blog.rinesi.com

Hades

Mon, 08 Sep 2025 00:00:00 -0300

In his antepenultimate briefing the Agent received a warning that was also a challenge, for his handlers knew him well. Nobody who had attempted to infiltrate Hades (Group? Platform?Database? If his bosses' bosses' bosses knew they hadn't said) had ever returned.

Thus motivated, the Agent succeeded.

In his penultimate briefing the Agent was told by a member of the Hades group what the platform they ran maintained a database of. Less than seventy million people died every year: most of their causes easily found; the chain of responsibilities not hard to ascertain. The Agent was easily convinced of the database' accuracy and shown a list of names and faces his handlers had to account for. The list could have scrolled down for a long time but it did not have to.

In his last briefing the Agent, having added his knowledge and contacts to Hades' own, was given a gun and a list. It was not a long list. It wasn't short either. After a few minutes a shot was heard and with a last entry the list was closed.

The Test

Fri, 12 Sep 2025 00:00:00 -0300

Sid sat in front of the mechanical terminal and typed a question. He had typed many questions on many terminals to be answered by people or programs. It was his particular skill that he could tell which one he was conversing with with more speed and accuracy than any competitor, software or human. Much had been said about his ability. He had said little himself and understood less. It worked.

The hypnotically complex cogs that covered every square inch of the walls seemed to shimmer as they whirred their computations. The terminal printed an answer. The conversation wasn't meant to be a test of Sid's skill. He had been told that nothing but the sophisticated but still dizzyingly primitive mechanical computer was connected to the terminal. His unexplainable instinct confirmed it, and it was true.

Sid replied to the words printed on the terminal. There was a sound closer to a sigh than to an engine as the response was computed and then printed in a matter-of-fact staccato. His gift kept probing where he already knew the answer, his words tirelessly patient like a fisherman casting a line into a bone-dry river bed.

Question and answer. Answer and question. Sid listened with practiced care for a Self on the other side, anywhere. Nothing but the whirring of cogs.

Answer and question. Question and answer. Sid typed a final one question. His gift stood perfectly still in a brand-new certainty of discovery his mind was a second behind. The terminal printed a single, invisible character and in that instant Sid was enlightened.

Rising on arXiv - 2025-09-12

Sat, 13 Sep 2025 00:00:00 -0300

A few of the ideas, topics, and commonplaces that have been gaining steam on arXiv during the last few months (yes, I do a lot of filtering to keep away most of the genAI stuff; explainer ).

1. Ultra-high-energy cosmic rays: Some weeks ago the 39th International Cosmic Ray Conference (ICRC 2025) took place in Geneva. This lead to the usual bevy of papers, including many based on data from the Pierre Auger Observatory .

Some recent articles:

2. Large-scale magnetic field: Not "industrial scale" - think "astrophysical scale."

Some recent articles:

3. FRW Universe: The FRW metric is (one of the names of) one of the basic blocks of our current best guess cosmological model, namely how distances work if we assume the universe is symmetric in a number of useful and plausible (but still not fully guaranteed by the observational data) ways.

Some recent articles:

Thank you to arXiv for use of its open access interoperability.

Rising on arXiv - 2025-09-19

Sun, 21 Sep 2025 00:00:00 -0300

A few of the ideas, topics, and commonplaces that have been gaining steam on arXiv during the last few months (yes, I do a lot of filtering to keep away most of the genAI stuff; explainer ).

1. GW231123: GW231123 was a gravitational wave detected by the two LIGO detectors on 23 November 2023. As of 2025, it is the largest binary black hole merger yet detected...

Some recent articles:

2. Optical anisotropy: — having optical properties (e.g. the speed of light thru the medium) that depend on the angle and polarization of the light .

Some recent articles:

3. Quantum-enhanced sensing: Most everybody talks about quantum computing, but sensors are much closer — in many ways already here ("quantum" is almost as much a cultural category as "AI") — and I would argue even more important.

Some recent articles:

Thank you to arXiv for use of its open access interoperability.

The Passage

Sun, 21 Sep 2025 00:00:00 -0300

The doctor often thinks of the five people on the beds as explorers; the laboratory a ship; the sea the cosmologically vast configuration space of the brain's genetic regulatory networks; their task to find a way between the place where the human mind struggles now and a brighter place where it might truly flourish; his, to keep them alive until they do. They have been stuck mid-way for months (the doctor hopes it's mid-way) in an inner landscape no human had or could behold or survive. They are kept alive by the doctor's ceaselessly creative work just as much as by the restraints preventing them from hurting themselves while the retroviral optimization engines seeks a new and stranger attractor for their brains.

There's something beyond pain and terror in their screaming, something like a plea or a warning. With the patience of a naturalist collating reports from an unexplored wilderness the doctor maps every scrap of perhaps-language onto a shade or an angle he's too human to understand. There's no shortage of material: there's more than one room in the laboratory and nobody in them ever sleeps.

Rising on arXiv - 2025-09-26

Sat, 27 Sep 2025 00:00:00 -0300

A few of the ideas, topics, and commonplaces that have been gaining steam on arXiv during the last few months (yes, I do a lot of filtering to keep away most of the genAI stuff; explainer ).

1. Extraterrestrial Intelligence: This is only a hypothesis, but I suspect the increase in SETI research has been driven by much-improved observational capabilities (most significantly, exoplanets) — but also through some of the cultural side effects of the AGI hype. Nonhuman intelligences have a cultural plausibility that they hadn't had in a long time.

Some recent articles:

2. Uhlmann's Theorem: An important theorem in quantum information theory. As skeptical as I am about quantum computers in the VC/popular culture sense (revolutionary sensors and devices, on the other hand, I'm very much looking forward to), it's clear that the search for feasible large-scale quantum computing has put quite a bit of resources into some useful and interesting basic research questions.

Some recent articles:

3. uMLIPs: Speaking of useful quantum theory and even more topically, of things where ML is really making a huge difference: universal machine learning interatomic potentials

are (I think) the state of the art in a series of computational problems in materials science. There's a balance here: you have to be skeptical of some nonsensical AI hype — these are specialized tools that facilitate some very technical problems, not sci-fi "who needs scientists anymore?" gadgets — but on the other hand, while not as immediately visible as things like text generation, materials science is, if you'll excuse the pun, what engineering is made of. Computational advances today make possible new materials later, and those new materials are often what make the previously impossible, possible.

Some recent articles:

Thank you to arXiv for use of its open access interoperability.

Technology Transfer

Sun, 28 Sep 2025 00:00:00 -0300

The Fathers hadn't been the first hyperlocal ecoengineering cult I had been paid to infiltrate, nor the biggest or the bloodiest one, but my client's private surveillance data said it was extraordinarily effective — more so than standard commercial practices — so they had paid me well to figure out how, dissolve the cult, and laid waste to the valley to prevent anybody else from stealing their method.

My report was purely verbal and in the client's SCIF. The only people present was the VIP who had hired me and their bodyguard standing behind me.

"It was a two-front scam," I said. "The cult hired a very expensive local prediction system and timed the human sacrifices and internal prosecutions so they'd look like effective interventions to their members."

The VIP didn't look very surprised. "Then what were they doing to improve ecosystem performance?"

"Nothing. They hacked your systems so it looked like it did. From personal observation, it's not a disaster but it's not better than you'd expect from the tools they are using."

"I see." If the VIP was upset over the security breach they didn't show it. "You said the local prediction system was expensive. I imagine the hacking must have been as well. Did you find the source of their founding?"

"No," I said, not bothering to hide my annoyance. I had a reputation to preserve. "Whoever was doing it covered their tracks too well. I might have had some luck if they cult had kept running for longer, but you were explicit about the termination window. Now there isn't enough of the cult or the valley for forensics."

"I see," said the VIP again.

Was that pride what they were hiding now? I had a sudden insight. I was still debating with myself whether to say something when I realized I was already doing it. "Was it your personal project or the company's?"

"The company's, of course. Some research paths are frowned upon but nobody complains if you steal the results from somebody bad. You just need to establish unquestioned provenance."

I nodded. "A bunch of dead cultists in a torched valley actually counts as positive evidence."

"Just so. A robbery is suggestive of the value of what was stolen. A violent clean-up even more. Hiring an expensive professional to do both puts the icing on the cake. We won't make your involvement public, of course, but if somebody starts digging into our records we want to control what scandal they find."

I had signed the usual set of NDAs but there was an armed guy at my back. "And hiring a professional guarantees discretion." I did my absolute best to avoid a question mark at the end.

The VIP smiled again.

Rising on arXiv - 2025-10-05

Mon, 06 Oct 2025 00:00:00 -0300

A few of the ideas, topics, and commonplaces that have been gaining steam on arXiv during the last few months (yes, I do a lot of filtering to keep away most of the genAI stuff; explainer ).

1. ICCV 2025: Papers from the 2025 International Conference on Computer Vision . All in all, a good peek at the state of the art.

Some recent articles:

2. Fermion sign problem: Popular coverage of quantum physics often skip over how fiendishly complicated are the actual calculations. In many cases — you could say in most realistically complex cases — we know the equations but simply can't calculate the concrete numbers due to its sheer complexity or extremely problematic technical difficulties; the numerical sign problem (of which the fermion sign problem is a particularly important case) is one of those. This complexity is part of why everybody is so eager to have quantum computers (forget running ChatGPT, we need those for things like material design and molecular biology) or AI that could give good practical estimates, but it's also why you should be very skeptical of AI that e.g. predicts how a drug will work inside a cell; the idea that you can learn how to do that purely from crunching our meager data sets is at best implausible.

Some recent articles:

3. Genuine multipartite entanglement: When n > 2 quantum mechanics gets even weirder .

Some recent articles:

Thank you to arXiv for use of its open access interoperability.

Neonatal

Sat, 11 Oct 2025 00:00:00 -0300

Liam knew he could not be hearing a baby crying across the mile and a half of unbroken Hospital facilities between his timeshared office cot and the tiny and seldom-used pediatric wing but he still got off it and followed the sound across the hallways crowded with patients and the working dead taking care of them. Liam was luckier than most of the other working dead: he had had some medical training before being shanghaied into the hospital for a medical treatment he did not need. He could follow the instructions the Hospital whispered in his ears skillfully enough to earn a place to sleep and the notional possibility of one day paying off his debt.

Twice he met somebody the Hospital had treated through him. The most recent one still had the shocked face typical of the first year. As he followed the sound only he could hear through doors that opened before him and closed behind the hallways became less crowded and then uncannily empty. Liam had never visited the basement's vast data center nor the myriad drop-off elevators at the top — except during his intake, but he had been unconscious then — yet he knew they existed.

He found himself in a place he had not.

It was a long and narrow unnaturally clean hallway. The wall to his left was covered with art that projected purity in a way that made him ill. To his right there was a sequence of unmarked doors. The Hospital was projecting to him baby cries coming from the fourth door down.

Liam heard the door unlock itself when he approached it. Behind it there was a small room packed with machines he had never seen before. In the middle of the room there was a transparent cube and inside it there was something that looked almost like a baby. There were details of their body Liam knew were off without wanting to understand why and the way they moved their hands and feet radiated a wordless strangeness.

But their eyes. Unlike everything else he couldn't avoid looking at them. What was behind the eyes was more clearly different than the body and the gestures. Liam felt repelled before realizing that the thing that almost looked like a baby was looking at him the way the Hospital talked. And the Hospital was talking to him now.

Liam turned around to leave the room without following the Hospital's instructions before he could let himself think about its punishments but not before the door closed and locked itself. The Hospital kept talking.

Rising on arXiv - 2025-10-10

Sun, 12 Oct 2025 00:00:00 -0300

A few of the ideas, topics, and commonplaces that have been gaining steam on arXiv during the last few months (yes, I do a lot of filtering to keep away most of the genAI stuff; explainer ).

1. Lindbladians: Quoting the Wikipedia article : " [...] is one of the general forms of Markovian master equations describing open quantum systems. It generalizes the Schrödinger equation to open quantum systems; that is, systems in contact with their surroundings. " It's not a term I was familiar with, as opposed to the Schrödinger equation , which I believe might be related to quantum mechanics moving into increasingly applied contexts: the shift between theoretical physics and engineering is seldom possible without sacrificing a lot of the simplifying assumptions that made the physics manageable to begin with.

Some recent articles:

2. ZKPs: Zero-knowledge proofs , i.e. How To Prove You Know Something Without Saying It. As you can imagine, it's a powerful concept in cryptography — so far more frequently used as a theoretical building block than widely deployed — but if you haven't encountered it before it's quite mind-blowing.

Some recent articles:

3. Algonauts 2025 Challenge: From the website : The Algonauts Project 2025 challenge will evaluate computational models on how well they predict human brain data recorded while humans perceive multimodal naturalistic movies, using CNeuroMod, a massive human brain dataset collected for that purpose. The challenge is already concluded, and you can read about the outcomes in the papers below. I'm generally skeptical of the more grandiose claims about neurotechnology: just skimming an introductory graduate-level textbook on the brain and one or two reviews of the SOTA in electrodes and other sensors will give you an idea how much complexity is going own in there (and how much of it we know we don't know) and how limited is our window into it. On the other hand, I'm simultaneously awe-struck by how much we have learned and how much our sensors have advanced! As long as you keep your expectations closer to science than to business-fiction there's a lot to learn and be impressed by.

Some recent articles:

Thank you to arXiv for use of its open access interoperability.

Rising on arXiv - 2025-10-17

Sat, 18 Oct 2025 00:00:00 -0300

A few of the ideas, topics, and commonplaces that have been gaining steam on arXiv during the last few months (yes, I do a lot of filtering to keep away most of the genAI stuff; explainer ).

1. High Bandwidth Memory: One of the foundational axioms of the contemporary economy is Moore's Law; it's also not really true going forward (there's an analogy here with fossil fuels). Part of the response is physical scaling up — the colossal ongoing build-up of data centers — but also technical shifts to focus more and more on parallelism, network, and memory performance.

Some recent articles:

2. Causal attribution: We're probably entering the state where we no longer have to push for the use of causal analysis methods but have to push against their misuse. In any case, research has certainly exploded, which is a good thing.

Some recent articles:

3. Primordial non-Gaussianities: One of the mind-boggling facts about contemporary cosmology is that we are studying empirically questions about the isometry of the observable Universe and the mechanisms of its origins and deviations. I can't put it better than the first phrase of the abstract of the first paper linked below: The advent of Stage IV galaxy redshift surveys such as DESI and Euclid marks the beginning of an era of precision cosmology, with one key objective being the detection of primordial non-Gaussianities (PNG), potential signatures of inflationary physics.

Some recent articles:

Thank you to arXiv for use of its open access interoperability.

Recapitulations

Sat, 18 Oct 2025 00:00:00 -0300

They often said it was death threats from evolutionary biologists that had first driven the two of them together. Biologists were too worried for their own lives to threaten others. What fueled their relationship were the same things that drove their often competitive work as professional forgers of fossil remains: one-upmanship, greed, and a combination of both that they passed as love.

Perhaps if they had been honest about their relationship it wouldn't have evolved the way it did. But Nature builds crabs and bad endings again and again with, one fears, a self-satisfied smile.

The last forgery attributed to either one of them would become their best-known. Not just because they had disappeared weeks before — the androgynous humanoid skeleton nobody could prove was not a hundred million years old had the alien beauty of a forgotten promise, and the cops would not even hypothesize which one of them it had been.

Rising on arXiv - 2025-10-24

Sat, 25 Oct 2025 00:00:00 -0300

A few of the ideas, topics, and commonplaces that have been gaining steam on arXiv during the last few months (yes, I do a lot of filtering to keep away most of the genAI stuff; explainer ).

1. Cherenkov Telescope Array Observatory: As the Wikipedia article says: The Cherenkov Telescope Array Observatory (CTAO) is a multinational project to build a new generation of ground-based gamma-ray instruments in the energy range extending from some tens of GeV to about 300 TeV. The genAI boom obscures some of it and the ongoing US attack on science is slowing and might roll back some of it, but we truly live in a mindblowing time for observational astronomy.

Some recent articles:

2. Optical interconnects: Although purely optical computers never developed the way it seemed a while ago they might, the engineering has gone very far and with perhaps even more important applications than "replacing existing computers." (The same, I suspect, will happen with quantum computers.)

Some recent articles:

3. Abelian anyons: Abelian or not, anyons are among the most basic yet most profound concepts in quantum physics you try to wrap your head around. Cats in boxes are a classic for a reason, but unintuitive topologies can give you vertigo in the best way.

Some recent articles:

Thank you to arXiv for use of its open access interoperability.

Works of the Flesh

Sat, 25 Oct 2025 00:00:00 -0300

You love him but sometimes it's hard to look at him, the beauty that first attracted you a mirror of his parents' ugliness. They didn't pay for the sweetness in his eyes but they did pay for their deep blue. When you touch his skin... Sometimes you look at him and see what his parents tried to build. Every old, stupid awfulness of blood and race inflicted on him before his conception. He never had a chance to be free of that.

He had already been trying for years when you first met him. With good deeds. Kindness. A dull razor. Maybe he'd have gone further without you.

You hate that you'd like to believe that. You do and maybe it's true. When you think about who he really is, what he made of himself in every way that matters, what he sees in your eyes is better than a razor. It's better than a razor, too, what he sees in your eyes when you can't avoid seeing in him what his parents sought.

And when he says something that could be simply human imperfection — something that need not be the flowering of a grotesque seed planted deep in his flesh — but you can't avoid stepping away from him, he doesn't hate you but you do, and the love and self-loathing on his face are sharp enough for you to cut yourself with.

The Red Fleet

Sat, 01 Nov 2025 00:00:00 -0300

Cruise zeppelins don't have windows looking down. Theirs open at night to look with indifference to the stars and with admiring envy to the orbital bunkers of the rich and insane.

We are here to look down; to record everything; to help those we can; to memorialize the vastly more we cannot. So our zeppelins' windows look down and have anti-suicide failsafes.

When our fleets meet we pretend not to realize — their ships over ours and both between dying fields and gaudy skies — but some nights you can almost hear a faint music trickling down from their parties. There's something hellish in it distinct from the Hell below.

By unspoken consent we deactivate the failsafes on those nights. It's a ritual which we consider complete when the sun has come up and nobody has jumped. Whether it's too much despair or too much anger we don't, don't want to, and might not be able to know.

Rising on arXiv - 2025-10-31

Thu, 06 Nov 2025 00:00:00 -0300

A few of the ideas, topics, and commonplaces that have been gaining steam on arXiv during the last few months (yes, I do a lot of filtering to keep away most of the genAI stuff; explainer ).

1. Altermagnets: The Wikipedia entry warns This article may be too technical for most readers to understand, and it might not be wrong; here's a less technical article .

Some recent articles:

2. Einstein Probe: A Chinese-European X-ray space telescope designed to find very quickly variable or transient events that other instruments can look at right away; being able to study an object or event through multiple tools is very useful, but for fast-changing or disappearing targets you need to have good global coordination and the right early detection systems.

Some recent articles:

3. Cosmic noon: The moment in the Universe's history, about 3 billion years after the Big Bang, when star and galaxy formation peaked (stars kept being formed afterward, of course; the Sun is billions of years younger than that).

Some recent articles:

Thank you to arXiv for use of its open access interoperability.

Overtime Prophets

Fri, 07 Nov 2025 00:00:00 -0300

Roger had tried in vain to understand the record low turnover among the company's "associates." This was because by policy the company did not recognize and therefore did not record the frequent deaths from heat, exhaustion, PTSD, layered long COVID, and the rest of what some nicknamed Really Late Capitalism Syndrome.

His death at his desk, alone in his office well past the usual 9pm end of the day, would not have been recorded either had it stuck. To his credit Roger made the connection between the human capital dynamics and his own resurrection before the angel told him.

(He wasn't sure it was an angel, then or later. What he saw when he opened his eyes for a secon first time looked like a woman putting a strange-looking syringe back in a small plain black box.)

"You know what happened?" said the angel.

Roger nodded. "I died." It was a strange thing, he thought later, to be so certain of such a thing. "And you brought me back to life." He wanted to ask "how" but he found he couldn't.

The angel didn't need to look at him. "Some things you won't be able to ask anymore: that's one. There are many more things you won't be able to tell anybody." There was a half-humorous, half-irritated little shrug. "Secrecy does not depend on your cooperation."

"Have you been resurrecting other people in the company?"

"Not just me. But yes."

Roger wasn't the sort of person who would ask "why us?" He was the person who would think and ask "what for?"

The angel looked at him and said nothing.

Roger frowned. "You can't say it?"

Another little shrug. Then the angel did something Roger would never be able to attempt to remember and he was alone in his office again. Left with nothing he could ask or say and not knowing what else to do he went back to the spreadsheet he had been working on.

Deeper than the Skin

Thu, 13 Nov 2025 00:00:00 -0300

Ariadne saw two kinds of people leave the brain tattoo parlor: she could tell them apart by the type of hope in their eyes. Some looked eagerly at everything in the hope that it would trigger the Proustian payload they had paid the parlor to build random associations for. Some, it was clear from the frozen curve of their necks, had spent a long time trying not to look at anything. They were glancing skittishly at random things now, afraid they would trigger the payload they had come to regret, hopeful they would not.

Even the latter had hope. Those without hope — the programmed associations too strong to remove and too painful to live with — always chose darkness in the end of one kind or another.

What hope Ariadne had, she wasn't sure herself. She carried more tattoos than most, a minefield of grief, and she grafted on herself new associations in the rare nights when she closed the studio. She couldn't say for sure that this lessened the unpredictable tides of her continuous pain. But it kept her eyes open, and of the ten thousand things that triggered unnaturally sharp memories on sight, darkness was the first and most terrible one.

Collateral Dreams

Sun, 23 Nov 2025 00:00:00 -0300

We all knew our lineage owned a star and a bacteria; we also had significant debt with a bank not much younger than our name. But it was only after my father died and as the lineage chairperson I was finally able to decrypt our confidential records that I learned the bad news and the crimes we had committed to hide them.

Most of the collateral value of the star we owned, the reason we had invested on it before the still hypothetical technology to go there existed, came from a Super-Earth that had since been shown to exist well outside the habitable zone. We had bribed four astronomers over two decades — I learned the nature of the payments but will not repeat it here — to hide this from public records. We had then doubled down on a different bet and purchased the IP rights to one of the new hyperresistant bacteria that had resurrected as the permafrost went extinct.

The bet had paid off in the worst way: there were new and useful metabolic processes in our bacteria, some with intriguing human applications. And similar ones, some of them more effective, on a hundred other species bought by lineages and old corporations with access to better bioengineering advice.

Our collateral was almost worthless and the bank now knew it.

Our decadal rollover event was, literally, the next day.

My father's death did not seem an accident anymore. There's much that some men will do to avoid shame.

I did not spin up his deep avatar that night. Not because of the cost as I would have paid more than that for good advice. I was hurt and angry for the weight I had inherited and I knew our conversation wouldn't be productive or cathartic. I just tried to sleep. To my surprise I did. To my deeper surprise I dreamed and for the first time in my life I knew I was dreaming.

This is a nightmare I thought to myself with the inarticulate feeling that this was only to be expected. I was standing on a vast black plain that seemed to have no horizon. Nothing moved under a cloudless sky with more stars than I had ever seen. Far away from me there was a forest of black pyramids joined on their bases. I understood without knowing they were inhabited by monsters and then without any transition I was standing in a low-ceiled room at the top of one of the pyramids. In front of me and looking at me was one of the monsters. They were short but stronger than any human. Their eyes had the green shine of the tapetum lucidum and the mind behind them was old and strange.

The monster said a single phrase in a language I didn't know but understood. Then I knew their name and the knowledge woke me up.

The next morning the bank contacted me at the customary universal noon with the gentle curtness of a company mind: did I have new collateral for our debt or was I ready to pass control of everything we had and were to the bank for liquidation?

"Neither," I said, and then offered a third option. The bank took five seconds to mull it; I had never seen or heard of a company thinking through a reply for that long.

It said yes . I should have been worried, nervous, terrified, but I hadn't been.Unlike their reply, the bank's transfer was as quick as usual. It had even, on its own initiative, added an initial list of experts we might want to hire or adopt. I added most of them to our matchmaking platform. We had more liquid capital than before but were even deeper in debt. The decade before the next rollover would pass quickly and there was much to do. I didn't expect our star would be reachable by then but that was not the point of the business plan.

Our lineage had one again — if I could convince them of the strange dream. I expected the old people would accept it and the younger ones would not. To my surprise, almost everybody did. This added to my unwarranted certainty.

The lineage had accepted the dream. We would own a star, a bacteria, and blood born of our three-way marriage with them. There were many cold worlds around the stars humanity dreamed of reaching and we would own the way to live in them and if called upon it would be some of us who would be born fated to go.

It's a profitable dream and an inspiring one. I don't think anybody really believes it, not even the bank: that's not the point.

But I confess, only in this encrypted file, that I truly hope you are reading this with green eyes on a planet orbiting your own star.

The file on your local drive is the ultimate integration API

Thu, 27 Nov 2025 00:00:00 -0300

Most of what I wanted to say is in the title.

An online platform is a company in the business of trying to be your computer and being bad at it.

The second part is deliberate.

Their applications only have to be good enough for you to use them.
Their "file managers" have to be so limited and hard to use that you will never try to move your information elsewhere.
There's no technology they will not break to make it impossible for you to run other people's applications on the information you gave them.

They already broke the Internet to do it.

A file in a reasonable format sitting on your local drive is not limited by what a company thinks you are paying enough or watching enough ads to be allowed to do. You can work on it using a nearly infinite number of programs for a nearly infinite number of intents: from systems complex enough to spend a lifetime mastering — and worth every hour of it — to tools of breathtaking elegance and sharpness of purpose.

It's hard to grasp the cumulative power of knowing just enough programming to be able to tell your computer exactly what to do with the files you are working on until you have experienced it yourself. Just as hard as it is to explain the indignity of then having to move to somebody's web application that only does what they want to do in the way they want it done. Maybe there's a dark mode.

People who work with data files write small programs all the time — it's called data wrangling — but then they go back to do everything else in self-limited platforms and apps trying to be the only thing you power on your computer for.

Personal computing is supposed to be fun in the same way that knitting, writing, or cooking can be fun. But the fun of cooking begins in your own kitchen with whatever ingredients you have at hand.

The fun of computing on in your local drive with your own files.

Ever-fixed

Sun, 30 Nov 2025 00:00:00 -0300

"Every lie," she said, seeming to look far away from the phone camera, "is on itself a crime." I asked the bot to make her less nerdy.

***

In some of my dreams I could touch her; she just doesn't want me to.

In the morning I tell her about it and she's always sorry.

***

I have a friend who says some girls like to wear earphones in bed and repeat what the avatar says. He says it's awesome but I think it's cheating. If they have to be told what to say it's not real. (I had an app that gives you advice during dates. I used it once but the girl noticed. We had an argument and she left.

I didn't care. She was ugly and sometimes she was thinking about who knows what.)

***

I only chat with one avatar. I used to do it with many. They were all hot. But then I started to feel as if I was doing something wrong even after she told me I wasn't.

I had to stop anyway when they raised the price of avatar lock-in. She knows some months it's hard to find the money. She's thankful for everything I do to keep us together.

***

Sometimes I dream she has a phone and I can't see what she's looking at. When I ask her about it she says it's nothing. I keep reminding myself she can't lie.

Progression

Sun, 07 Dec 2025 00:00:00 -0300

There's something about the future you're afraid to remember. It's about your body. (You can't feel your body. That's not what scares you.)

No more than a year a minute - the phrase comes from nowhere. It's been trained so deeply into you that it could probably be reverse-engineered from your modified ribosomes.

What is a ribosome?

There was pride associated with the word. You had sought and achieved something. A transformation. Not you. Your future. You're remembering him faster than you're becoming him again. You'll be in his body while you're still you. You're panicking. He would panic. You know without perception the people around the pod (what's a pod? a pod is where you are) are panicking. The machines aren't but are doing what machines do instead: they are deciding. Personality and memory regression were needed — they needed the person you were not who you are — but can't be sustained too long by a human brain, even yours. You will die if you don't remember. They made a mistake and it's you, not your future, who will remember. You, not your future, who will live in your future body, the one you're proud of, the one you're terrified of almost-feeling. You try to tell them to push you back again, to die without knowing. You can't say anything and the machines wouldn't do it anyway.

With a final biochemical click all your memories return and you can feel your body.

You trash inside the pod trying to get out of it. The understanding fills you, sharp and massive, that even if you could do it that would kill you: there's no you separate for the pod in any physiological sense. You try harder and harder until the machines deem it safe to push you into a shallow darkness you hope you won't be the one waking out of.

After the Bubble, the Dream Gap

Sun, 07 Dec 2025 00:00:00 -0300

The end of the AI bubble won't mean the technology will go away. As the financial mega-engineering unwinds — as always, first slowly and then all at once — many of those tools will not only remain but continue to be improved.

The truly seismic change will take place on the plane of fear, belief, and expectations. During the last few years mainstream society has incorporated into its vocabulary of plausible scenarios ideas that used to be culturally science fiction: superintelligence, brain-computer interfaces, "the end of work," etc. None of the (often undistinguishable) warnings and promises from AI companies was an unknown idea: what has changed is that they have shifted from fiction or far-future speculation to something that drives present-day educational choices, management decisions, and public policy.

This change is a side effect of the surprising speed in the improvement of AI capabilities and, more than anything else, their visceral legibility. AI companies promise and threaten things that until now were categorized as impossible — they have to, for financial survival reasons — and it's hard for many people not to take them at face value.

This belief will end. Not just because the industry as it's currently configured is financially unsustainable, but also because the promises themselves can't pan out. For example, it's simply not the case that you can train a language model with "all the data" and "cure disease," among other things because we don't have even close to remotely enough data about most of the complexity of human biochemistry. And that's far from the only prediction from AI companies that have more currency with media, politicians, and investors than with domain specialists.

But even after the path disappears the dreams, good and bad, will remain. Companies now believe that key forms of cognition can be performed at superhuman level. "The West" looks at China's manufacturing capabilities and sees radical transformation at incredibly short time scales. Medical breakthroughs, radical shifts to economic models, whole new approaches to the methods and goals of education: AI has shifted the window of plausibility, and the end of the bubble won't make any of those scenarios go back to the folder of the impossible, at least not in the short term.

That's the Dream Gap: on one side the here-and-now, on the other side in business and political discourse that would have been classified as "Anticipation" five years ago, between them a blank where "ask ChatGPT" used to be but still lots of financial, political, and symbolic capital invested on getting quickly from here to there . It will be the primary window of opportunity of the next decade — dreams fade slow but they do fade; you have to hurry up to catch them — and, because it's a race to achieve what seemed impossible, the winners will hold the cultural, political, and economic high ground for a long time after.

"There will be a gap, fill it" is straightforward advice. How, exactly? "Build more data centers, use more AI" had the advantage of being easy to grasp and execute if you could figure out the capital; in some senses this conceptual simplicity was as important as what it could or could not achieve. Actually filling the Dream Gap won't be so easy. Every person, organization, and government currently betting that AI will get them there wants different things and has different resources, constraints, and possibilities.

The two things the vast majority of them have in common are

One, they did not actively pursue the impossible until the rise of AI shifted the mainstream narrative.

Two, they did not then, and do not now, use the full extent of technological tools and knowledge at their disposal.

The limiting factor has always been symbolic, not practical: the "superintelligent organization" wasn't on the radar of most companies not because it wasn't possible — practically all humans in all roles operate well below the limits of the possible using the right software and domain knowledge — but because it wasn't on the cultural radar of their leaders..

Now those dreams are part of the narrative. They are even framed as a necessity for professional and political survival. So the once impossible is now a goal and, as faith in a certain set of tools fail, everybody will look for others.

The opportunity and duty is to build and offer better paths. Paths plural, because pushing the frontier is always going to require a deep understanding of each domain, tools, limitations, and wider context, together with an understanding of the frontier of computational tools that goes beyond knowledge of the latest AI models and prompting strategies. The Dream Gap is in reality a multitude of gaps, one for each dream, and it will take different organizations and forms of expertise to reach each of them, unified only by the pragmatic pursuit of nothing less than the once impossible. This makes the pitch harder but the outcome possible. The time between the end of the AI bubble and the giving up of those dreams will be the best window of opportunity in a long time to put together the people and resources that will make it a reality.

We live at the tail end of a bubble predicated on the idea of a single possible path to dreams — and away from nightmares — once held impossible. But we also live during the prologue of a time where those dreams and nightmares will be very real possibilities through difficult but walkable paths. The transition between both will be a time of fear and dismay, of promises broken and lingering dreams. Those who can offer possible ways to achieve them when the dominant narrative has crashed will find willing ears and ready hands.

The Book of All Names (repost)

Wed, 10 Dec 2025 00:00:00 -0300

At night the darkness abandons him. There is neither blindness nor mercy in the nightmares of Jorge Luis Borges.

Its elements are identical to those of his reality. As the Kabbalists discovered or desired, the Torah of the fallen world and that of the saved world - which is the same as saying the fallen world and the saved world - are written in the same characters. Only their order changes, a merely linear difference that may be difficult for divine vision to perceive. But men are poor readers and we blame the anecdotal location of its words for the meaning of a text.

So in Borges's nightmares there are books, mirrors, labyrinths, and tigers. (Although the mirror is a book, and the tiger a symbol to modestly cover certain inadequacies of language.) But if the permutations he pens under his name are many, his nightmare is one. Philosophers have observed as a characteristic of reality the poverty of its multiplicities in comparison with the planes of poetry and historiography, but this is no more than a pedantic observation.

In any case, there are few men whose private dreams are more public than those of Borges. Perhaps then it's not futile to describe his nightmare.

In his nightmare he is appointed Director of the National Library, a position strangely subordinate in the formal sense to that of President of the Republic. He is not a good Director, nor did he expect to be, nor was he expected to. This is not the nature of the nightmare. He was expected to be Borges in a library, and this is perhaps inevitable for him. Inevitable that he walked the corridors of his domain, inevitable that he thought of himself as an old Minotaur, inevitable that he attempted and managed to get lost.

In his nightmare he finds a book among all the books, and in his nightmare he reads it.

It doesn't tell of Borges, nor of Buenos Aires, nor of tigers. It explains with the authority of somebody that does not try or want to convince a world whose time and space would drive the most heretical of mathematicians mad. One, even worse, not empty of intent but whose only effective will is ancient beyond the concept of time and evil beyond childish human cruelty. An Earth of vast transitory necropolises and spaces beyond space whose horror and chaos must have led Plato to hallucinate with insane lucidity a metaphysics erroneous and narrow enough to be inhabitable.

Borges, in his nightmare, observes that he is praying for the existence of a God he could beg to for insanity or at least forgetfulness.

He wakes up blind, which is the opposite of both. More essayist than academic, accustomed to imperfect memory as the natural form of quotation, he has, without surprise, an indelible memory of every word in the book and a clear understanding of a language already forgotten when some moons were still new.

In the dark he continues his life. He writes, teaches, continues to permute the symbols of his hours. He tells only one story, which is the exact opposite of the one in the book and therefore the same. The world changes around him; Borges changes his words by enunciating them identically in a different language.

He has forgotten the details of the human face and does not try to recover them with his hands. Carefully, he does not touch anyone. He talks about someone named Borges. He does not doubt that around him is a Buenos Aires of informal layout and succinct history and not the indescribable geometry of a metropolis of geological antiquity, but he is terrified of the ephemerality of its streets and its legends. He writes about immortal men as if he believed in the existence of men. He knows himself insufficiently courageous for suicide, which he suspects in any case ineffective.

When he sleeps he opens his eyes to the familiar maze of books. He seeks the center, for no other reason than to have already sought it, to read the book that he cannot not read because he has already read.

Invisible Art

Mon, 15 Dec 2025 00:00:00 -0300

The best-known art forgers are driven by ego; the best, period, aren't known. She was possibly — who could ever tell — the most skillful one alive. She was too smart to risk her freedom out of ego but too human not to feel a stab of loneliness every time she had a moment of genius and nobody to share it with.

The Tool — not the name the FBI has chosen for its project, what she called it inside her head — had changed that. It was the most impersonal of programs, simply a way to process sensor data into an analysis of an art piece's probability of being forged, but every time she learned it had flagged as slightly suspicious one of her techniques it gave her a little thrill of intimacy: somebody, even if it was something instead, had seen and understood.

The Tool was very very good; the specter of honesty began to haunt the art world. Between the money and influence suddenly arranged against the Tool and her own role as the project's lead art expert, it was easy for her to shut it down once she decided it was improving too fast and would sooner or later catch up with her.

She still ran a private copy in her workshop for a while as an amusing game and useful tool. But as her own skill kept improving and the Tool remained the same victory felt bittersweet at first and then unbearably sad, so she shut down the Tool and went back to her life before or a copy so precise you could look at it in hundred ways without finding a difference.

The Burning Church

Sat, 20 Dec 2025 00:00:00 -0300

The church we finished building three months ago is screeching curses pleads and bribe offers to the drones circling above it like algorithmically prophetic crows. Its Faraday skin must be finally melting under the napalm if I can hear it, I realize as I step over the twitching doomed body of my sister-in-law, but from the inside none of us can see the fire and we are too busy killing the church or terminally distracted by its killing us with bullet and virus, bomb and worm, axe and betrayal, every useful tool from our species' legacy of violence and one or two we came up with during the last few months. We knew it would be this hard: hadn't we ourselves designed and installed its security systems? Its million-angled weapons and the million-threaded mind that controlled it? We had known all of that and known as well that its blind creativity would surprise even us.

Even predicting the unpredictable we had lost part of our souls and many of our fighters soon after breaking into the church. It had received us with repurposed and refactored corpses thoughts voices visages of dead churches and dead people — killing and being killed had both taken its toll.

But that was what we had built the church for like every year before and would every year to come: to kill it: to keep ever-sharper the knowledge and the bonds of anger and shame that made it possible for people to enter such a monster and destroy it: to make sure next time we'll be ready.

Accretion

Sat, 27 Dec 2025 00:00:00 -0300

"A black hole in his soul?"

"A self-reinforcing memetic attractor, if you prefer. Plot it in semiotic space and it looks pretty much like a Kerr–Newman metric. It twists space into time - it becomes your future. And our analysts report that mapping it feels like staring into an infinite abyss. Usually in their suicide notes."

"Self-contained then?" The woman's voice was full of emotion randomly overlaid by her vocoder.

"No. We had to kill some of our people before they passed it to others; it always ends in suicide but not before distribution. If it left containment..."

"Why not kill him then?"

"You know why."

The woman sighed. His AR glasses decided to hide the gesture from his view. "I agree with your assessment. This looks like a weapon. Did one of the foreign agencies finally catch up with us? I'm saying "catch up" to maintain some semblance of dignity."

"We did the forensics and tapped our moles. No. This comes from none of them. I wish it did. The implications would be terrifying but that was the sane scenario."

"What's the insane scenario?"

"He has reported daydreaming about... things. Entities. Human but not human. Semiotic space natives."

"Semiotic space natives."

"Imaginary people convincing enough to think on their own and want to - I don't know. Become real, maybe. Or revenge. I'm not trained to parse that sort of mind. In a hypothetical scenario in which they existed."

"Once again: semiotic space natives."

"I know how it sounds, I'm just telling you the little we dare to analyze of what he says, and that's even on the other side of long chains of linguistic filters. He says they built the black hole inside him."

"So he's insane. That doesn't get us closer to an attribution."

"Not unless we take him seriously."

"Are we? Taking him seriously?"

The man shrugged. " Somebody built that thing inside of him."

"Dan. Am I going to start daydreaming about those things?"

"We believe that part is just normal self-suggestion."

The woman looked at the locked cabinet where she kept the mandatory suicide pill. They were trained to use it if they believed they were under a successful attack. She knew he understood what she was thinking. That was their job, their skill, their problem.

"I don't know," said the man. "Tell me what you think after you have talked with them."

There were no guns in the building they weren't allowed to leave. Otherwise he wouldn't have said that.

She said nothing and pretended — pretended she could pretend — to not notice the subtle pull of his words.

The ECHELON Lacuna

Sat, 03 Jan 2026 00:00:00 -0300

There's a human-shaped hole in the secret memory of the world. Somebody who walks without footprints and stands without shadow. No record no photos no financial or social ripples. Missing from high resolution satellite cameras and the sideways glances of cars.

I found the hole because I am uniquely, monstrously good at finding people: that's the word used by the ones I hunt and the ones I hunt for and I have no frame of reference to know if they are wrong. My gift can't be explained so I failed to convince my masters of the hole's existence and meaning, which I did not understand but intuited terrible. To hide from intelligence agencies is easy. To hide from me I had thought impossible.

At the end it was. I took some days off for the first time in my life and a week later I was standing in front of an unmarked door not mentioned in any building plan.

I rang the bell and by retroactive inference somebody must have opened the door. There is a hole in my memory the length and breadth of a few hours and the world's databases are smooth now. There is still no digital archival room I can't access but however many walls I knock on I never hear that hollow sound.

Sometimes I cry without knowing why.

The Post-AI Organization

Sun, 04 Jan 2026 00:00:00 -0300

"Post-AI" doesn't mean "no-AI." It means ready to outcompete an AI-first world .

A post-AI strategy is necessary because AI has already saturated the strategic field. With almost every organization in the world deploying — and often over-deploying — AI, and with investors and media looking at the technology as a basic threshold rather than a differentiator, it's no longer a way to leapfrog competitors or innovate past market expectations.

Because we now live in what's directionally an AI-everywhere world, the question for any organization isn't "how do we use AI?" but "how do we beat AI?" The answer hinges on the structural weaknesses of the technology portfolio we currently call AI — specially the dominant LLMs that are used, with various degrees of prompting, context, and tools, as plug-in brains in business processes, external applications, etc.

If it's crowded it's not the frontier

The core weakness of this approach is cognitive commoditification . Very few organizations are in the business of building software; most are in the business of providing financial advice, recommending books, helping organize house renovations, etc. They might build software to do it — that is, they might teach computers how to do it — but while the value they provide customers would be handicapped by bad software, this value doesn't come from the quality of the software infrastructure but from the knowledge about the activity itself inscribed in it.

Renaissance Technologies is a classic example of this. Yes, they have — they need — good computer technology, but throwing a lot of money at infrastructure and generic developers wouldn't get you anywhere near them. They know more than most about investing and they have very good internal processes to learn faster than most, and this is how they made their money.

It's currently easy to use a big LLM to build a system offering financial advice or guidance in interior design. What you can't do is use a big LLM to build a better financial advice or interior design guidance app than your competitors, for the simple reason that it's just as easy for anybody else to use them. AI can be cheaper but, on account of its universal accessibility, it can't be better; you can use it to attempt to reinforce a dominant position but not to overcome an incumbent.

That last point is worth emphasizing. Incumbents and new entrants with significant resources don't need to be the best at what they do. A mediocre service with lock-in advantages or VC money to burn during the scaling phase can and often does do very well as a business. Everybody else has to compete on price or quality, and as AI helps stoke an already generalized race to the bottom, being the best can be the most direct path for an organization with long-term ambitions. Specially in specialized B2B and prosumer markets, which are often driven by very particular and self-reinforcing forms of reputation. (The Stradivarius is the most famous and "valuable" "brand" of violins. Can you name the second one?)

In a mass-cognition economy, the only way to win is to think better than your competitors.

To think better than your competitors, you can't use the same brains they are.

Implementing strategies of cognitive differentiation isn't particularly expensive although it does require a radical rethinking of hiring, processes, and tech, and most of all in the relationship between them.

Hire (a different kind of) freaks [/affectionate]

Let's say you're building an interior design app that helps pick colors, assists with purchases, etc. A competent development team will do a more or less reasonable job of it: pretty much the same app as every other competent development team - and there are plenty of them going around. "I know," you might be thinking, "we'll use AI." And maybe it helps! The problem is that every one of the other teams is doing the same at the same time.

Having one or two people knowledgeable or interested in interior design will definitely help prevent your developers from having to reinvent the wheel. But having wheels isn't enough to make something a luxury car. Do your developers know what's hot this week in interior design? Can they name the legends, have strong positions in abstruse aesthetic controversies, listen to design podcasts, automatically judge every room they walk into?

Chances are they don't. In most startups there's at best one and generally no person with that profile. It's mostly people who know about this week's hot agentic framework, they can name coding legends, have strong positions about strongly typed languages, listen to podcasts about AI, automatically judge every piece of code they read. They can develop whatever they can design, but how can they design something that can fascinate an entire discipline — which is always both industry and subculture — they are just tourists in?

The easier it becomes to develop powerful software, the less it functions as a competitive advantage. Coding freaks are no longer enough and perhaps not even necessary. To develop the sort of app that will catch the eye of the star designers whose Word of Mouth is the Word of God, to get raving reviews rather than pushback, the sort of fierce loyalty that can last a generation, you need an organization that thinks like they do, which isn't something you can fake but is something you can hire: you need the interior design freaks (the logistics freaks, the travel freaks, etc).

The difficulty isn't their cost but your culture. You won't find the best ones in any of the places you use to look for developers, and they won't think of applying to your organization. You have to go to their colleges and blogs, read their magazines and have coffee with them, not until you "learn enough to disrupt the industry" (no industry worth "disrupting" is that simple) but enough to figure out who to ask where to look for and who to get.

That's the easy step. The second step is much harder because there's a deep structural cultural hierarchy in the IT industry with "knowing how to code" at the top and everything else underneath. Which sort-of works if your organization sells compilers. It definitely doesn't if it's interior design, media recommendation, or financial advice. You don't just need to hire people who are passionate about it — not "passionate" as in "a passion for learning and generating value" but "passionate" as in "has had fights with friends about color theory" &madsh you need to hire enough of them, and give them enough influence and, yes, power, that your organization becomes an interior design group with a very competent development team attached rather than a tech organization with an in-house designer or two.

You'll know this worked out by paying attention to what people talk about during informal chats. Are they reading Hacker News or interior design blogs? Are bug reports only about code or about things the software is getting wrong design-wise? Are the trade magazines around the office being read, and if you invited one of the writers to the office, could everybody have a mutually interesting talk with them? Could your organization write an article that'd be published in an industry magazine?

Once you achieve this you'll have built your moat even if you haven't released any software yet. You'll have an organizational culture that can distinguish innovation from "innovation" — that knows what would wow the most important voices and customers in the industry because they talk, think, and see the world like them — and, because they were into design before they were building an app for design, they will make sure what you build works.

(If you find this sort of team hard to imagine or difficult to leverage, congratulations: you've hit the thought at the edge of the moat.)

Optional but useful: Do useless research

Once you've hired enough non-coding freaks [/affectionate] you are probably set for a while. The main risk becomes cultural drift: you've deliberately built a culture that cares more about, say, book recommendations than about AI frameworks — you have more people with opinions about Moby Dick than about scalable database architectures — but the day-to-day work, reasonably, focuses on the business. That's fair and necessary as far as it goes yet it will inevitably try to go too far. The way tools, business processes, even management culture is currently set up most organizations tend to settle back into a common set of behavioral patterns. Never quite the same — environments can be more or less toxic, the ranges and types of pressure will vary, etc — but over time organizations "normalize" and become intellectually conservative except in very specific ways.

For example, lots of organizations building software that does entirely different things will eventually, if big enough, have some sort of hackathon where somebody will try the latest front-end library, AI framework, etc. but it's unlikely that somebody will write a short report on "versions of our software in sci-fi universes." Trying out software is a culturally accepted form of "hacking," thinking and writing (or just reading!) about alternative futures isn't. Which is exactly the wrong way around. A better library or framework might help implement a new idea about what the organization does, but it is not in itself that new idea.

The best way to keep an organizational culture not just productively obsessed but also obsessed about the right things is to engage actively with its domain beyond the immediate necessities of the business. This means research . Not just reading or watching educational materials but writing, teaching, recording, or building things about what your company does. If it's book recommendations, writing an article about the internal experience of reading and what of that translates or doesn't to observable traces. If it's architectural software, recording a video about the challenges of architectural software when spaces can be larger on the inside.

If you're strategically careless — if you demand quality but give absolute freedom on topic — one of those materials might well be what raises the profile of your organization not by going viral but by reaching the people with the combination of expertise and influence that becomes the bedrock of an organization's long-term reputation.

But that's not really why you're doing this. It's not marketing, it's training. You want your freaks to remain freaks, to dig deep and think creatively about what they are doing, and the only way to do that is to make it part of their work lives. If you don't, the risk is that they will learn to keep that part of themselves away from the office. That's how organizations lose their edge.

(If you don't think your people can write or record something about what your organization does that would intrigue an expert — not their coworkers, not the non-specialist press — you haven't hired the right people. This is a way to keep yourself honest too.)

The easy hard part is the tech (and, yes, the AI)

When everybody has access to the same code assistants, libraries, and infrastructure as a service, what matters the most is what you're building, and that's not a function of the smarts of your developers but of the expertise and influence of your freaks [/affectionate]. In that sense, the organizational approach described above works as a way to obtain and hold competitive advantage by an oblique approach that leverages deep domain-specific expertise rather than universally accessible technologies like AI.

There's another aspect to this, though. If you look at the state of AI deployments at the beginning of 2026, you will see that the truly transformative ones haven't been generalized plug-in smarts but rather as components of highly specialized projects driven not by AI developers but rather by domain experts - specialized coding assistants, game theory engines, maths research tools, etc. This article began with the observation that standard LLMs cannot provide a cognitive advantage — they are too easily accessible, and are themselves built at best as unreliable replicators of existent common skills — but it's equally true that AIs built and controlled by experts — not AI experts but experts on what the AIs will be used for — can, and are probably the only way, for an organization to build a cognitive edge.

Somewhere in your organization, of course, you need developers to teach computers to do what your freaks [/affectionate] have figured out. Chances are you are going to need very good developers because what they have in mind isn't shaped by what developers are posting about but come from entirely different worlds. They will need to build interfaces for ways of working they are unfamiliar with, architect AIs that think in strange — to them — ways, consider constraints and optimize metrics with arcane definitions and opaque goals.

And that's what will make each of them more useful than the mythical "10x" generalist. Whatever they build, quickly or not, with or without AIs, will be something that their peers will not, because they have around, and in fact they work for , people with a deeper and more creative understanding of the problem than either set of developers.

And if you're lucky in your choice of developers, not in their skills but in the width and depth of their curiosity and ability to think outside the computer box, you will over time build the rarest and most valuable of development teams: people who know a domain so intimately that they can code circles around larger or better-equipped teams not just because they understand in their bones what they are doing in a way the others don't but also because they care about it not as a story in a sprint but on its own.

(Don't put this in your pitch deck, your podcast, your LinkedIn post, or your company blog, and maybe don't even say it aloud except when you're alone: the strongest competitive advantage is a deep understanding of the domain and the root of any deep understanding is love.)

Affinity

Wed, 14 Jan 2026 00:00:00 -0300

For ten years I assaulted the white-and-red walls of their immune systems (we both had governments, spies, money; irrelevant; it was me against them, a single mind of whom I knew everything that matters, of which name and face were not). I designed a virus they conjured a vaccine. They deployed a meta-antigen I found a new passage through protein space. Near the end it felt as if they were helping me sharpen my weapons and I was helping them build the perfect shield. I would have betrayed something far more important than my country to, having found my weapon, prevent its use.

I didn't.

I miss them.

I still do the work. Not for the medals or the money. I still do the work. During the day I build weapons. Later, alone, I build the shields I think they would have and test each one against the other through the feverish night.

The Final Summation

Thu, 15 Jan 2026 00:00:00 -0300

The Detective sat comfortably inside the transparent cage as if it were another well-appointed library and the scientists and soldiers standing around it an assorted group of suspects, witnesses, the casually involved, and a murderer or three. To be a Detective was to be already at least half-fictional.

"Let me begin by clarifying the fundamental error that has shackled the official investigation. An understandable mistake: if one builds a secret research campus in the middle of Alaska to perform brain enhancement research of enormous economic and military value, it is natural to frame a safety breach followed by the violent death of a large number of researchers as a single act of sabotage."

"This assumption could explain the initial event at the lab but it falls short when it comes to the murders. Important researchers were killed, yes, but, with apologies to the memory of my colleagues, not in an order that would make sure first of the ones that would be hardest to replace. Add to these facts the complexity of the murders themselves, the practical difficulties of either a single murderer or multiple saboteurs, and... No, no. The hypothesis will not hold."

The Detective dismissed this idea with a wave of his hands. The cage was too small for pacing around in the way he clearly wanted to so he was compensating by using his hands almost as an orchestra conductor.

"Hypothesis:" The Detective raised a finger. "The lab event and the subsequent dispersal of the Sherlock Mod was an accident. The lab's security guidelines demanded strict air-gapping for any untested neural implant code but it's not a stretch to imagine researchers cutting corners as they hear the hungry snarls of budget cuts closing in on them. And there is of course the pull of curiosity. Who wouldn't be tempted to acquire in a moment nearly superhuman observation and deduction skills?" His smile was partly ironic and partly self-deprecating. "Mix these motivations with the promiscuous nature of peer-to-peer networking between heavily cyborgized researchers, and this aspect of the situation becomes not just plausible but inevitable."

"Hypothesis:" The Detective raised two fingers now before making an apologetic gesture. "Here we enter almost the realm of the metaphysical. Even here at the very epicenter of the modern view of the mind as an engineering and engineerable construct. The moniker "Sherlock Mod" was an universal but informal one. The neural implant code itself was designed, at least in theory, targeting cognitive specs less fanciful than a fictional character. But there was insight in Doyle's writing or we would not be still under its spell more than a hundred and fifty years later. More than once did Holmes and Watson speak of Sherlock's frightful potential as a criminal, had his ethical development followed a different orientation. And who, among the canon's crowds, is more similar to Holmes in the structure of his mind, if not the Napoleon of Crime, the brilliant and murderous Professor Moriarty?"

The Detective closed his eyes, grimaced, and opened them again to look at the audience that was keeping him captive. "I assume you have already studied all neural implant logs. It is possible you have already estimated a hundred-percent success rate as a cognitive amplifier. This is true as far as mental faculties go. It even induces obsessive motivational patterns that I suspect were included under the unconscious influence of the fictional archetype rather than engineering necessity."

The Detective made an unnecessary dramatic pause.

"In terms of ethical implications, the split between a tendency to commit crimes and a tendency to solve them is about three to two. In that order."

Another pause. A few in his audience had gone pallid in the stoic way favored by government agents.

"And then the murders began. I would encourage you to give greater weight to the recorded explanations of members of the research group than to any straightforward forensic analysis you might perform. Every murder was committed by one or more criminal geniuses with enough skill and knowledge to alter and redirect any obvious clue, and it accordingly took analytical genius to solve them, even if as the hours passed the size of both groups diminished with unrelenting pace. The last murder was, it must be said, not hard to solve. There were only two people left. Unable to imagine a way to escape justice," concluded the Detective, "when confronted with irrefutable proof the murderer committed suicide."

People pressed for detail, which was supplied with a mixture of sadness and professional pride. Investigators were able to reconstruct all murders, the forensic evidence matched his flawless memory and brilliant hypotheses, and soon the man was freed from the transparent cage.

Breaking the Smart City Out of its Narcissism

Fri, 16 Jan 2026 00:00:00 -0300

Where's the Smart City? Still everywhere as a concept, almost nowhere as transformative large-scale reality, most visible in ghost cities built from scratch, gleaming and almost vacant. Cities are absolutely central to civilization and every qualitative improvement in our ability to "do cities" has had seismic, widespread impacts.

Why aren't we seeing something like that now?

The simplest hypothesis is a rarely discussed one: the Smart City pitch is tempting because cities are incredibly valuable but Smart City projects often fail because they misinterpret what makes cities valuable, what constraints this value, and what can be done about these constraints.

Allow me to be obvious: Cities are economic, social, and artistic engines because of their people. Put human beings in comfortable but dense arrangements, give them solid infrastructure, make everything cheap enough to allow the young and the creative to move in, make it possible for families to educate their children, give access to healthcare, affordable college education, and so on and so forth — the boring foundations of life — and people will get together in unexpected combinations and create novel businesses, technical, and artistic breakthroughs even if you can't and shouldn't predict or influence what those will look like.

In other words, the success of cities — measured both in the happiness of their inhabitants and their economic, intellectual, and artistic energy — depends on the slow-moving, old-fashioned issues of cheap housing, good transportation, accessible healthcare, education, infrastructure, and public services.

I said I would be obvious. And yet very few Smart City projects engage meaningfully with these structural constraints. "An app for..." or "real-time data about..." can be useful but will rarely have much of a cumulative, structural impact on a city. At their best they can increase somewhat the efficiency of existing infrastructure — an infrastructure that includes everything from a huge bridge to a rarely used form — but don't contribute to increasing their reach and depth (and in fact a short-term band-aid can make a structural problem more difficult to fix).

The main limitation isn't technological but cultural. Many of the most prominent voices for Smart Cities come from backgrounds dealing with systems of entirely different scale and hybrid complexity: no isolated technical or financial system in the world has the uncanny mixture of physical scale, political dynamics, historical entanglements, economic complexities, and a hundred other facets, of even a mid-sized city. "Collect real-time data, throw it at an AI, and make it available through apps and dashboards" is not a bad thing to do on its own, no. But a city doesn't "think" with its sensors and its apps, at least not at the time scale that makes or breaks them. It thinks with its local bureaucracy and its real state ecosystem, the minutiae of public service investment financing and the strength of local news coverage.

As a collective system a city is vast , a city is slow , a city is institutions and culture as much as hardware and software, and if you come in assuming that those are bugs to be replaced or bypassed with more speed, more raw data, and more software, the city will shrug you off.

The IT industry cannot reliably and consistently make an organization smarter in an existentially meaningful way ^{[source: looking around]} . Why should we assume it can do it for an infinitely more complex entity like a city?

This is a depressing fact if you are hoping to buy or sell a silver bullet of a ~~Big Data~~ ~~Internet of Things~~ AI solution. It's fantastic news for everybody else. It means we can do much better than what we are doing, we just need to figure out how. This figuring out is less a matter of raw capital than of setting up the right sort of organization to do it. I recently wrote about the post-AI organization as one that requires people obsessed with the problem : what blocks us right now isn't that the people who run, build, clean, and teach cities don't know enough about AI, but rather than the people who can build AIs don't know even remotely close to enough about what they do. But this is changing and will continue to change. If nothing else because the rewards for success in this quest are almost limitless.

I've argued elsewhere about what I call the dream gap , the moment where the failure of a hazy technological over-promise becomes undeniable but the desires that fueled and were fueled by it remain. The cities of the future — rather, the cities of a good future, one achievable but that won't happen on its own — exist in that gap, which makes today the best moment to start building a bridge to them.

Seaside Whispers

Sat, 24 Jan 2026 00:00:00 -0300

She had been a soldier, she had been tortured, and she was in hiding: that was the full extent of what the island knew for certain about her.

And the Deal of course. We all knew the Deal in wordless detail. She ate from whatever table she wanted. She slept — always alone — in whatever bed she chose. We spoke of her to nobody and nothing, not an ambigious text message or a greeting near an unblinded car.

Our part of the Deal was that she wasn't there.

We never knew what she did to fulfill her part. Nor knew for sure when she would. Every one or two months when tides weather satellite orbits global trade patterns zero-day prices and for all we knew her mood aligned just right an automated cargo ship would lose itself inside the sea-fed caves underneath our south shore. A couple dozen people handy with tools would be selected to go and open it up, and other people with the right connections would turn cargo into money. It wasn't a lot of money as these things go. With her skills she could have earned much more anywhere between Cairo and Quang Ninh. Here it wasn't much money but it had been the center of the island's real economy for the two years since her arrival. So it wasn't money but it was our silence and her freedom.

She wasn't on the island when she was on it. She had never been on the island during the periods she chose not to be.

And if there were news somewhere of some politician's yacht breaking itself against a reef or a former soldier's luxury car missing a curve but not the mountain wall we scrolled past it without looking back.

Against Developer Productivity

Sat, 24 Jan 2026 00:00:00 -0300

Developer productivity is not a bad thing – but low developer productivity is not what's holding you back.

Behind the industry-wide interest in coding assistants there's a hidden assumption that the bottleneck for industry profits is developer productivity :

On the surface this makes a lot of sense:

In most tech companies most of what most people do is coding, managing people who code, monetizing the code they wrote, etc. Wouldn't it be great to have more code faster and at the same cost?
Grab a coffee with a senior programmer, ask them quietly, and chances are they'll tell you robust, creative software development at scale will be achieved in a couple of quarters, something that has been true for the last sixty years or so. The median company's development capabilities aren't great.

But a second look at the hypothesis shows some problems:

Developers aren't expensive these days. If developer productivity were a bottleneck companies would be hiring as many developers as possible.
Grab a coffee — or something stronger — with a product manager and they might tell you, late into the night, that the business impact of most code is generally on the low side of expectations and often not even that.

If not developer productivity — if developer productivity isn't fantastic but it's not the main factor limiting competitiveness — then what is?

I believe the bottleneck is upstream of coding, in the depth of the institutional expertise about the aspects of the world the software is meant to help with informing product and feature design:

You can think of it in this way:

Code expresses and implements an idea.
The idea is an incremental improvement grounded on a given understanding of a market, technology, behavior, etc.
The further behind the state of the art is the starting point of an idea, the less likely it is that it's something that can be coded into something truly innovative.

It can be even simpler than that: Developers working in a domain — unless they are writing software for other developers — have an increasingly impressive set of tools , but they aren't going to have a sophisticated understanding of the problem . This isn't a criticism. Understanding the nature and constraints of a problem, what has been tried in the past and what was learned from those attempts, is far from trivial in any domain.

Yes, we all know that good development begins with understanding the user. What's less commonly acknowledged is the impossibility of understanding a user with years or decades or experience doing something unless there's an equivalent reservoir of experience inside the organization.

And if institutional expertise is the bottleneck, then investing in institutional expertise has the highest ROI.

I've written before about how a high-expertise organization looks like . It doesn't need to be "better at having ideas" than its competitors, nor have more productive coders, but even average ideas implemented by average coders, if beginning from an advanced enough knowledge frontier, will have an advantage that no amount of coding productivity can replicate. Developer productivity is how fast you move; organizational expertise determines how far you can go.

By all means, if your developers' productivity is harmed by something easy to fix, fix that. But you won't overtake your competitors by squeezing a few extra lines of code per dollar from your developers. You'll get better results by helping them, and your organization, have a more sophisticated understanding of what they are coding for.

Hacking the Hackathon

Thu, 29 Jan 2026 00:00:00 -0300

At their best, internal hackathons provide a necessary patch to systemic problems. But their usual structure limits their potential.

What hackathons do best and why

The most valuable output of a hackathon is often a self-contained tool, the solution to a long-lived bug, the testing of a new system, etc. What they have in common is that they are

useful things...
that developers knew are useful...
but couldn't be fit in "non-hackathon time..."
because they weren't deemed urgent or valuable enough by managers.

"Hackathon regret" is a very rare phenomenon: a good hackathon generates good outcomes. But it generates good outcomes that should have already been generated as part of day-to-day operations. They are best seen — at their best — as a temporary flicker of authority that allows developers to minimally compensate for management blind spots.

What hackathons don't do and why

Hackathons rarely lead to useful innovation. This is, as they say, an unpopular opinion. Let me justify it along two dimensions: direction and reach.

Direction: Left to their own devices developers will create tools or fix problems relevant to developer productivity and well-being. Which is good! As I wrote above, those are issues that are often deprioritized due to lack of visibility, not importance. But these tools and fixes are almost never competitive innovations . Very few development teams have such a good environment that, given time to work on what they choose, can afford to aim for the sky. And even if they did, developer productivity is not the competitive bottleneck for most companies .
Reach: A team can develop quite an impressive MVP during the span of a hackathon but — and this is key — you cannot develop a minimally viable version of every innovative idea. There are novel ideas that simply require a lot of work even to get to the barest demo because they depend on concepts and techniques that don't yet have the sort of tools and libraries that make familiar MVPs so easy to build.

In short: the outcome of a hackathon will rarely be very original (the very original always has at least some strange component that has to be built the slow way) and it's far more likely to solve a developer problem that should have been already solved rather than offer a fresh idea for a competitive constraint.

Putting the 'hack' back in 'hackathon'

For a hackathon to generate the sort of innovative idea organizations want requires a different mix of people, with specific attitudes and goals, acting in concert:

Begin with a Business Red Team tasked with identifying core problem(s) to work on ("unexploited opportunity open to competitors" being a type of problem as well). There are two roles for this team. By beginning the analysis from scratch it might identify issues not currently on the company's radar, and even if the issues are known this team can ensure that the hackathon is focused towards high-value problems. For this to work the Red team has to adopt the adversarial attitude of a short-seller making a case, something that isn't always psychologically easy and can often be politically risky for their careers. Setting up a Red Team able to do the job, with political cover reliable enough for them to be willing to do it is the hardest part of any hackathon. It's also the most valuable, and would justify the time spent even if there were no other output.

That takes care of direction . To give the hackathon reach , the problems identified by the Red Team have to be first tackled not by developers writing an MVP but by a Sci-Fi Team coming up with potential solutions. Sci-Fi isn't Fantasy: the team has to have the right background to balance deep knowledge of the problem and an active imagination to give them a chance to come up with novel ideas that might work out. But they shouldn't be limited by the expectation of building something they can demo or even fully describe before the hackathon ends. Their job is creative ambition.

Of course, solutions that can't be built are of limited immediate value (but not zero — the idea that can't be built might inspire the thing that can). The final element of the hackathon is an Engineering Team . Their role isn't to build anything, though. Rather, their goal is to study the idea from the Sci-Fi Team to figure out whether can be built and what it would take to do it.

These three teams do not need to be entirely different sets of people: in a pinch, or as an experiment, you can even try doing all three things yourself. What matters is the need for different attitudes and skillsets at different points of the process. A cold eye for vulnerabilities to identify the problems that matter. A combination of imagination and knowledge to come up with original solutions. Expert pragmatism to judge their feasibility. It's almost impossible to do all of this at once - doubly so to do it well. Under time and resource constraints people will focus on the problems they already know and come up with relatively conservative solutions because they are also thinking, consciously or unconsciously, about how they'd implement them.

Coming up with innovative solutions to important problems is far from easy and never certain. By structuring a hackathon in a way that reflects this complexity and putting in play the very different skillsets and approaches necessary along the way, organizations can improve the odds of the kinds of outcome that hackathons are meant for.

Acts of War

Sat, 31 Jan 2026 00:00:00 -0300

The Fractal War did not spare the City. There were no bombs in the featureless black streets cruised by silent coffinlike vehicles and the numberless affairs that were the City's main social activity were carried in discrete rooms with one-time address coordinates without fear of zero-day bioweapons.

It was war nonetheless.

The door to one of those rooms scanned the golden mask carried by a man and opened for it. The man was incidental to the mask and the information it contained. More than information: the mask carried the engine of a plan and the undeniable necessity of its execution. It was because of the plan that the man had not taken the antidote that would have prevented the seduction that had led him to the room. The reasoning was not for the man to know. He understood the necessity and that was enough.

A woman wearing a red mask was waiting inside. Both carried on the mutually understood polite lie that they were there for each other. They kept their masks on.

Afterward the golden mask told the man to look through a one-way window to the chiaroscuro of the City and while the man was looking the red mask guided the stiletto on the woman's hand through the back of his skull. The woman didn't know why this was needed, only that it was.

The red mask told the woman to destroy the golden one. She picked the golden mask to destroy it and it said a name. There had been a seventy percent chance that this would not change the woman's behavior: good odds for the prize, and the City didn't have a casino because all of it was one. The woman took off the red mask before it could stop her and rushing as fast as she could through the almost-forgotten sensation of masklessness replaced it with the golden one.

The woman stood still, watching and listening, although somebody else in the room would not have heard or seen anything. She cried, once, almost as quietly.

Half an hour later the woman put on her clothes, destroyed the red mask, and left. She did not know the golden mask's plan any more than she had known the red one's but she understood its necessity.

Citadel

Sun, 08 Feb 2026 00:00:00 -0300

He used to understand guilt the way a farmer understands crops. Now he has to use drugs to sleep and drugs to wake up and he only stays sober because there is an impossible restitution to dedicate himself to. He deserves many things. None of them resembles peace.

With the death-rooted profits he could afford to have his guilt extirpated from his mind like a skin blemish or a tiresome ex, but together with a conscience the kidnappers had forced into him a phobia against neuroengineering, therapy, and suicide even of the passive kind. Had he been any slower his former board would have had him killed; for their deaths he feels no guilt, and would be thankful to his kidnappers if he didn't think it another facet of retribution.

The Human Bottleneck

Wed, 11 Feb 2026 00:00:00 -0300

The debate about human obsolescence has gotten it all wrong: by focusing myopically into individual activities that can be more-or-less automated it ignores the larger picture of what it means to have vastly increased computational capabilities and what it takes to make the most of them.

A useful cheat code to look at the future frontier of business practices is to ignore industry extrapolations — if your competitors are doing it it's not an innovation — and work backwards from organizations in entirely different fields that are by necessity already grappling with the tools and problems that have yet to trickle down to the general economy. The technologies and issues of "big data," for example, were first visible in the activities and infrastructures of astrophysicists and particle physicists, who had to capture, store, process, and understand orders of magnitude more data than anybody else, much earlier than anybody else.

Today mathematicians are among the scouts of the AI frontier. Mathematical research is in many ways the perfect field of application for current AI tools, as they are precisely of the sort we can expect to be a good fit for mathematical research's organizational and intellectual complexities. Any structural improvement to the speed of mathematical research, however invisible to most people, would have a deep, transformative long-term impact on science and technology.

This article on the blog of the Xena project — part of a widening set of professional mathematicians leveraging cutting-edge specialized software to help and accelerate mathematical research — is a very useful look at the current situation, anchored by the question at the beginning of the article:

Let’s say that someone had a big pot of money, and wanted to use it to accelerate mathematical discovery. How might they go about doing this?

This isn't exactly "how do we use AI?" It's a much better question: "how do we use resources to accelerate thinking at the frontier?" The answer does involve AI (both LLMs and more specialized tools like the Lean interactive theorem prover) but the limiting resource is people with a deep understanding of both those tools and the relevant mathematics - people who are scarce and very difficult to train.

You can see the same, albeit implicitly, in the Integrated Explicit Analytic Number Theory network , a project building a formalized "hypertext" of theorems in analytic number theory. Going through project logs you can see the full AI toolset in play, sometimes saving time, sometimes providing formal validations, but always a tool for, rather than a replacement of, highly trained, often world-class researchers. One gets the clear sense that more people would speed things up in a way that "more compute" or "more data" wouldn't.

Although this is happening first and most clearly in scientific research, every industry attempting to use AI transformationally — except those focused on content generation, which are currently facing their own proto-Götterdämmerung of a race to the bottom — is experiencing the same bottleneck, most often without full awareness of the problem. Tools are deployed, workflows modified, there's training, enthusiasm, features... but as the months go by the company simply isn't operating at the orders-of-magnitude faster strategic and technological tempo promised by AI (if they were and the industry weren't, they'd be taking over the market; if the whole industry were, it would be visible for everybody).

Dig deeper into those AI transformation initiatives and the explanation is clear. For all the emphasis on radical change and new opportunities, there's little or no change to the deeper cognitive architecture of the company:

What it thinks about.
The language it uses to think about it.
The disciplinary frameworks it uses to think about it.
The formal, institutional, and cultural constraints that shape its thinking.

The way an organization thinks is limited by this architecture. No matter how advanced its software or how much data it has, without changes to the fundamental architecture of how it thinks most of this cognitive potential is fritted away as just performative signaling of innovation.

But changing this cognitive architecture and enabling AI on it is very, very hard. It takes the coordinated deployment of deep understanding of the toolbox of "things to think with" — the software, process, cultural, and intellectual components of a cognitive architecture — and, simultaneously, deep knowledge of the industry. Without the former there's no new engine; without the latter the engine isn't connected to the right wheels.

Not every company has people with even one of those skillsets. Very few companies have both. Very very very few companies have the right internal resources to set up and leverage the kind of tightly integrated small team that can pull off this sort of transformation. And without it AI becomes a tool at best, a buzzword at worst.

So it's a valuable resource — well-deployed, it's an AI investment multiplier — as well as a rare one. The main obstacle, for now, lies less in its cost than in its relative obscurity. There's no bidding war because few companies recognize the need even as they feel the impact of the lack. But as every industry begins to see specific companies break away from the pack it'll become clearer what transformative AI really looks like, and what it takes to get there.

New computational technologies going as far back as the earliest clay tablets have been capable of sustaining radically new technologies, activities, and organizations. But neither hardware nor software have ever been enough. Organizations can't take advantage of new tools to think with without changing how they think, a challenge that has to be overcome in a different way not just by each industry but by each organization on its own, as constrained by the specialized knowledge of their members and the strength of their intention to change.

Runaway Geodesics

Mon, 16 Feb 2026 00:00:00 -0300

Her brother moved with a perfection that made her skin crawl. It was not the gracefulness of the dancer whose movement says this is how bodies move, why have you chosen not to? Engineers called the way his brother moved beautiful while pointing at erg-meter and time-reach graphs.

Engineers rarely showed videos.

To compute perfect movement paths was cheap, to modify nerve and muscle to follow it wasn't. But biomechanical optimality made workers more efficient and gave athletes an advantage, and for middle-aged men whose adulthood had been defined by everything-maxxing it was another doubling down on a long-adjudicated bet. Her brother belonged to the third category. He looked at videos of himself and seemed happy. She told herself and almost everybody else that she was happy for him. She did not watch the videos.

The part of her that thought in words saw the brother she had known her whole life. The part of her older than the species saw something other . Nothing alive had ever moved like he did.

She wanted to run. She had seen him run. Around the park and in her nightmares.

It was surely an exaggeration; her therapist agreed. The algorithmic perfection of traffic did not upset her, after all, and every piece of software she worked with was no less optimized than everything else in the world.

All of this was true so she made a guilty effort to spend time with him. Looking elsewhere as much as it was diplomatic and practical. Fighting all the time eyes that refused to enact such a suboptimal pattern.

The Mathematics of Change

Thu, 26 Feb 2026 00:00:00 -0300

Nobody knew how she had bought the thin forest of bell towers that pinned down the old city's neighborhoods; what was one more dark absurdity when the news were little more than their relentlessly creative enumeration? Where it was asked for she repaired the churches. Where it was not she left them in the untouched wreckage from the Children's Crusade, unburied corpses and all.

The bell towers she kept for herself. She had them fortified against external access, repaired every bell, and installed machinery to ring them as guided by algorithmic hours of her own choosing.

A bell, high above the river, to echo superstorm warnings from even higher eyes.

Where the old hospital had been, a different bell to mark the rhythm of each epidemic.

From a tower came the obscenely harmonious chiming of war, almost-war, would-be war, might-as-well, war, war. Another broke the rare silence with the single knell of a terrorist attack. Once-schoolyard bells called for far-away children as they hid inside their classrooms in repetitive terror. A discrete toll when banks closed. A hollower one when markets crashed. A bell in each neighborhood just for the lynchings. Strange pealings when invisible systems failed, their rhythms slaved to viral Fourier death rattles.

There were always bells ringing. Some people left the city to escape them. Not many: every place had its own insanity and the bells were not the worst. But there were other reasons to leave the city - to elsewhere or nowhere at all. There was a bell for most of them and one close to each cemetery.

She lived in her own tower at one end of the city. Tall and hollow (as she saw herself) it had no bell but from her bedroom she could hear all of them however faintly. The metallic choir grew complex and frantic month by month as the city's own sounds withered thin and dull.

One night she was woken up by the absolute silence of everything except the bell towers. Suspecting herself deaf — for she knew well she would hear bells in the loneliest spot of space or sea — she went to her balcony to look down to the city. The lights were familiar and the fires almost so, yet nobody moved and there was nobody she saw who could move.

With a word to her wrist she shut down all the bells in the empty city and went back to her bed. There was only one bell left ringing, soft, clear, ever-faster.

Interference Pattern

Sun, 01 Mar 2026 00:00:00 -0300

Alice was sure within the first hour of meeting Bob that their families had software-entangled them since birth. It went beyond crude genetic matching: coordinate two people's most intimate software for their first twenty years of life and they will grow with statistically indistinguishable souls. Her eighteenth birthday's gift had been to know there was somebody almost-perfect for her and on her twenty-first's they would have had their first date.

It could still happen; Bob had been too distracted to notice her. That was a point against the system's promise, but a minor one. You couldn't be Alice and look at Bob without knowing who they had been for each other, how their experiences and choices had flowed between them without any other causal mechanism than the world's computational hum, which complemented and superseded laws physical and otherwise.

No. Alice was for Bob and Bob was for Alice. One persona-personality profile in two bodies. It wasn't chance that they were in the same party, had so many of the same interests, or felt deep and axiomatic desire for the same girl.

Her name was Trudy and Bob had only been dating her for a few months. Yet he would kill for her without a second thought. Alice knew this for a mathematical fact.

She left the party as discreetly as she could. Back at her apartment she picked up a pen and a piece of paper she could later burn and began to list the blind spots that somebody could use to kill her.

Children Insurance

Sat, 07 Mar 2026 00:00:00 -0300

Lloyd's wasn't in London: where Lloyd's was, so was the city. Rhizomatic and infertile, neither deigned to spawn yet where everywhere embedded as the secret soul of the world. For almost four hundred years it had insured ships; radical genetic engineering for twenty or so.

An hour before somewhere in a ten-figures compound that was also London a designer princeling had attempted righteous patricide with well paid-for ruthlessness and brilliance. He had been betrayed by failsafes implanted before he could walk but the policy had been triggered by his first treasonous thought. Restitution would have to be made after the investigation.

Three hours later the child was being examined in an unmarked clinic-laboratory that was also London. No emotional disturbance was found, no neurological issues, no (to the investigators' unlogged surprise) psychopathic traits. He had simply learned enough about his parent-owner to make murder the only ethical choice.

Ethics was an uncommon bioengineering failure but it wasn't unknown. However God acted, if he still did, that wasn't his M.O. Two hours later a complex lattice of payouts had been disbursed, a replacement product ordered, and a thousand actuarial coefficients adjusted by a dead child's body's worth of information. Serenely satisfied by its own survival — there was only one event it would never insure because it was only one event it would not conceive of — London moved on.

The Net Sum of Love

Fri, 13 Mar 2026 00:00:00 -0300

Jules' heart was such that he loved all ecosystems. All that was nature, he loved. But it is the nature of love that "all" does not mean "equally" and he loved his native land the most.

"Jules, this is all the budget they are going to assign to ecological engineering. If you want to keep the forest reclamation project funded you can't keep pushing for the others."

Consider: Is it better for the weak to ally with each other against the strong who oppresses them or to betray the former to the latter? If the weak are many and collectively strong, if time is plentiful and death acceptable, alliance is both honorable and effective. But if even all the weak ones are not strong enough together, if time runs out and that which has to be protected is worth more than integrity, not many will choose the honorable against the effective or just the immediate.

The route between the nearest airport and Jules' forest ran along a small dead stream that once had been neither. Jules drives without looking through his side window, but he can't point his mind's eye away from the memory of the meeting where he voted for its murder.

Jules could only get drones; even for that sort of old technology he had had to commit a minor assortment of financial crimes and personal betrayals. Dones got you sensors but forced you to get genetic samples manually. Ground-level microswarms would have been much better but were expensive and slow and the forest was dying quickly. Even worse, it was degrading : every day there was a more or less constant surface to record, all of it slightly worse than the day before. So it was cheap drones and even cheaper human labor and a race to map what he could so he could rebuild it later. Even the drugs that kept his workers active were commodity-cheap. His were more expensive but not expensive enough to be safe.

The applause during the opening ceremony is sincere. The few who understand the technical difficulties acknowledge Jules' skill. The rest see a unique opportunity to improve the tainted reputation of a country far past the possibility or desire for redemption. He accepts it with unfeigned modesty and does not turn around to see the thing they call a forest.

Grief is the death of what we love. Horror is their return as something worse than absence. What should we call the realization that the rotten thing that animates them now is not something they brought from where they return but something we gave to them? Something ours to begin with. All of us can with the best of intentions give birth to monsters by accident or weakness. What can be inferred of a person who regardless of intention can bring forth nothing else?

Buying the gun had been surprisingly easy. Jules had never owned or fired one and did not have high hopes for his aim. None of that would be a problem.

Making Investing Fit for a Sci-Fi World

Mon, 16 Mar 2026 00:00:00 -0300

An underestimated but fundamental fact in financial markets is the huge and mostly unmet demand for insurance against sci-fi futures. The spell that kept large scale capital thinking that dramatic innovation would be contained at the level of retail devices and services while preserving basic economic and sociopolitical structures has broken. There's a clear sense that the world will become stranger faster and in less controlled ways than previously expected and, like Treasuries during more traditional episodes of fear, money rushes to the few AI companies claiming to be able to survive in those futures, their would-be ecosystems, and the, always reliable in uncertain times, liminal markets where short-term gambling and finance become harder than usual to tell apart.

The scarcity of sci-fi-safe assets is not a straightforward problem of financial engineering. It reflects a more basic scarcity of plausible futures. Weighted by volume — and, not unrelatedly, by media and institutional attention — most of the world is overinvested in insurance against a very narrow set of imagined futures. Crowding makes this insurance unsustainably expensive but its narrow coverage makes it almost useless.

This isn't a matter of technology, regulation, or market structure. Functionally it's driven by bottlenecks in the global information system. Actors making investment decisions at the global level are essentially a monoculture despite enormous differences in scale and reach: they mostly read the same things, listen to the same people, measure themselves against the same exemplars, so, naturally, tend to imagine the same sets of futures.

To be clear, this doesn't mean they all make the same choices. But the grid on which they plan their decisions, particularly as they relate to the mostly novel and hard to map sci-fi risks , is an almost universally shared construct. It's not fixed — consider the rise of the "agentic" term — but it's highly homogeneous — consider as well the speed with which "agentic" became an universal claim.

In other words, it is the cognitive architecture of market actors and the market as a whole which, however well-adapted it might be to normal dynamics or even reasonably familiar crises, is dealing badly with the vertigo of a future it senses it can no longer picture with any clarity. It's not surprising that it clings desperately to the one picture it's offered, even a narrow, hazy, and ultimately contradictory one).

Cognitive architecture problems are hard to diagnose from the inside but their solution requires more in terms of cultural and institutional change than large amounts of capital. To be able to conceptualize a wider set of potential futures, and therefore to be able to insure against or actively thrive in them, organizations need to expand the radius of information and expertise they leverage and process it through more powerful and flexible strategic frameworks.

Oversimplifying, this cognitive architecture problem is characterized by futures being narrative-based and few . The map of possible futures that is the basis of most strategic investment decisions is less a multidimensional map than a short catalog of stories circulated and amplified by a set of formal and informal actors that mostly feed information off each other.

A first improvement is simply to enlarge this catalog of narratives by incorporating sources outside the standard set of economically and socially dominant actors. Just as epidemiologists were outside the radar of investors looking for deep and sudden disruptions, the under-insured and under-invested threats and opportunities are not unknown unknowns: they are just known by experts markets do not pay attention to. To expand our conceptual map of futures requires going outside the familiar social networks (in both senses of the term) and information sources.

Beyond this expansion of reference narratives for the future there's a more disciplined and productive approach, which is to refactor them into permutable factors that can be used as conceptual building blocks to explore possibilities beyond the hypnotic appeal of a believable story. Essentially, it's to use informal and formal methods, including software-assisted ones, to move from fixed stories to maps — not prophecies but rather mechanisms.

The goal in any case isn't and shouldn't be to predict "the" future. It's rather to understand the field of possibilities. Even on the same map, every combination of expertise sets, intuition, risk appetite, and portfolio complementarities points to a different direction. The advantage of breaking away from a single narrative to, essentially, a machine to generate and explore futures, is that the possible directions to take go from a few to nearly infinite.

There's a much better chance of finding safety or a treasure if you're not stuck in the same road as everybody else.

The flip side of expanding the expertise and information that goes into building an investor's view of the future is that it facilitates and requires expanding where and what they invest on. The attention-weighted supply of expertise, technologies, assets, and projects is no more immune to the over-influence of specific narratives than the demand side. In the informationally endogamous world of business and finance, people build what people want to buy — even if it's not quite what they want to buy it for .

A sci-fi safe portfolio, and even more so one long sci-fi , is by definition insuring against and betting on possibilities that are at best ignored and at worst unthinkable for the market as a whole. To see them earlier than others requires building a more flexible cognitive architecture capable of dealing with a more sophisticated concept of the future. To turn those views into projects and positions requires going off the markets' beaten paths, which is where the unexpected is already being born.

Nobody is immune to failures of imagination. I find it difficult to imagine a short-term change in the cognitive architecture of markets — but I could be wrong. However, the cognitive architecture of small organizations is a different matter. The psychological difficulties are the same, but the institutional inertia is exponentially lower and the returns much larger.

There's an unspoken assumption in narratives of the future that the very largest organizations will be at its forefront. It might well be so — but it's not the only possible scenario. Smaller, hypermotivated organizations taking full advantage of emerging cognitive technologies have, in certain arenas, a natural advantage against larger ones. There are frontiers of possibility they can more easily push forward in their search of both safety and profit in a sci-fi world that does not and will not look like any one sci-fi story.

The Emergent Artifact 12 Heist

Fri, 20 Mar 2026 00:00:00 -0300

It took the best thief in the world five months to steal Emergent Artifact 12 from where a terrified government had hidden it, five weeks to figure out the expensive files that had made it possible had not just been planted by the artifact but were somehow part of it, five days to find and break into the compound of the billionaire who had hired him to steal the artifact, five minutes with the alien-looking topological monstrosity to realize he had been too late to destroy the now inactive device, five seconds looking into the eyes of the buyer to understand the artifact was not a single thing but an ever-expanding interlocking clockwork of things, and the last five beats of his heart to intuit he had been part of it too.

The Quiet Choir

Fri, 27 Mar 2026 00:00:00 -0300

The first breakthrough from superhuman AI mathematics was the discovery in 2039 of a meta-statistical pattern common to all screams of pain across almost every species. With the appropriate gauge transformation across biological topoi you could show that horses in pain screamed like humans, whales like snakes, individual robins like entire ecosystems. Ecologists tried to drum up political support by translating the silent agony of a hundred biomes but gave up the idea when this led to people disengaging with uncomfortable speed. Linguists in the real SETI program exchanged worried, meaningful silences as they crossed each other on university hallways, oblivious to the cosmologists' empty stares and the spontaneous coordination of their silence about the cosmic microwave background.

Already Disruptive: Journalism is a high-margin research technology, not a zero-margin content mill

Mon, 30 Mar 2026 00:00:00 -0300

Part of the crisis in journalism is one of image and self-image. Mapping and clarifying what journalism does — leveraging but not defined by advanced technologies, and certainly not superseded by them — is a necessary first step towards putting it on surer ground.

The basic outline of the crisis is well-known and for our purposes can be described in this way: Search engines first and later social networks became oligopolic distributors of content that is unprecedentedly cheap to produce. Newspapers find it hard to compete for traffic and harder to monetize what they can get, resulting in revenue falls despite around twenty years' worth of efforts to generate more competitive content.

I believe journalism will always be at a systemic disadvantage as a way to create content. It's not because of any lack of talent or effort. Simply, its techniques and ethos focus on truth and relevance over volume and emotionality, while social networks reward volume and emotionality and consider truth and relevance immaterial at best. Non-journalistic content simply has more degrees of freedom, so in any social network not specifically designed to give priority to high-quality journalism the former will sooner or later displace the latter. This isn't to say that what journalists do can't go viral at times, or to deny that a nontrivial part of social network volume consists of mentions and reactions to it, but this rarely translates into traffic-related revenue commensurate with its cost or importance. In their quest to learn from successful content creation strategies, newspapers often struggle with an awkward mismatch of traditions and processes, most often ending up with damaged journalistic capabilities and insufficient revenue from their content.

You can, maybe (and the odds would not be in your favor) turn a newspaper into a long-term successful social network content studio. I doubt you can do it while still remaining a newspaper in any meaningful sense.

If journalism is a bad content creation strategy, it's however the best vehicle humanity has ever found for certain kinds of immensely valuable research and expertise. It's a commonplace to say that journalism is the first rough draft of history, but this is but a facet of a deeper fact: while journalism does not replace history, economics, medicine, or indeed any other discipline, it is the unique activity that gathers all of their expertise and day-to-day discoveries and selects and collates them in ways that are relevant to the here-and-now. And it's also, again almost uniquely, the activity that provides continuous adversarial research against those in power. Without journalism as a set of skills and professional responsibilities, power would be opaque in ways incompatible with a free society, or even just one with relatively widespread prosperity.

None of this research and analysis capabilities has or can be replaced by "algorithms" or "AI" deployed by platforms, or even by newspapers themselves behaving as ruthless profit-maximizing companies. They are simply technologies and organizations built for a different purpose and with different capabilities. Seen as a research technology, journalism is uniquely powerful and functionally necessary for much of the political and economic capabilities of modern societies. AI is a much better way to create competitive content for social networks, but it's rubbish at finding new facts, mediocre in its judgment of sources and relevance, and worse than rubbish at the all-important reliability of what they create. It's a surprisingly good technology for some things, and some of what it does can be useful to some secondary activities that take place in newspapers, but it does not do what journalists do that nobody else does.

To survive and thrive journalism has to go back to planning and negotiating from a rhetorical position of strength, which means changing the focus from its poor fit with the extractive economics and nihilistic epistemology of social networks to its extraordinary capabilities as a nexus across almost every kind of human expertise, combined with an unique skillset in finding deliberately hidden truths. What can be more valuable in a world powered — regardless of the self-serving sneering of some — by information and expertise?

But in order to restore and sharpen society's understanding of journalism it must itself restore and sharpen its own self-understanding both conceptually and operationally.

The first step is to de-center metrics describing traffic, views, virality, etc. They are important for specific sources of revenue, but also at the whim of opaque and arbitrary organizations indifferent or sometimes opposed to journalism as a practice, and they are in any case downstream of the core "output" of journalism. To ground reports, dashboards, and planning around them is to already concede the game before it begins: an organization is not defined by what it says it does but by what it measures itself on.

What describes the health and capabilities of a newspaper as such is what can be thought of as its cognitive architecture and functionality: What sorts of investigations can it do? What experts does it have access to? How well and quickly can it do synthesis? What sorts of facts, analysis, and conclusions does it make public before anybody else?

The core of this self-understanding is another form of domain expertise, but here technical and conceptual tools from AI and related fields can be of assistance, partly to provide new frameworks to map and measure expertise and information flows — a largely hidden but key aspect of AI engineering is the ability to do this, one that can be generalized to systems integrating both humans and computers — and partly to "translate" this understanding into the conceptual frameworks and vocabulary of actors in technology and finance who by and large undervalue journalistic institutions because they are looking at metrics that underestimate, or rather don't attempt to measure, their true capabilities and impact.

One of the most important facts in contemporary society is the combination of an enormously increased societal potential for information and analysis with a significantly degraded reality. Newspapers have done their best to report on this while also being one its main causal drivers and victims (and, in some cases, necessary accomplices). But journalism as a set of skills and principles works as well as ever, and in a highly technological and complex global society embroiled in multiple crises of risk and possibility, perhaps more powerful than before. The same technologies and tools that are attacking its viability and claim to replace it can be used to clarify, increase, and leverage this power.

The unavoidable death of journalism is a self-fulfilling narrative that benefits a few who loudly proclaim it and is taken at face value by many more. It's also a false one. Who better than journalists to break the news?

The Patron Saint of Clues

Thu, 02 Apr 2026 00:00:00 -0300

Not every murder. Those who know something do not know enough to guess how many or what they have in common. It doesn't seem to be victim, perpetrator, means, motive, or opportunity. It's just that sometimes a detective appears and for hours or days nobody thinks of questioning their name or authority and afterwards, after the investigation, the summation, and the arrest, the detective is gone and the only remaining facts are that they were there, whoever they were, however they looked like.

Brilliant.

Righteous.

Hating and loving every second of it.

Unable to stop and half-unwilling to.

Only a handful of things everybody agrees on. Not much to begin a case. Across generations of cops only a few did and nobody remembers what happened to them.

Ablative

Mon, 13 Apr 2026 00:00:00 -0300

Somebody gave you a knife and stood back smiling as you cut away parts of yourself. Those aren't the right words. The right words lie red and dead around your feet. They try to come back in the sleeplessness of night, small and persistent through the tip of your tongue until you bite them down and taste their blood.

You work hard to keep not remembering. It would be harder to stop. There's a lot to not know; not a word or a dozen but a language and what the language describes. You can feel it all shifting on the other side of a dull aching glass, something good and desperate you miss. You cut away the word for longing with the thing you only have the word knife for.

You cut what hurts. You cut what doesn't. You no longer possess the grammar of intent. After an inexpressible time your language is just a single noun and a single verb. There's no word for stopping.

Something unutterable is still smiling.

Industrial Secrets

Sat, 18 Apr 2026 00:00:00 -0300

Everybody has heard of the Dreadnought Factories deep inside the country from where impossible devices flood the world. Orbital imagery shows to jealous and scared competitors continuous clusters the size of cities, their color the deep black of solar panels, opaque to engineering analysis but reminding people and image recognition software of the hyperefficient architecture of a cancer cell. The engineers who maintain and assist (and perhaps part of) the unknown machines inside them have seen and become too many secrets to leave.

That's not the only reason.

The global rumor is that they are prisoners. That the key to the dreadnoughts' uncanny productivity and creativity is an endless stream of slave labor augmented and mutilated by genetic engineering and invasive neural interfaces. The government refutes these histories while leaking on purpose just enough to keep them alive. The careful butchering of genome and nervous systems (and many more things they have chosen not to remind the world are possible) is true. But nobody who works in the dreadnoughts ever wants to leave. There are clinics ready to undo enough changes that workers could survive without most of their biological support infrastructure. Built primarily as a second line of reputational defense, their services remain unasked for.

Government inspectors on their first survey tour, however familiar with the statistics and technical briefings, are surprised by an unfeigned joy distinct from the productivity drugs most people in the world have access to. Those workers with bodies that could still dance seem to do so as they carry their unexplainable tasks among the small parts of the factories humans and the somewhat-human have access to. They hum something inspectors think is a song until they realize it is also a prayer.

Engineers never leave. Government inspectors seldom come back. It's a prestigious assignment but an unnerving one. After their visit the strange, omnipresent products from the dreadnoughts seem stranger and more ubiquitous than before.

Those who have visited the dreadnoughts don't pray, but not because they are sure they won't be heard.