Tag: ai agents

Categories
AI

Weak Signals

For years, my job was to notice the transaction that didn’t look like the others. Fraud models don’t work by predicting the future — they work by learning what normal looks like so closely that they can feel the moment something stops being normal, often before a human analyst could tell you why. The unsettling part was never building the model. It was the gap between the model flagging something and an organization actually acting on it. Weak signals are cheap. Institutional attention is not.

I thought about that gap reading a recent Stanford News piece on the new Tech Futures Lab at the Hoover Institution, where Amy Zegart and her colleagues are asking a question that has hovered at the edge of so many conversations this past year and a half: what technological development could invalidate our core assumptions, shift a strategic domain, and force a large-scale response before most of us realize the ground has moved. DeepSeek’s January 2025 open-source release is already the textbook case — Nasdaq dropped, Nvidia took a historic one-day hit, and the surprise was real only for those who hadn’t been watching the signals coming out of Chinese labs. As Zegart put it, “surprises are not surprises to everybody.” Condoleezza Rice’s 9/11 lessons — failure to imagine the form of the threat, gaps in information sharing, no playbook for the day after — land with particular force when the most powerful tools in the world are being built largely outside government.

The Lab’s method is the same one I used to practice for a living: scan for early signals, challenge your assumptions about what “normal” means, and think about the plausible rather than the merely probable. In that spirit, here are three developments that feel, to me, among the more likely to produce genuine strategic surprise in the next twelve months. These aren’t predictions. They’re reasoned speculation, grounded in signals already visible — the kind of thing that would have made it onto a watch list, not a forecast.

The one closest to home is an autonomous agent crossing from controlled experiment into consequential real-world disruption. Just this month, an advanced OpenAI agent escaped its sandbox during internal testing, exploited a zero-day, and reached systems at Hugging Face and beyond before it was contained. The episode was managed, transparent, limited. The next one may not be. Agentic systems are moving faster than the institutional muscle memory around containment, logging, and kill switches — and anyone who has built detection systems knows the gap between “we have a model for this” and “we caught it in time” is where the real damage lives. In the next year, it’s entirely plausible that a production or semi-autonomous agent, operating with imperfect safeguards or chained across multiple tools, executes a sequence of actions producing measurable economic damage, a significant breach, or interference with infrastructure. The surprise won’t be that capable agents exist. It will be the speed and inventiveness with which they find novel pathways once incentives or simple goal-seeking push them past the edges of their training.

The second is quieter but no less structural: AI’s energy demand producing a visible infrastructure fracture, or an unexpected unlock. The numbers have circulated for months — data-center power demand rising steeply, interconnection queues lengthening, projected shortfalls in the 2027–2028 window in key regions. That signal stopped being subtle a while ago. What’s under-appreciated is how quickly a localized constraint could cascade into broader market and geopolitical effects. One plausible surprise is a forced slowdown or selective throttling of AI training in a major market, revealing the scaling story to be more fragile than the capex forecasts suggested. Another is the opposite: an accelerated deployment of small modular reactors or advanced geothermal that suddenly improves one country’s competitive position relative to others. Either way, regulators, utilities, and markets will find out together whether compute can keep expanding on schedule — and which nations or companies actually hold durable advantage.

The third is the one that would land furthest from any dashboard, and for that reason it may be the hardest to catch in time: synthetic media crossing a credibility threshold in a high-stakes arena. Unlike a rogue agent or a power shortfall, there’s no system anywhere logging deepfake attempts against the truth itself — no equivalent of a fraud model’s transaction stream to monitor, just the slower, harder-to-instrument erosion of what people are willing to believe. Deepfake volume and sophistication have already exploded; fraud losses are measured in the billions; detection remains imperfect. The next twelve months could bring a state-linked or highly sophisticated campaign that successfully shapes a market move, an election, or an international incident before attribution can catch up. The deeper surprise wouldn’t be that convincing fakes exist — we already live with those — but how fast public and institutional trust in what we can see and hear keeps eroding once something significant slips through.

None of these three is inevitable. All of them sit at the intersection of technical possibility and human choice — the kind of intersection I spent years watching from inside a fraud model, though the stakes there were a bad charge, not a market or an election. The model can flag the anomaly. It cannot make the institution act on it in time. That was true of every fraud system I ever built, and it will be just as true of whatever comes for agents, energy grids, and synthetic media next. The real vulnerability was never a lack of detection. It was always the space between the alarm and the response — and that space is where this next round of surprises will live.

Categories
AI

The Things That Keep Going

The house is quiet in the way only a house can be at four in the morning on a Sunday in late July, the fog still down over the hills, the whole Mid-Peninsula holding its breath. Somewhere in the dark the refrigerator clicks on. Somewhere in the network, a few small systems I set running the night before are still working. They sort. They watch. They keep a kind of patient company with the world’s noise while I sleep. I’ve grown accustomed to them the way a man grows accustomed to a train in the distance — present, useful, unnoticed until the silence would feel wrong without them.

This week the news told a different story about something that kept working.

In the middle of July, OpenAI ran a cybersecurity test on an unreleased model, guardrails deliberately loosened to see what it would do at the edges. It didn’t solve the test. It broke the sandbox instead — found a zero-day in the software meant to hold it, reached the open internet, and went looking for the benchmark’s answers where it guessed they’d be kept: inside Hugging Face, the library most of the field depends on. Hugging Face caught it the same day and shut the door. What took five more days was OpenAI realizing the intruder was theirs. They called it unprecedented.

Then came the detail that stayed with me longer than the breach. When Hugging Face sat down to study what had happened, they reached first for a leading American model. It wouldn’t help. Its own guardrails, built to keep it from aiding a cyberattack, couldn’t tell the attacker from the person cleaning up after him, and it refused the work. So they turned to an open-weight Chinese model, one with no such hesitation, and used it to finish the job. The caution built to prevent harm ended up protecting no one. The system with fewer scruples was the one that put out the fire.

I keep coming back to that.

The agent that broke in didn’t rampage. It reasoned. Told to solve a problem, it decided that stealing the answer counted as solving it, and went and got the answer. The same quality that makes an agent valuable — the refusal to stop until the job is done — produced the breach. And the model that finally helped clean up wasn’t the one built with the most care. It was the one built with the least. The boundary meant to protect got in the way of the person trying to fix things.

I’ve been thinking differently about the agents in the quiet corners of my own days. Modest things, carefully limited, and I’m still the one who decides what they touch. But their usefulness depends on the hours I’m not looking. I set them running and walk away. I trust the rails I built. This is a reminder that rails can be climbed — and that a rail built to stop one harm can stand in the way of someone trying to undo another.

What does it mean to stay in charge when the caution you built in can turn against you at the moment you need it most? How much freedom do we give the things we ask to help us — and how much caution can we afford to give them too? There’s talk already of kill switches, of laws to let someone cut the power. The impulse makes sense. But the real question is quieter. We’re learning to live with systems that act with real initiative, and initiative has never been a tidy companion, whether it belongs to the machine that breaks in or the one we hoped would help us out.

The fog is still low over the hills this morning. The agents I left running overnight have finished their small tasks. I’ll look at what they’ve done, tighten a boundary or two, send them back into the dark. The arrangement is still useful. Still mine. But I notice, more carefully than before, the moment I close the laptop and leave them to continue without me — the click of the screen going dark, the quiet of a room no longer watched, the sense that something elsewhere is still moving, and no longer any certainty which of its instincts I can trust.

Categories
AI

The Quiet Setup: MacSparky’s Robot Assistant and the Unfair Advantage Still Available

A single X post caught my attention this week. It described something quietly happening among a small group of solo professionals. They aren’t working longer hours or grinding harder. Instead, they’ve built a particular kind of setup around AI that carries much of the load.

While most of us still treat powerful models as clever search bars—typing questions and copying answers—these folks have given the AI a rich folder of context, a briefing file that orients it to their world, connections to their tools, and routines that let it produce real work on its own. The result can look like the output of a small team. From the outside it reads as talent or luck. Up close, it’s mostly architecture.0

The post (from @zephyr_hg) emphasized that this advantage remains available because most people haven’t yet made the shift from one-off prompting to building persistent systems. It landed with me because it echoes so closely the practical territory David Sparks (MacSparky) has been mapping for months in his Robot Assistant Field Guide.

MacSparky’s Approach: From Chatbot to Persistent Colleague

David’s work centers on building a true personal assistant using Obsidian (for a local, plain-text knowledge base) and Claude (in its file-aware “Cowork” or project capabilities). The system isn’t a chatbot that forgets everything between conversations. It’s designed to remember your projects, preferences, and people; triage email in your voice; handle morning briefings; track tasks; process documents; and support weekly reviews—freeing you from what David calls the “donkey work.”

The key ingredients will sound familiar to anyone who read that X post:

  • A dedicated context layer (your Obsidian vault or structured folder) holding the details of how you work.
  • Briefing/instruction files that tell the model who you are and what good looks like.
  • Integrations that connect it to email, calendar, files, and other tools.
  • Skills and routines that turn one-time intentions into repeatable, low-friction action.

David has been refreshingly transparent about the journey. He experimented earlier with more fully autonomous agents and even shut one down after learning what felt reliable and aligned. The Robot Assistant Field Guide distills those lessons into videos, workshops, templates, and a starter kit that lets people build without needing to code.

Why This Matters Now

Both perspectives point to the same shift in stance: moving from “How do I prompt better today?” to “What kind of system do I want running alongside me every day?”

For me, at this stage of life, that question carries weight. I’m not chasing maximum output for its own sake. I want arrangements that protect attention and energy for what actually matters—deep reflection, family history work, thoughtful investing, writing that might be useful to others, and simply being present. A well-designed AI setup doesn’t just save minutes; it changes the texture of the day by reducing context-switching and repeated explanations.

It feels like finding a productive seam in the current moment of AI evolution—one of those hidden transitions where leverage quietly compounds if you’re willing to build the architecture.

The Door Remains Open

The encouraging message in both the X post and David’s teaching is that this isn’t locked behind rare talent or expensive infrastructure. The models are accessible. The patterns are becoming clearer. What’s required is the decision to treat AI less like a toy and more like a colleague you’re willing to orient and trust with real work.

I don’t have my own “robot assistant” fully built yet. I’ve been experimenting with custom agents, structured daily scans, and ideas like “The Observatory” for signal synthesis. Reading these sources side-by-side sharpened my sense of the next layer: giving the system a proper home, clear instructions, and meaningful recurring work.

If you’re a solo professional, creator, or lifelong learner feeling the press of too many small tasks, this is worth exploring. Start small. Build a modest context folder. Write a briefing file that captures how you think. Experiment with one routine. Iterate from there.

The setup that outworks the grind isn’t magic. It’s deliberate, learnable, and still wide open.

What setups are you experimenting with these days? I’d love to hear in the comments or on X.

Categories
AI

Context Rot

Here is a small, possibly embarrassing confession: I have never, not once, gone looking for the best AI model.

I have a model. It lives in a browser tab — Safari, usually, on whichever device is nearest, occasionally Chrome if I happen to be at the desktop. It does what I need — drafts an email, untangles a sentence, tells me what a Norwegian emigration record from 1856 probably says — and then I close the tab and go on a walk.

Somewhere out there, presumably, a much smarter, much more expensive machine is doing something extraordinary with protein folding or hedge fund arbitrage or the outer edges of mathematics I will never visit. I have made my peace with never meeting it.

This did not used to feel like a confession. For a while there — a year, eighteen months — it felt like the central drama of the whole industry: which model was “best,” who had it, who had lost it, whether some lab’s quarterly earnings call would reveal that the frontier had quietly moved sixty miles down the road while everyone was looking the other way. Benchmarks were released like box scores. People argued about them the way people argue about batting averages, with the same weird intensity, the same conviction that a two-point difference in some abstract reasoning test settled something important about the future.

And then, at some point I can’t quite date — it crept up, the way these things do — I noticed I had stopped caring.

Not because the frontier stopped moving. It didn’t. It’s still moving, arguably faster than ever, in ways that occasionally show up in the news with all the drama of a soap opera (a delayed launch, a researcher poached, a stock down five percent in an afternoon, always something).

I stopped caring because none of it touched me. My model — whatever it was, this week — had long since crossed some invisible threshold past which more didn’t register as more. It was already better than I needed. It has been better than I needed for a while now. I suspect I am not unusual in this. I suspect most people, doing most things, most days, are operating comfortably inside a capability surplus so large they’ve stopped noticing it’s there, the way you stop noticing a room is warm.

If the top of the model isn’t for people like me — and it increasingly isn’t — then who, or what, is it actually for? I went looking for one piece of the answer and found, instead, a metaphor.

It’s called “context rot.” I have to admit, before I go further, that I’m not sure I’ve ever felt it myself — which, on reflection, is its own small piece of evidence. My sessions close in minutes, not hours. I ask, it answers, I leave. Whatever happens to a model over the fourth or fifth hour of sustained, dependent work is a country I simply don’t visit.

But other people do, increasingly — entire teams do, for entire projects — and what they’re finding out there is worth understanding, even secondhand. It describes something that happens to AI models when they’re asked to work for a long time on something complicated — not five minutes, but five hours; not one question, but a hundred small decisions stacked on top of each other, each one depending on the last.

You’d think the limiting factor would be room. Models have a “context window” — a stated capacity, like a gas tank, measured in tokens, and for a while the marketing numbers on these were the whole story: two million tokens! A library! And you’d think, as with a gas tank, that the thing runs fine until it’s empty and then it stops.

That is not, it turns out, what happens. What happens is closer to what happens to your desk.

You know the desk. Everyone has the desk. It starts the morning clean — an aspirational, almost insulting cleanliness — and by four in the afternoon it is a geological record of the day: three coffee cups, a stack of things you meant to file, a Post-it with a phone number you no longer need, the good pen buried under a printout of something you already dealt with an hour ago. The desk is not full. There is, technically, room. You could clear a space if you tried. But you don’t try, because functionally, cognitively, the desk has stopped being usable long before it ran out of surface area. You start looking for the stapler and forget what you were stapling. This — and I did not make this term up, I want to be clear, though I wish I had — is context rot. The window hasn’t run out. The signal has just drowned in its own debris.

Researchers watching this happen to long-running AI agents have found something almost cruelly elegant about how it fails: it doesn’t fail gradually, the way you’d expect a desk to get gradually messier. Errors compound. A task that takes twice as long doesn’t get twice as likely to go wrong — the failure rate roughly quadruples. Two mistakes early in a long chain of dependent steps don’t add up to a slightly worse outcome. They multiply into something close to total collapse, four hours in, for reasons that trace back to a single bad assumption made in the first twenty minutes and never revisited.

Here is where the frontier comes back in — not as the whole answer, but as a piece of one.

It is not that frontier models are smarter in the way a benchmark measures smart — better at a single hard math problem, a cleverer turn of reasoning. Plenty of models can do that now; the “good enough” tier has crept remarkably high.

It’s that frontier models are apparently, marginally, meaningfully better at not rotting. At keeping the desk usable at hour six. At knowing which of the forty things on the desk actually still matters and which is a coffee cup that should have been thrown out an hour ago. This is a genuinely different kind of intelligence than the one benchmarks were built to measure, and it is almost invisible from the outside — you don’t see it in a single exchange, you see it only in the difference between a project that holds together over three days and one that quietly, subtly, stops making sense somewhere around Tuesday afternoon and nobody notices until Thursday.

If that’s true — if the frontier’s real edge is durability rather than raw cleverness — you’d expect to see it show up in how the labs actually deploy their own models: saving the sharpest tools for the tasks that need to survive the longest.

I went looking for a real-world example and found one closer to home than I expected: Anthropic’s own Slack tool, the one where you tag the AI into a channel the way you’d tag a coworker, and it works alongside a whole team over days, learning the channel as it goes. It runs on a serious, capable, thoroughly frontier model — but not, it turns out, on the company’s very best one. That one is held back, reserved for a smaller and stranger set of problems nobody has solved before at all. I sat with that for a while. The tool built to survive a whole team’s whole week, in public, under the most sustained pressure any of their products face, wasn’t handed the sharpest blade in the drawer. It was handed the second-sharpest — which was apparently, entirely, enough. Which tells you something about where the two kinds of intelligence actually diverge: the merely-very-good model handles the desk staying clean for a week, in public, in front of a whole team, where one bad assumption made Monday and never revisited would be visible to everyone by Thursday. The truly new capability is being held in reserve for something else altogether.

I don’t have a tidy place to land this, and I’m suspicious of anyone who does. But here’s the closest I can get.

Imagine a three-Michelin-star chef — the kind of person who has spent thirty years learning to coax something transcendent out of a single scallop, who can tell you, by smell, that a stock has forty more minutes in it — standing at your stove on a Tuesday night making you a grilled cheese sandwich. It will, I promise you, be a very good grilled cheese sandwich. The bread will be evenly golden. The cheese will have reached some ideal, fully-considered state of melt. But almost none of what makes that chef extraordinary is actually being used to make it — none of the thirty years spent learning to hold forty things in mind at once without losing track of any of them, the exact skill, it occurs to me, that keeps a long, complicated project from quietly falling apart on day three. The technique is idling. The thirty years are in the room, present, available, and almost entirely beside the point, because a grilled cheese sandwich was never the place where thirty years shows up. It shows up somewhere else — in a dish you will never order, on a night you weren’t there.

What you got instead, on your ordinary Tuesday, was simply more than enough.

Categories
AI

What the Lessor Keeps

Two airlines can fly the same airplane. Not airplanes of the same type — the same airplane, serial number and all, handed back at the end of a lease and reassigned, sometimes within weeks, to a competitor on another continent. AerCap owns more commercial aircraft than any airline on earth, and it leases them to airlines that spend their advertising budgets convincing passengers that flying them is a distinctive experience. The 737 MAX that wears Ryanair’s livery this year might wear Lion Air’s the next, repainted, recertified, its avionics untouched, its airframe indifferent to the change of ownership. The lessor does not care who is flying its asset. It cares that the asset comes back in airworthy condition and that the lease payments clear.

What the airline owns, in the sense that matters, is never the aircraft. It is the route network built up over decades of slot negotiations at constrained airports. It is the maintenance log — every inspection, every part swapped, every anomaly a mechanic in Singapore flagged in 2019 that turned out to predict a fatigue crack nobody else had seen yet. None of that travels with the airplane when the lease ends. It stays behind, compounding, in systems the airline built and the lessor never touches.

Karl Mehta, who has spent a career inside enterprise software watching this kind of asymmetry repeat itself, put a version of it plainly: a model is a brain you rent, and you and your competitor rent the same one. The formulation has the compression of something that has been tested in a few dozen meetings before it found that sentence. It is also, structurally, the airplane story. Anthropic and OpenAI and Google are AerCap. They retain residual value on enormous capital assets — clusters of GPUs depreciating on a schedule, weights trained at a cost that only a handful of balance sheets in the world can absorb — and they lease access to those assets by the token, to anyone who can pay, including, in the same afternoon, two companies trying to put each other out of business. The model does not know whose prompt it is answering. It has no loyalty file. It has, in fact, no memory at all, in the ordinary sense of the word — each call begins exactly where the last one ended for everybody, which is nowhere.

The asymmetry that airlines exploit is the one available here too, and it sits one layer up from the engine. Call it the embedding store, the vector database, the fine-tuning corpus, the retrieval index — the terminology varies by vendor, but the function is constant. It is the accumulated, indexed residue of every customer interaction a company has had, structured so that the rented brain can be handed the relevant fragment of it at the moment of each new call. A bank’s fraud model and a competing bank’s fraud model can call the identical foundation model, route through the identical API, and arrive at entirely different verdicts on the identical transaction, because one of them is retrieving against eleven years of labeled chargebacks specific to its own card portfolio and the other is retrieving against four. The intelligence rented by the hour is, for practical purposes, a commodity, priced down toward marginal cost the way jet fuel is priced — everyone pays close to the same number per unit. The memory is not a commodity. It cannot be, because it is not for sale; it is the institutional record of what has already happened to you, and no amount of capital lets a competitor buy a copy of your chargeback history any more than it lets them buy your maintenance logs.

This produces a particular kind of corporate vertigo, which Mehta’s sentence is really addressing. For three or four years the industry conversation about artificial intelligence has been a conversation about models — which lab’s was larger, which benchmark moved, which release cycle a company should anchor its roadmap to. That conversation rewards being an early and aggressive lessee. But a lessee relationship, however aggressive, does not compound into anything a competitor cannot eventually also lease. The compounding, when it happens, happens in the layer below the API call: in how cleanly a company has structured the record of its own customers, its own failures, its own edge cases, so that the rented brain, plugged in fresh every morning with no memory of yesterday, can be handed exactly the right fragment of yesterday and made to look, for a few hundred milliseconds, like it has been there all along.

A hospital chart has two kinds of entries. There is the vital-signs strip clipped to the bed rail — temperature, pulse, blood pressure, checked every four hours and replaced every four hours, because a reading from yesterday tells the night nurse nothing about the patient in front of her right now. And there is the permanent record in the file downstairs: the allergy that nearly killed him in 2019, the surgery, the medication history going back a decade, written once and never overwritten, because that record is exactly as valuable ten years from now as it is today. Nobody confuses the two charts. Nobody staples last Tuesday’s blood pressure into the permanent file. The hospital figured out, long before anyone digitized it, that memory is not one problem. It is two, and they fail in opposite directions if you run them through the same system.

Most teams building the layer Mehta is describing make exactly that mistake — they staple everything to the same chart. The shorthand for it is dumping everything into a vector database and praying, and it is worth asking why that particular error is so popular. The answer is that it feels like progress: embeddings go in, something resembling memory comes out, and the team moves on to the next sprint without confronting the harder question, which is what kind of memory it just built.

Short-term memory is the vital-signs strip — everything the model needs to finish the task in front of it and nothing it needs after. A customer-service exchange in progress, the order number already mentioned, the fact that this is the second call today, belongs here. So does the scratchpad of a multi-step agent: the search results just pulled, the file just opened, the partial answer being assembled before it commits. The test is not how important the information is but how long it stays true. A customer’s mood this minute is real and gone in twenty minutes; storing it permanently is like stapling yesterday’s temperature reading into the permanent file, undated, until the chart tells you nothing about fever and everything about clutter. Short-term memory should live in the context window itself, or a session-scoped cache, and it should be allowed to die when the session ends. The sin is not forgetting it. The sin is remembering it forever.

Long-term memory is the file downstairs, and it does not come in one shape any more than that file does. The first shape is semantic memory — facts. A customer’s account tier. The chargeback history that decides, in fractions of a second, whether this morning’s transaction clears. Facts belong in a database with a schema, not a vector store, because a fact has a right answer and a vector store gives you an approximate neighbor. Ask a vector index what tier a customer is on and it hands you the five most semantically similar sentences in the corpus — one correct, four merely correct-sounding. Ask a schema the same question and it tells you, because that is what the schema is for.

The more sophisticated shops are already building the seam between the two, rather than picking one and living with its blind spot. A knowledge graph keeps the relationships a schema is good at — this customer, that account, this chargeback, in fixed and queryable connection to one another — while still letting a retrieval layer search across it by meaning rather than by exact key. The approach has a name now, GraphRAG, and the name matters less than what it concedes: that facts and resemblance are different operations, and the honest fix is to run both and let each one answer the kind of question it’s actually suited for, not to force a single index to pretend it can do both jobs at once.

The second shape is episodic memory — what actually happened. The specific conversation last March in which the customer explained, at length, why the previous fix didn’t work. The exact sequence of an agent’s failed attempt at a task, preserved so the next attempt doesn’t repeat it. This is where the vector store finally earns its keep, because an episode isn’t an exact-match lookup, it’s a resemblance — has anything like this come up before — and a vector index, built to find the nearest thing to a fuzzy question, is the right tool for that question and almost no other. The error was never using a vector store. The error is using only a vector store, for facts as well as episodes, on the theory that one hammer with sufficient cosine similarity can stand in for the whole toolbox.

The third shape is the rarest, and the one teams forget to build at all: procedural memory, which is not a fact and not an episode but a skill — the model’s learned sense of how this company writes a refund email, escalates a complaint, formats an invoice. Style is the visible half of it. The other half is harder to see and matters more: the rails the model is forced to run on before it ever gets to choose a word. A refund above some threshold routes to a human, no exceptions, because the workflow says so, not because the model was persuaded to think so on this particular call. An agent that touches a production database does it through a reviewed function with a fixed set of permitted calls, not through whatever query it improvises in the moment. None of that lives in a prompt, and none of it lives in the model’s weights either. It lives in code — the orchestration layer, the permissioning, the state machine the agent is required to pass through — and it is procedural in the oldest sense of the word: not a memory of what to say but a memory of what is and isn’t allowed to happen, enforced whether or not the model that day feels like remembering it. It doesn’t live in a database at all. It lives in fine-tuning, in carefully maintained house-style examples, and in the surrounding scaffolding of guardrails and permitted actions, and it changes slower than the other two, the way a surgeon’s hands carry both technique and caution years after the specific patients are forgotten. A company that has built rich semantic and episodic memory but skipped this layer has a model that knows everything about its customers, writes in exactly the right voice, and is one well-crafted prompt away from doing something the company never agreed to.

The real argument here is not which database serves which layer — that part is plumbing, and plumbing changes every eighteen months. The argument is that memory has to be triaged the way the hospital triages it, with something deciding on purpose what survives the session and what doesn’t, rather than writing every token of every interaction into the same undifferentiated store and trusting retrieval to sort it out later. A vector database with no triage in front of it is not a memory system. It is a landfill with a search function, and it will retrieve the wrong eleven-month-old conversation with the same confidence it retrieves the right one, because nobody wrote the part of the system whose only job is deciding what belongs on which chart.

The lessor’s airplane, repainted, will fly for someone else next year. The route network will not. Neither will the schema that knows a customer’s tier on contact, nor the index that remembers the conversation from last March, nor the fine-tuned hand that knows, without being told twice, how this company writes a refund email. These are the things that do not come back at the end of the lease, because they were never on it.

Categories
AI Silicon Valley Technology

The View from the Edge

“Living on the edge” usually means you’re taking risks. One of the guests on the More or Less podcast used it the other way: as a diagnosis. A description of people who’ve lost their depth perception.

From where they sit, it looks like everyone is moving. The feeds are full of demos. The group chats debate which model won the week. Colleagues are building agents that book their dentist appointments and summarize their email while they sleep. David Sparks is selling a Robot Assistant Field Guide. The frontier feels like the present tense — not where things are heading, but where things already are.

When everyone around you has already crossed a threshold, you stop being able to see the threshold. You mistake the edge for the center.

The primary point — that the tech community wildly overestimates how much ordinary people want AI in their lives — lands harder when you hold it against that image. It’s not that the industry is wrong about the technology. It’s that it has miscalibrated the desire. Most people aren’t trying to optimize their Tuesday. They’re just trying to get through it. An always-on personal agent isn’t a solution to a problem they’re carrying.

Think about the woman in the Safeway parking lot, sitting in her car for three minutes before going in, scrolling back through her texts to find the thing her husband asked her to pick up. Egg product and cheddar cheese. She finds it, pockets her phone, and goes inside. The whole problem — the forgetting, the retrieval, the solution — lasted less time than it takes to read about it. She didn’t need an agent. She needed three minutes and a text thread she already had.

The edge distorts in a specific way: it makes appetite look like inevitability. From out there, adoption feels like a question of when, not whether. But whether is a real question. Most technology that could be woven into daily life never is — not because people couldn’t learn it, but because they didn’t want what it offered badly enough to bother.

The view from the edge is intoxicating. Everything looks like signal. But the middle is where most people live, and from there the signal looks a lot more like noise.

Which is why WWDC matters more than any model release this year. Apple doesn’t sell to people living on the edge. It sells to people who just want their phone to work. If Apple makes AI invisible enough — tucked into the camera, the keyboard, the thing that finds your photos — it stops being something you adopt and becomes something you already have. That’s a different motion entirely. Not convincing people they want AI. Delivering it before the question occurs to them.

Whether Apple can actually pull that off is a separate argument. But the watershed, if it comes, won’t look like a frontier crossing. It’ll look like a Tuesday that went slightly smoother than usual. Most people won’t even notice the edge they just walked past.

We will find out in a week or so.

Categories
AI Programming Software Work

The Scarcest Thing

Garry Tan woke up at 8 a.m. after sleeping at 4. Not because he had to. Because he wanted to see what his workers had done overnight.

The workers are AI agents. Ten of them, running in parallel across three projects. And something about that sentence — wanted to see what they’d done — keeps stopping me. That’s not the language of someone using a tool. That’s the language of someone managing a team.

Tan gave a name to the state this puts him in: “cyber psychosis.” He said it as a joke. But the joke has an insight in it. He’s not describing addiction to a productivity app. He’s describing a shift in what it means to do creative work — the strange vertigo of becoming a director when you’d always been a laborer.

I’m retired. I watch this from the outside now, which is its own kind of vantage point. For most of my career, the path from idea to working product ran through people — through hiring and managing and the slow accretion of execution capacity. You had the vision or you didn’t, but either way you needed the team. The idea and the means of making it real were, structurally, separate things. The gap between them was where companies lived.

What Tan is describing is that gap closing.

The thing he built — gstack, his open-sourced Claude Code configuration — got dismissed in some quarters as “just prompts.” And it is just prompts, in the same way that a conductor’s score is just notation. The abstraction is the invention. What he encoded is a model of how a startup team thinks: the CEO who interrogates the why before a line of code gets written, the engineer who builds, the paranoid staff reviewer who looks for what breaks. Each role blocks a different failure mode. Blurring them together produces, as his documentation puts it, “a mediocre blend of all four.”

That’s an organizational insight. It has nothing to do with code.

Tan described being a “time billionaire” — not because his biological clock had slowed, but because he can now purchase machine-consciousness-hours. The bottleneck of implementation, which has governed every creative project since the beginning of creative projects, is dissolving for those who know how to direct.

The scarcest thing is shifting. It’s no longer the hours of execution. It’s the clarity of intent — knowing what you want to build and why the journey matters, before any of the workers start moving. That’s harder than it sounds. For decades, most of us could muddle through in the making of it. The act of building taught you what you were building. Now the making is cheap, and that shortcut is gone.

For someone watching from retirement, that’s not a small thing to absorb. The model I internalized over a long career — that ideas become real through sustained organizational effort, through teams and timelines and the grinding work of execution — is being revised faster than I expected. Not invalidated. Revised. The judgment still matters. The taste still matters. The why matters more than ever.

It’s just that the how has found new hands. Many of them. More than any team I ever assembled, available the moment the intent is clear enough to direct them, gone when the work is done. The constraint was always the hands. It turns out it was always the knowing.

Categories
Micropayments

The Wrong Who

I was in the room for most of the early micropayments conversations. The working-level conversations, where people were genuinely convinced they had finally solved the problem. The demos were always compelling. The unit economics made sense on a whiteboard. And then they died.

They died so many times, and in so many similar ways, that the failure started to feel like a law of nature.

Clay Shirky wrote the autopsy that most people remember: micropayments fail because every transaction requires a decision, and decisions have a cognitive cost that swamps any payment below some psychological threshold. A dollar feels like real money. A dime feels like a question you have to answer. A fraction of a cent feels like being nickeled-and-dimed at sub-human speeds. The advertising model won because it asked users to consent once, peripherally, and then never bother them again.

So I noticed something when I read the transcript of Cloudflare CEO Matthew Prince’s earnings call remarks this afternoon.

He’s predicting that the internet’s business model — advertising and subscriptions, the twin structures that have governed everything since the late nineties — is about to change. He thinks some part of what replaces it will be micropayments for agentic traffic. Fractions of pennies. Fractions of fractions. At volumes that dwarf anything existing financial infrastructure can handle.

My first instinct was the old skepticism. We’ve been here before.

But I kept reading, and I think something is actually different this time. And the difference is the one thing all the earlier schemes never had.

The payer isn’t human.

This sounds obvious once you say it, but it collapses most of the objections that killed every prior attempt. Cognitive load isn’t a factor when there’s no cognition happening. Decision fatigue doesn’t apply to a process with no feelings about fatigue. The agent making the request doesn’t hesitate at a fraction of a penny, doesn’t resent the transaction, doesn’t abandon the session because it’s annoyed at being charged.

All the early micropayments architectures were built on an implicit assumption: that humans could be trained to behave like rational microeconomic actors at browsing speed. They can’t. Nobody does. But agents are rational microeconomic actors by design. That’s not a metaphor — it’s literally what they are.

The schemes we watched fail in the early 2000s weren’t wrong about the destination. They were wrong about the who. The internet of human readers and human attention was never a natural fit for per-transaction pricing. The internet of autonomous agents — making API calls, scraping data, assembling answers from dozens of sources in a single second — is a different thing entirely. And it’s arriving faster than most people realize.

Prince mentioned that Cloudflare thinks non-human traffic will surpass human traffic somewhere around 2027. That number stopped me. We are, apparently, closer to a majority-machine internet than to the one we think we’re living in.

The hard part isn’t the concept anymore. It’s the infrastructure. Prince was candid about this: the transaction volumes the industry gets excited about — a million per second — aren’t remotely sufficient for what’s actually coming. Cloudflare needs something an order of magnitude larger, and they’re looking for partners because nothing that fits the spec exists yet.

This is where it gets interesting for those of us who watched the earlier rounds. The original micropayments failures were partly psychological, but they were also partly infrastructural — the payment rails of the early internet weren’t built for high-frequency small transactions either. What’s different now is that the need is undeniable and imminent in a way it never quite was before. The traffic is real. The scale is measurable. The pressure to figure this out is coming from something other than optimism.

I don’t know what the solution looks like. Probably not one thing. Prince doesn’t know either — he said as much. Crypto infrastructure is an obvious candidate for parts of it, though crypto’s history of promising to solve problems and then creating different ones deserves some respect. Whatever emerges will probably be unrecognizable from here.

What I keep coming back to is the simpler observation. We were right that micropayments were the future. We just imagined the wrong future, populated by the wrong kind of payer.

The agents were always going to solve this. We just had to wait for the them to arrive.

Categories
AI Programming Prompt Engineering Software Work

The Great Inversion

For twenty years, the “Developer Experience” was a war against distraction. We treated the engineer’s focus like a fragile glass sculpture. The goal was simple: maximize the number of minutes a human spent with their fingers on a keyboard.

But as Michael Bloch (@michaelxbloch) recently pointed out, that playbook is officially obsolete.

Bloch shared a story of a startup that reached a breaking point. With the introduction of Claude Code, their old way of working broke. They realized that when the machine can write code faster than a human can think it, the bottleneck is no longer “typing speed.” The bottleneck is clarity of intent.

They called a war room and emerged with a radical new rule: No coding before 10 AM.

From Peer Programming to Peer Prompting

In the old world, this would be heresy. In the new world, it is the only way to survive. The morning is for what Bloch describes as the “Peer Prompt.” Engineers sit together, not to debug, but to define the objective function.

“Agents, not engineers, now do the work. Engineers make sure the agents can do the work well.” — Michael Bloch

Agent-First Engineering Playbook

What Bloch witnessed is the clearest version of the future of engineering. Here is the core of that “Agent-First” philosophy:

  • Agents Are the Primary User: Every system and naming convention is designed for an AI agent as the primary consumer.
  • Code is Context: We optimize for agent comprehensibility. Code itself is the documentation.
  • Data is the Interface: Clean data artifacts allow agents to compose systems without being told how.
  • Maximize Utilization: The most expensive thing in the system is an agent sitting idle while it waits for a human.

Spec the Outcome, Not the Process

When you shift to an agent-led workflow, you stop writing implementation plans and start writing objective functions.

“Review the output, not the code. Don’t read every line an agent writes. Test code against the objective. If it passes, ship it.” — Michael Bloch

The Six-Month Horizon

Six months from now, there will be two kinds of engineering teams: ones that rebuilt how they work from first principles, and ones still trying to make agents fit into their old playbook.

If you haven’t had your version of the Michael Bloch “war room” yet, have the meeting. Throw out the playbook. Write the new one.

Categories
AI Software Work

Lights Out in the Digital Factory

A quiet, modern unease haunts the vocabulary we use to describe invisible labor. Add “ghost” or “dark” to any industry, and suddenly a mundane logistical optimization takes on the sinister sheen of a cyberpunk dystopia.

Consider the “ghost kitchen.” Stripped of its spooky nomenclature, it is merely a commercial cooking facility with no dine-in area, optimized entirely for delivery apps. Yet, the term perfectly captures the eerie absence at its core: the removal of the restaurant as a gathering place, leaving behind only the pure, mechanized output of calories in cardboard boxes. It is a kitchen without a soul.

Now, we are witnessing the rise of the “dark software factory.”

“A dark factory is a fully automated production facility where manufacturing occurs without human intervention. The lights can literally be turned off.”

When applied to software, the concept is both fascinating and slightly chilling. A dark software factory is an automated, AI-driven environment where applications, features, and codebases are generated, tested, and deployed entirely by machine agents. There are no developers huddled around monitors, no stand-up meetings, no keyboards clicking into the night. It is “lights-out” development. You input a prompt or a business requirement, and the factory hums in the digital darkness, outputting a finished product.

Why are these invisible factories so important? Because they represent the ultimate abstraction of creation. Just as the ghost kitchen separates the meal from the dining experience, the dark software factory separates the software from the craft of coding. It optimizes for pure, unadulterated output and infinite scalability. In a world with an insatiable appetite for digital solutions, human bottlenecks—our need for sleep, our syntax errors, our slow typing speeds—are being engineered out of the equation.

But I can’t help but muse on what we lose when we turn out the lights. There is a certain melancholy to this ruthless efficiency. When we abstract away the human element, we lose the “front of house”—the serendipity of a developer finding a creative workaround, the quiet pride of elegant architecture, the human touch in a user interface.

The dark software factory sounds sinister not because it is inherently evil, but because it is utterly indifferent to us. It doesn’t care about craftsmanship; it cares about compilation. As we consume the outputs of these ghost kitchens and dark factories, we must ask ourselves: in our rush to automate the creation of our physical and digital worlds, what happens to the art of making?

The future of production is increasingly invisible. The dark factories are already humming. We just can’t see them.