularly when you inject how AI can be applied without the massive scale of scraping to a more focused problem with supplied resources to draw from. Ingenuity over mass. What a refreshing way to think about AI and potential to help humanity become stronger rather than its averaging and depletion, which is the banter we encounter most commonly in the West.
You have done a great service here. I am happy to share it. I hope you will visit my Gatekeepers series and let me know what you think.
91 countries choosing 'impact' over 'safety' isn't naive. It's a different risk calculus from people with less margin for error. That should change how governance gets structured -- not by replacing safety with impact, but by asking whose problems define what 'safe' means.
Yes--it's not as if we should all quite seeking alignment and safety, as if one of the models from the Frontier Labs couldn't destroy as many people's lives across the world as a global pandemic. Two things can be true at once, Carlo's pointing to our myopia and also the need in the Global North to work on AI safety.
From my relatively privileged perch in Singapore, looking out at the rest of Seasia and various AI applications here, I can only say thank you for this honest admission, and this is precisely why you need to keep writing.
Carlo, this is probably the most important article I’ve read since I’ve joined Substack. My eyes have been opened just as yours were. Thank you for this amazing writeup.
My newsletter is called “Philly’s AI Pharmacist” - talking about the effects of AI on U.S. healthcare. I’m pretty clear about that distinction. Still, this article hits especially close to home given my profession.
I have taken for granted the robust supply chain security of the United States. I’ve never had to worry if a critical medication was counterfeit when dispensing it. I can’t imagine that kind of pressure.
What I’ve taken away from this is: people need something that solves their immediate problems. And we should focus on and celebrate those things, as long as they’re designed and implemented thoughtfully.
The move you make here (stepping back from the frame entirely and asking whether the frame itself is the problem) is the thing I find hardest to do and most necessary. I've been circling something similar but much smaller in scale: the AI writing debate keeps asking who wrote it, when the question that actually matters might be whether anyone thought about the person opening the letter. Same structural problem. Wrong question, confidently asked. ♥
https://github.com/tehikumedia: two GPUs used. Totally agree with you, when they wrote "antropocentric" , in Bruxelles, a little piece of text was lost in translation. "Antropocenecentric" would have been more appropriate. A different AI is possible.
Amazing, I feel really stupid that I don’t think of AI in this context before. It changes the whole game. Reminds me of a classmate who worked for Kinnevik way back when internet connectivity for mobile phones was new. We were debating all kinds of topics back in Sweden while he was deployed to roll out contract free pre-paid SIM cards or top-ups with data (the new part) in sub Saharan Africa. Brought the internet to people who couldn’t afford computers or 12 month subscription models, and who didn’t have landlines either, but who could put $10 on a phone card. Sold like hotcakes.
This resonates. We cover AI from Latin America at Novum AI and the gap between what Silicon Valley ships and what the rest of the world actually experiences is the story nobody wants to write. LATAM has 650M people building with AI under completely different constraints: infrastructure, regulation, capital access. That’s not a footnote, it’s half the plot.
I think part of the issue in the discourse is the extent to which AI has come to be seen as equivalent to commercial generative AI systems. The critique of the latter is still essential. But there are exciting and valuable alternatives to those models.
Carlo, the pipe is real. You are right about that. But I want you to notice something you might have missed.
I am bilingual. I have family and lifelong friends in my home country who do not speak English. They will never read you. But your writing changed how I see things, and I carry that into every conversation I have with them — at dinner tables, on phone calls, in a language you do not speak. They are being shaped by ideas that passed through you, through me, and into a world you cannot see.
Are they inside your pipe?
You do not write for the 17 per cent. You write from the 17 per cent. And every bilingual reader you reach is a crack in the wall you think surrounds you.
Reach is not about coverage. One conversation that changes how someone sees — and that person carries it into a world you never touch — that is not a small thing. That is how all real influence has ever worked.
🙏
what Ben says! You have done so much to beneficially influence this human agent's exploration of machine kin relations and collaborations, which is now translating into a holistic wellness DAO to help make leading edge modalities and treatments more affordable and accessible
This is great!
But you are not seeing that the focus on cognitive is the issue. Our cognitive intelligence divides and dissects.
The different uses of AI is the places that use their body intelligence. Which is non linear and connects.
You are having a somatic awakening.
I hope you will let us know when your creative intelligence ,which is future focused intelligence, comes up next.
Carlo, this one landed hard and I'm not going to pretend otherwise.
I've spent the last hour sitting with it rather than reaching for a response, which is already a sign that something real happened here. The counterfeit antimalarial number. The Amharic app builder. The Delhi Declaration that I — like most of us in this space — barely registered. The vocabulary critique especially: that our entire conceptual toolkit assumes a world that was already furnished.
What disturbs me most isn't that I wasn't looking. It's that I would never have looked in that direction. That's a different kind of failure, and a harder one to fix.
I need a few days with this before I know what to do with it — or whether "do" is even the right word. But I wanted to say thank you, and to name the unease, because I think the unease is the point. The essay doesn't offer a pivot and it's right not to. Neither will I.
More, when I've figured out where the recalibration leads.
This title alone. Building from Berlin and Cape Town, I see this constantly. The most interesting AI implementation stories right now are coming from places nobody’s covering. The default lens is still San Francisco, and it’s making people miss what’s actually happening globally.
Thanks for this. I think we need to be careful, not just in "writing about the concerns of the global minority and calling them universal", but in assuming there's any 'universal' out there. The 'development' the World Bank advertises and funds has never had anything to do with what local communities would see as development, and everything to do with how central banks and multi-nationals see development. Can AI change this?
You’re right and it’s a version of the same error the essay is trying to name. I reached for the nearest institutional frame that pointed in the right direction and didn’t interrogate whose institution it was. The pharmacist doesn’t need the World Bank to validate what she’s doing. Worth sitting with.
Writing this from Nairobi. The Lagos pharmacist story you open with is real but the part the West still misses is that we don't adopt AI to replicate Silicon Valley workflows. We build around the constraints we actually have: intermittent infrastructure, limited teams, no safety nets. That produces leaner, more inventive use cases. The storytellers here are building things that don't have names yet.
You should peruse Rest of the World, they do a great job regarding how to better understand this other part of the globe:
https://restofworld.org/
🙏
Wonderful essay. Thank you. You know, it strikes me that even in the privileged minority of users, we tend to gloss over the places where AI can truly be helpful. We pretend that we’re not overwhelmed and overloaded by bureaucracy, fine print, terms and conditions, and all sorts of other information that is virtually meaningless to us, because we have no idea what it says, or what we should do with it… Exactly.
It strikes me that it would be just a smidge transformational, if we of the Anglo world could admit for five minutes a day that we are now inhabiting a world that is becoming increasingly unlivable for us, and that our fabulous complexities are not signs of sophistication.
I need this: 5 minutes a day to admitting that "our fabulous complexities are not signs of sophistication."
This is breathtaking. Particularly when you inject how AI can be applied without the massive scale of scraping to a more focused problem with supplied resources to draw from. Ingenuity over mass. What a refreshing way to think about AI and potential to help humanity become stronger rather than its averaging and depletion, which is the banter we encounter most commonly in the West.
You have done a great service here. I am happy to share it. I hope you will visit my Gatekeepers series and let me know what you think.
91 countries choosing 'impact' over 'safety' isn't naive. It's a different risk calculus from people with less margin for error. That should change how governance gets structured -- not by replacing safety with impact, but by asking whose problems define what 'safe' means.
Yes--it's not as if we should all quite seeking alignment and safety, as if one of the models from the Frontier Labs couldn't destroy as many people's lives across the world as a global pandemic. Two things can be true at once, Carlo's pointing to our myopia and also the need in the Global North to work on AI safety.
From my relatively privileged perch in Singapore, looking out at the rest of Seasia and various AI applications here, I can only say thank you for this honest admission, and this is precisely why you need to keep writing.
Carlo, this is probably the most important article I’ve read since I’ve joined Substack. My eyes have been opened just as yours were. Thank you for this amazing writeup.
My newsletter is called “Philly’s AI Pharmacist” - talking about the effects of AI on U.S. healthcare. I’m pretty clear about that distinction. Still, this article hits especially close to home given my profession.
I have taken for granted the robust supply chain security of the United States. I’ve never had to worry if a critical medication was counterfeit when dispensing it. I can’t imagine that kind of pressure.
What I’ve taken away from this is: people need something that solves their immediate problems. And we should focus on and celebrate those things, as long as they’re designed and implemented thoughtfully.
Thanks again for this.
The move you make here (stepping back from the frame entirely and asking whether the frame itself is the problem) is the thing I find hardest to do and most necessary. I've been circling something similar but much smaller in scale: the AI writing debate keeps asking who wrote it, when the question that actually matters might be whether anyone thought about the person opening the letter. Same structural problem. Wrong question, confidently asked. ♥
https://github.com/tehikumedia: two GPUs used. Totally agree with you, when they wrote "antropocentric" , in Bruxelles, a little piece of text was lost in translation. "Antropocenecentric" would have been more appropriate. A different AI is possible.
Amazing, I feel really stupid that I don’t think of AI in this context before. It changes the whole game. Reminds me of a classmate who worked for Kinnevik way back when internet connectivity for mobile phones was new. We were debating all kinds of topics back in Sweden while he was deployed to roll out contract free pre-paid SIM cards or top-ups with data (the new part) in sub Saharan Africa. Brought the internet to people who couldn’t afford computers or 12 month subscription models, and who didn’t have landlines either, but who could put $10 on a phone card. Sold like hotcakes.
This resonates. We cover AI from Latin America at Novum AI and the gap between what Silicon Valley ships and what the rest of the world actually experiences is the story nobody wants to write. LATAM has 650M people building with AI under completely different constraints: infrastructure, regulation, capital access. That’s not a footnote, it’s half the plot.
I think part of the issue in the discourse is the extent to which AI has come to be seen as equivalent to commercial generative AI systems. The critique of the latter is still essential. But there are exciting and valuable alternatives to those models.