The Bridge Builder
He built a real-time sign language translator. Then he started a newsletter. The sign language project let machines understand people - and now the newsletter does the same thing, just with prose instead of code, and for 40,000 readers instead of a product demo.
Alberto Romero writes The Algorithmic Bridge from Madrid, Spain, every week, twice a week, with no ads, no sponsors, and no softening of positions. He is an analyst at CambrianAI Research by day, a Substack author by cadence, and by inclination something rarer: a former ML engineer who chose to write sentences instead of functions - and who still knows enough code to call out the industry when it lies.
The newsletter launched in June 2022 with around 700 subscribers. ChatGPT arrived five months later and made AI a kitchen-table topic. Romero was already there, waiting, having built his audience before the gold rush. By the two-year mark, he had 26,000 subscribers. He now sits above 40,000, ranked among Substack's top rising technology publications.
"I'm determined to bridge the gap between people who love AI indiscriminately and people who hate it just as indiscriminately."
- Alberto Romero, on why he named the newsletterThree Degrees of Separation from Hype
His educational path reads like a dare. First, aerospace engineering at Universidad Politecnica de Madrid - the discipline of making things go very fast in very unforgiving environments, where precision isn't a preference but a survival requirement. Then, a master's in Brain and Cognition at Universitat Pompeu Fabra in Barcelona - the study of how biological intelligence actually works underneath all the metaphors.
Between and after these, he spent years as an ML developer at Showleap, building a system that could translate sign language in real time. That is not a side project. That is years of work on a problem where the gap between "almost working" and "actually working" is the difference between communication and silence.
The consequence of this background is that he reads AI research differently. He doesn't see benchmarks; he sees the gap between what a benchmark measures and what a benchmark proves. He doesn't see "AGI is coming soon" announcements; he sees the specific architectural bets those announcements are making, and whether those bets are structurally sound.
"The AI bubble is not like past bubbles - it's not fueled solely by speculation about excess, but also by speculation about viability. AI is a fancy general-purpose technology, but its unreliability may trump infinite fanciness and generality."
- Alberto Romero, on the AI bubbleWhat "No Sponsors" Actually Costs
He runs the newsletter on reader subscriptions. No brand deals. No "this issue is brought to you by." This is not a minor stylistic choice - it changes everything about what he can write. He called OpenAI's structural problems ("You Have No Idea How Screwed OpenAI Is") years before they became polite dinner conversation. He argued Google DeepMind was winning while half the industry was still writing post-mortems about Google's ChatGPT panic. He wrote a piece titled "OpenAI's Dead" - not as trolling, but as a considered argument about what happens when the foundational premise of a multi-hundred-billion-dollar enterprise turns out to be uncertain.
A writer with sponsors doesn't write those pieces. Or they write them softer, with hedges that serve as emergency exits. Romero writes them as if the reader deserves the actual argument.
The Publishing Machine
The cadence is metronomic: how-to guides on Mondays, curated AI news with context on Fridays. This is not accidental scheduling - it reflects a considered split between two modes of value. The Monday pieces teach; the Friday pieces contextualize. The former builds skills in readers, the latter builds perspective. Together, they describe a publication that treats readers as people who want to understand, not just to feel informed.
He writes about AI through three lenses he names explicitly: Humanity (the effect on people), Honesty (no editorial compromises), and Quality (practical value, not content volume). These aren't marketing copy. They explain why his newsletter can publish "30 Things I've Learned About AI" and make it worth reading - because each item is earned through eight years of actual study, not assembled from takes.
Calling Shots, Then Waiting
His 2025 predictions for AI ran 20 items long. His analysis of OpenAI's o3 model ran eight insights that cut past the press release to the structural implications. His piece on "The Shape of Artificial Intelligence" is a geometry problem - he literally maps the ceiling versus floor of AI capability as a spatial argument, the kind of move that comes from someone who once studied the physics of flight and the topology of neural networks in the same decade.
In April 2026, he published "AI Will Be Met With Violence, and Nothing Good Will Come of It" - a look at societal desperation and AI backlash that many mainstream tech writers were still treating as a fringe position. He tends to arrive at the argument before the consensus does. The newsletter exists, partly, because he wants to share the view from that vantage point before it gets crowded.
"Honestymaxxing is probably the best writing hack that I know of. It forces you to deal with yourself in a way that you usually wouldn't to get out the truth of what you think and feel and put it on the page. It should make you slightly uncomfortable."
- Alberto Romero, on his writing philosophyThe View from Madrid
He covers an industry that is overwhelmingly concentrated in San Francisco and the northeastern United States - and he does it from Spain. The distance is not a handicap. It is, arguably, the source of his clarity. Silicon Valley is a context that rewards enthusiasm, penalizes doubt, and makes careers by proximity to the next hot thing. Madrid has no such gravitational field. It is possible to think slowly there.
He joined Twitter in December 2020, when the AI conversation was still technical and niche. He wrote on Medium before Substack became the obvious home. He was building the readership and the voice before the audience materialized. When ChatGPT arrived and made AI suddenly legible to everyone, Romero was already mid-sentence.
Where He's Going
His stated aspiration is to bridge the gap between uncritical AI enthusiasm and reflexive AI fear - to build a media platform that holds the industry accountable while helping ordinary people understand what is actually happening. At 40,000 subscribers and growing, he is doing exactly that, with the stubborn independence that comes from funding it through his readers rather than his subjects.
The sign language translator gave machines a way to understand humans. The newsletter gives humans a way to understand machines. He has been building the same bridge twice, from opposite ends.