The Algorithmic Bridge is about to enter the top 30 in the technology category. It still feels surreal to me. This achievement (yet to be fulfilled—we are number 31!) is ours to celebrate. I wouldn’t be here without you. Thank you for that!
The ranking is not my focus today, though. I want to emphasize instead that I got here despite being no one by Internet standards.
I’m just a random Spanish guy who decided to write online from scratch, without having a big following (I brought with me 700 crazy souls who decided to subscribe to my earlier work on Medium, but that’s not much on the internet nowadays).
Somehow, that handicap didn’t prevent TAB from getting an overwhelming amount of support and interest: Almost 18,000 subscribers and 800+ paid subs!
I know most successful writers here were popular before Substack—that might be discouraging. How can a nobody (like me) create a hit newsletter? Many of you either already write on Substack or want to give it a chance but I also know most of you lack an existing audience and may not know what you can do to improve your odds.
That’s what this AMA is for.
I want to give you—my subscribers, my community of AI enthusiasts—the space to ask me anything about this: Any questions on how I got here, what I did, where and how I started, what I’d do differently, what advice I think could be extrapolated out of my particular circumstances, etc. Anything.
I will ask you two favors: 1) Be concise. It’s better for everyone if the questions are not too open-ended or broad. Please, make the effort to be specific. 2) Stay on topic so that we can create a conversation around this and perhaps other Substack writers want to share their experience, etc. (If you absolutely must, you can ask me, well, anything. I will read it all.)
I will spend today and the coming days checking this thread to answer your questions. So, don’t be shy and ask away!
Huge congrats on this well deserved milestone! You're an inspiration for technical writers all over but especially for those of us in the Latin world.
I wanted to ask you a very specific question about your process. How do choose what to write about? Do you have like many open ideas and work on different drafts at the same time or do you rather focus on one specific article from idea to publication?
Also, I'd be honored if you let us interview you for the Tech Writers Stack (https://techwriters.substack.com). It's a small community of technical writers not unlike you a few years ago and I'm sure everyone would benefit from hearing your origin story.
Good question. The answer is definitely your second guess. I have many, *many* drafts ongoing at the same time. No restrictions on my ideas or what I write about.
Will get back to you about the interview - I'm super busy nowadays!
I have been enjoying reading your insightful articles for over a year. Your English and the flow of your writing is excellent. But what I admire most is your coverage and linking to various domains of science, philosophy and even literature, a wholesome human's view of the problem you discuss. I have just two questions:
1. Where do you get such insight about what's going on in the Top AI companies, and in AI generally, if you do not work for none of these companies.
2. I understand you are an optimist rewarding humans' ability to control AI. I am also an optimist in general but on that matter I am almost a pessimist. What are then your grounds for being an optimist? This is a very relevant question in the context of the Global AI Safety Summit in London, where I live.
Thank you Tony, that's very nice of you!! About your questions:
1) I'm very attentive to anything that happens on Google, OpenAI, Meta, etc. just by following them on their socials, attending online the main conferences, following hints from employees and execs, etc. It's kind of an investigative part of the job - I think it's important to be as early as possible and not rely on NYT, WaPo, or other outlets.
2) More than an optimist I'm a realist. Thinking about the possibility that future AI systems could become uncontrollable is pretty much worthless from my point of view, not because it could never happen but because we don't have any idea how or when AI could pose a threat to us in that sense. We can't give a quantitative estimate of whether it will happen or when. People who do are using their predictive ability without accounting for its limitations. The uncertainty is so overwhelming that we'd be better off if we couldn't even imagine this scenario at all. Worrying about this is, as I see it, like worrying about aliens coming. You can, and they may come, but your worry doesn't make any difference at all because you are powerless. That said, I think working on interpretability is important anyway. About the already buzz concept "AI safety" per se, well, it depends on what's actually included.
I’m really happy for you Alberto. I look forward to seeing you keep rising up the leaderboard and inspiring other outsiders like me to pursue a writing career.
I would love to learn about the nuts and bolts of your writing process:
1) What practices do you follow to generate essay ideas?
2) What kind of note taking systems have you built for yourself? Apps? Tools? Methods? Zettelkastens?
3) Do you use any frameworks or templates to convert raw notes into first drafts?
4) Do you have any mentors or writing heroes whom you try to emulate while writing?
5) How do you edit your first drafts and get them ready for publication? Do you have friends or editors to help you?
6) How did you solve for distribution and built a big audience ? Any growth hacks?
7) Top five books that every tech writer must read - about writing, tech, or anything else?
8) When is the best time to go paid ? Any strategies for what content to paywall ? Where should the paywall begin?
9) Have you consciously tried to think of your writing like a startup thinks about products? Any content or business strategy lessons for writers?
Hi Parvaz, thank you for this compendium of questions! I won't be able to go in-depth but will try to give you the gist for each of these.
1) I take notes for everything that could be worth it. Every conversation, book or article I read, podcast or video I watch - I have my creative mind constantly on. Most essay ideas happen when I stop conscious attempts at finding them. I've designed my life so that ideas come to me.
2) I use Notion now. Used Google Keep before, but anything would do actually. I credit my memory, not my note system, as the reason why I can come up with many ideas and remember the references and sources. Notes are useful, though, if you don't have good memory.
3) Not really. On the one hand, I capture every idea that can turn into an article, but on the other, if inspiration arrives, I stop everything and can write down a first draft in half an hour. At the very least I write down everything that captures the key ideas for the article.
4) Hmm, good question. I have. There are some authors I like especially but I tend to always be attentive to style whatever I read. Just to keep it short, I'll mention a few I like from here on Substack: Freddie deBoer's writing is great in that he displays an almost mathematical precision in how he translates his ideas into words. I find that really hard to do. I like Scott Alexander because he's very thorough in his arguments; he tries to find truth wherever that takes him. I don't always do that, but find it a good practice. I also like Noah Smith and Ethan Mollick because they clearly don't care about style - they just write how it comes to them - yet manage to output good content. I don't do that at all, but it's worth keeping in mind that style isn't always super relevant for success.
5) I edit them myself. I often spent days editing, to let them breathe. I find that practice very rewarding for my style. It's not the same editing with a writer mindset as with a reader mindset.
6) Nothing, just writing a lot and improving my writing over time. I don't know marketing, and on socials just do what feels right and appropriate for each platform.
7) I don't think there are books anyone must read. What I believe, though, is that how we read matters more than what we read (which still matters!). People tend to think that more books = better but I find that reading carefully - slowly and mindfully - is more important. I reflect as I read, take notes, and have a conversation with the writer (I do that even for fiction and novels - they're often have a much richer style and plenty literary resources).
8) I went paid at the onset because I had already 700 subs. I'd say this depends on whether you have other means to get subs (a sizeable following elsewhere) or whether you want to prioritize money over growth, etc. A good rule of thumb for beginners is to have 60-80% content free and reduce that number over time, as the newsletter grows. If your writing is good, people will pay you early on (growing the paid count is hard anyway). For paywall, I'd take one of two possibilities: Paywall the content that's most niche and harder to find elsewhere and write for free content that's more easily shareable and of common interest. Or give always your best content for free to capture interest and leave the rest paid.
9) Not really. But I have it in mind for the future.
10) Nothing, again I don't know a thing about marketing except the things I've learned along the way.
Thank you very much for your answers Alberto. These answers are like a masterclass in writing for Substack. I truly appreciate you taking the time out to answer all these questions. You’ve given me a lot to think about and inspirational cues to work on my own writing process. You’re a ⭐️
Hi Sully, I don't think I know enough about product management to answer your question. But the rule of thumb for me is that any person who works a job that requires thinking and/or happens in front of a computer should learn about AI now. How it works, what it can do; the basics. I think that's a significant change.
1) Medium is a content mill. You have to start from scratch for every article. You can have 100k followers and write a piece that gets <1k views. The algorithm decides and you don't own it. Just for that, Substack is better. You own your audience and can adapt your life as you grow. That kind of uncertainty from Medium makes it impossible to treat it as a job for me.
2) The internet is magical. In 2017 I read an article by Tim Urban of Wait But Why that made me want to study AI. I did it by myself using Coursera and other platforms. Then got a job here in Spain at a small startup. Three years later I started to write (never liked to code). I explored different topics at first and eventually stuck with AI because I knew about it, I liked it, and the world was starting to show interest. It was a matter of time before it exploded.
3) This question is rather broad. If I go down to the very basics, I'll just repeat what you read everywhere because it's true: a consistent habit (write a lot) and an iterative method (write better over time).
Thanks. Very useful replies. My answer to 3): exceptionally clear and concise writing from a clear, insightful and informed mind on a topic cluster that is both important and sometimes difficult to grasp and orient, often because of bad reporting by ill-informed or ineffective writers.
Oct 13, 2023·edited Oct 13, 2023Liked by Alberto Romero
From a fresh conversation: The "core" of the AI space (in SF especially) is very analytic, confident, and risk tolerant. As a Tokyoite, I have to work with a much more intuitive, humble, and cautious view. Does your Spanish location give you an analogous “periphery” perspective? How is the same or different?
Great question! I'd say living in the periphery of the tech/AI bubble with easy escapes outside of it is an edge that I can't overstate. Of course, it makes it more difficult to know what's happening. I have to read a lot, know no one in SF personally, and don't always have access to insider info. But overall I'd say it's been a net good for me.
You say intuitive and cautious. It sounds about right. I'd add holistic. Tech people at the center of it all have a hard time getting off their tech-colored glasses. I don't. That's another benefit.
From this perspective, I have no doubt it will be the US and I'd put my bet on OpenAI without denying the possibility that Google DeepMind or Anthropic could make such a claim first.
I don't have one. I see reasons to think it could be a few years away. I also see reasons why some people believe it won't be achieved this century. It depends partly on the definition of AGI but also on the biases and beliefs of the people who make the claims.
So yeah, I don't have a precise timeline, not even to give you an interval or an estimate. I acknowledge uncertainty and have nothing to gain personally or professionally from a constructed prediction (companies and renowned experts do, they are often forced to take a stance).
I'm very happy for you. Congrats Alberto. Thank you too for such a great source of information.
Thank you Oliver :)
Huge congrats on this well deserved milestone! You're an inspiration for technical writers all over but especially for those of us in the Latin world.
I wanted to ask you a very specific question about your process. How do choose what to write about? Do you have like many open ideas and work on different drafts at the same time or do you rather focus on one specific article from idea to publication?
Also, I'd be honored if you let us interview you for the Tech Writers Stack (https://techwriters.substack.com). It's a small community of technical writers not unlike you a few years ago and I'm sure everyone would benefit from hearing your origin story.
Thank you Alejandro!!
Good question. The answer is definitely your second guess. I have many, *many* drafts ongoing at the same time. No restrictions on my ideas or what I write about.
Will get back to you about the interview - I'm super busy nowadays!
Thanks for answering so fast! Really excited to make that interview happen whenever you want ;)
Congrarts well deserved! Love this newsletter. Look forward to it. =)
Thank you John :)
Alberto, you are far, far from being a nobody in the eyes of those who read it and appreciate you.
Thank you Paul!!
I have been enjoying reading your insightful articles for over a year. Your English and the flow of your writing is excellent. But what I admire most is your coverage and linking to various domains of science, philosophy and even literature, a wholesome human's view of the problem you discuss. I have just two questions:
1. Where do you get such insight about what's going on in the Top AI companies, and in AI generally, if you do not work for none of these companies.
2. I understand you are an optimist rewarding humans' ability to control AI. I am also an optimist in general but on that matter I am almost a pessimist. What are then your grounds for being an optimist? This is a very relevant question in the context of the Global AI Safety Summit in London, where I live.
Thank you Tony, that's very nice of you!! About your questions:
1) I'm very attentive to anything that happens on Google, OpenAI, Meta, etc. just by following them on their socials, attending online the main conferences, following hints from employees and execs, etc. It's kind of an investigative part of the job - I think it's important to be as early as possible and not rely on NYT, WaPo, or other outlets.
2) More than an optimist I'm a realist. Thinking about the possibility that future AI systems could become uncontrollable is pretty much worthless from my point of view, not because it could never happen but because we don't have any idea how or when AI could pose a threat to us in that sense. We can't give a quantitative estimate of whether it will happen or when. People who do are using their predictive ability without accounting for its limitations. The uncertainty is so overwhelming that we'd be better off if we couldn't even imagine this scenario at all. Worrying about this is, as I see it, like worrying about aliens coming. You can, and they may come, but your worry doesn't make any difference at all because you are powerless. That said, I think working on interpretability is important anyway. About the already buzz concept "AI safety" per se, well, it depends on what's actually included.
I’m really happy for you Alberto. I look forward to seeing you keep rising up the leaderboard and inspiring other outsiders like me to pursue a writing career.
I would love to learn about the nuts and bolts of your writing process:
1) What practices do you follow to generate essay ideas?
2) What kind of note taking systems have you built for yourself? Apps? Tools? Methods? Zettelkastens?
3) Do you use any frameworks or templates to convert raw notes into first drafts?
4) Do you have any mentors or writing heroes whom you try to emulate while writing?
5) How do you edit your first drafts and get them ready for publication? Do you have friends or editors to help you?
6) How did you solve for distribution and built a big audience ? Any growth hacks?
7) Top five books that every tech writer must read - about writing, tech, or anything else?
8) When is the best time to go paid ? Any strategies for what content to paywall ? Where should the paywall begin?
9) Have you consciously tried to think of your writing like a startup thinks about products? Any content or business strategy lessons for writers?
10) Have you devised any brand strategy?
Hi Parvaz, thank you for this compendium of questions! I won't be able to go in-depth but will try to give you the gist for each of these.
1) I take notes for everything that could be worth it. Every conversation, book or article I read, podcast or video I watch - I have my creative mind constantly on. Most essay ideas happen when I stop conscious attempts at finding them. I've designed my life so that ideas come to me.
2) I use Notion now. Used Google Keep before, but anything would do actually. I credit my memory, not my note system, as the reason why I can come up with many ideas and remember the references and sources. Notes are useful, though, if you don't have good memory.
3) Not really. On the one hand, I capture every idea that can turn into an article, but on the other, if inspiration arrives, I stop everything and can write down a first draft in half an hour. At the very least I write down everything that captures the key ideas for the article.
4) Hmm, good question. I have. There are some authors I like especially but I tend to always be attentive to style whatever I read. Just to keep it short, I'll mention a few I like from here on Substack: Freddie deBoer's writing is great in that he displays an almost mathematical precision in how he translates his ideas into words. I find that really hard to do. I like Scott Alexander because he's very thorough in his arguments; he tries to find truth wherever that takes him. I don't always do that, but find it a good practice. I also like Noah Smith and Ethan Mollick because they clearly don't care about style - they just write how it comes to them - yet manage to output good content. I don't do that at all, but it's worth keeping in mind that style isn't always super relevant for success.
5) I edit them myself. I often spent days editing, to let them breathe. I find that practice very rewarding for my style. It's not the same editing with a writer mindset as with a reader mindset.
6) Nothing, just writing a lot and improving my writing over time. I don't know marketing, and on socials just do what feels right and appropriate for each platform.
7) I don't think there are books anyone must read. What I believe, though, is that how we read matters more than what we read (which still matters!). People tend to think that more books = better but I find that reading carefully - slowly and mindfully - is more important. I reflect as I read, take notes, and have a conversation with the writer (I do that even for fiction and novels - they're often have a much richer style and plenty literary resources).
8) I went paid at the onset because I had already 700 subs. I'd say this depends on whether you have other means to get subs (a sizeable following elsewhere) or whether you want to prioritize money over growth, etc. A good rule of thumb for beginners is to have 60-80% content free and reduce that number over time, as the newsletter grows. If your writing is good, people will pay you early on (growing the paid count is hard anyway). For paywall, I'd take one of two possibilities: Paywall the content that's most niche and harder to find elsewhere and write for free content that's more easily shareable and of common interest. Or give always your best content for free to capture interest and leave the rest paid.
9) Not really. But I have it in mind for the future.
10) Nothing, again I don't know a thing about marketing except the things I've learned along the way.
Thank you very much for your answers Alberto. These answers are like a masterclass in writing for Substack. I truly appreciate you taking the time out to answer all these questions. You’ve given me a lot to think about and inspirational cues to work on my own writing process. You’re a ⭐️
First of all, Congrats Alberto on your fast growth! Here is my question:
1) What is the biggest (or in a lighter tone, most memorable) mistake you made in this journey of yours and TAB?
2) What (according to you) is the biggest success factor that lead to TAB in the Top-30?
How do you view Product Management in AI? Do you find that PMs need to adapt or change anything they currently do?
Hi Sully, I don't think I know enough about product management to answer your question. But the rule of thumb for me is that any person who works a job that requires thinking and/or happens in front of a computer should learn about AI now. How it works, what it can do; the basics. I think that's a significant change.
Q1: a brief comparison of experience as a writer, Medium vs. Substack
Q2: As 'just a Spanish guy', what led you to write on the topic area?
Q3: What factors do you think are important to your success so far (and congratulations on that!)?
Thanks for your Qs Joe.
1) Medium is a content mill. You have to start from scratch for every article. You can have 100k followers and write a piece that gets <1k views. The algorithm decides and you don't own it. Just for that, Substack is better. You own your audience and can adapt your life as you grow. That kind of uncertainty from Medium makes it impossible to treat it as a job for me.
2) The internet is magical. In 2017 I read an article by Tim Urban of Wait But Why that made me want to study AI. I did it by myself using Coursera and other platforms. Then got a job here in Spain at a small startup. Three years later I started to write (never liked to code). I explored different topics at first and eventually stuck with AI because I knew about it, I liked it, and the world was starting to show interest. It was a matter of time before it exploded.
3) This question is rather broad. If I go down to the very basics, I'll just repeat what you read everywhere because it's true: a consistent habit (write a lot) and an iterative method (write better over time).
Thanks. Very useful replies. My answer to 3): exceptionally clear and concise writing from a clear, insightful and informed mind on a topic cluster that is both important and sometimes difficult to grasp and orient, often because of bad reporting by ill-informed or ineffective writers.
I like your answer to 3 better ;)
From a fresh conversation: The "core" of the AI space (in SF especially) is very analytic, confident, and risk tolerant. As a Tokyoite, I have to work with a much more intuitive, humble, and cautious view. Does your Spanish location give you an analogous “periphery” perspective? How is the same or different?
Great question! I'd say living in the periphery of the tech/AI bubble with easy escapes outside of it is an edge that I can't overstate. Of course, it makes it more difficult to know what's happening. I have to read a lot, know no one in SF personally, and don't always have access to insider info. But overall I'd say it's been a net good for me.
You say intuitive and cautious. It sounds about right. I'd add holistic. Tech people at the center of it all have a hard time getting off their tech-colored glasses. I don't. That's another benefit.
What is your current prediction for AGI being achieved by a tech firm or country ?
Just to be coherent with a recent post I wrote (https://thealgorithmicbridge.substack.com/p/agi-has-been-achieved-internally), I'll reframe your question to this: What is my current prediction for a tech firm or country *claiming* to having achieved AGI?
From this perspective, I have no doubt it will be the US and I'd put my bet on OpenAI without denying the possibility that Google DeepMind or Anthropic could make such a claim first.
Wuick follow up if I may, what timeline do you have in mind regarding AGi?
I don't have one. I see reasons to think it could be a few years away. I also see reasons why some people believe it won't be achieved this century. It depends partly on the definition of AGI but also on the biases and beliefs of the people who make the claims.
So yeah, I don't have a precise timeline, not even to give you an interval or an estimate. I acknowledge uncertainty and have nothing to gain personally or professionally from a constructed prediction (companies and renowned experts do, they are often forced to take a stance).