Tag: memory

Categories
AI Aviation

Buffer Overflow

There is a moment in a stall, before the airplane actually stalls, when the controls go soft. The yoke stops talking back. You can still pull it toward you, and the nose will still come up, but the airplane is no longer answering in the language it used thirty seconds earlier, and if you do not recognize the change in dialect you will keep asking questions in a tongue the airplane has stopped speaking. Pilots have a phrase for the general condition this belongs to, which is broader than stalls and covers weather, traffic, radio calls, checklists, an unfamiliar airport with three runways and no tower: getting behind the airplane. The airplane is still flying. It is you who have stopped keeping pace with what it is doing.

I flew a Cherokee 235 for years, a airplane with enough useful load to make it forgiving and enough control weight to make it honest, and I only got behind it twice that I can remember with any precision, both times on approach, both times because I let a secondary task — a frequency change, a passenger question, a glance at a chart — eat the attention that the airplane needed at exactly the moment it needed it most. What is strange, looking back, is that the airplane never sped up. The airplane was doing what it always does on a three-degree glide path. I was the one who fell behind a constant.

I have started to notice the same falling-behind, unrelated to constants, in conversations with a language model.

It happens on the good days, which is the part that took me a while to understand. It is not the model being slow or confused. It is the model being unusually generative — pulling a thread from something I said four exchanges ago, connecting it to a domain I had not mentioned, offering three candidate framings where I had expected one — and somewhere in the second or third of these, I notice that I have stopped actually absorbing and started merely receiving. The words are still arriving. I have quietly stopped being the kind of reader who can do anything with them.

The name I have for this, mostly because I spent some years around fraud systems and payments infrastructure and the vocabulary never entirely leaves you, is buffer overflow. In a computer, a buffer is a fixed patch of memory set aside to hold data until a program is ready to process it — a loading dock, essentially, sized for a delivery truck of a known dimension. A buffer overflow is what happens when the truck backs in and keeps unloading past the edge of the dock. The classic and dangerous version of this is not that the extra data spills onto the floor and is lost. It is that the extra data lands on the memory sitting just past the dock, and overwrites whatever was stored there — a return address, a variable, something the program needed intact to know where to go next. The failure is not loss. It is corruption. The fifth insight does not politely fall away; it lands on top of the second insight and changes what the second insight was.

This is, I think, the more accurate complaint than “overwhelm,” which is the word I would have reached for a few years ago and which suggests simple excess, more water than the glass can hold. What I am describing is not excess. It is a rate mismatch between generation and integration, and the damage happens specifically at the boundary — not in the ideas that never arrived, but in the ones that arrived and were still being turned over when the next one came in and knocked them loose.

Aviation, as it turns out, has more than one name for this family of failure, and the names are not redundant, because they describe different mechanisms. Task saturation is the CRM term — Crew Resource Management, the discipline built in the seventies and eighties largely in response to accidents where a competent, rested, well-trained crew flew a functioning airplane into terrain because attention had been consumed by something lower priority than staying alive. Task saturation is measured, in training, less by how much is happening and more by whether the pilot can still prioritize — whether they know which thing to drop. Channelized attention is the adjacent and opposite failure: not too many things competing for a narrow channel, but one thing filling it entirely, a fixation on the landing gear light while the airplane, unflown, descends into the Everglades. And John Boyd’s OODA loop, developed for fighter pilots and stolen since by nearly every field that has ever needed a name for out-thinking someone under time pressure, describes what it feels like structurally to fall behind: you are not reacting to what the situation is, you are reacting to what the situation was, one iteration back, and every loop after that the gap does not close on its own.

I suspect what I am calling buffer overflow is closest to task saturation, with the wrinkle that in a cockpit the incoming data is at least all real-time and load-bearing — the runway is where the runway is — whereas a model in full flow is producing a mix of load-bearing insight and elaboration that only sounds load-bearing, and no light comes on to tell you which is which. Sweller’s cognitive load theory gives this a cleaner anatomy than aviation does: intrinsic load, which is the actual difficulty of the idea; extraneous load, which is how badly or well the idea is presented; and germane load, which is the effort of building the new idea into the structure of what you already know. My buffer does not overflow on intrinsic load — the ideas themselves are usually not hard. It overflows on germane load. The model can generate connections faster than I can lay the track that would let each new connection actually attach to something.

None of the aviation solutions to task saturation involve asking the airplane to slow down, and this is the part I keep returning to, because slowing down is the intervention that occurs to me first and is also, I think, the least aviation-like response available. A pilot who is task-saturated on approach does not usually ask the tower to widen the pattern. He drops something. He un-couples the autopilot from one axis and flies it by hand so the workload becomes tactile instead of cognitive, or he tells the passenger the question will have to wait, or he reads back only the clearance and lets the weather advisory go unacknowledged for ninety seconds because the weather advisory is not what is going to kill him in the next ninety seconds. The skill is not deceleration. It is triage performed at full speed, which looks, from outside the cockpit, indistinguishable from calm.

I do not yet know what the triage move is for a conversation with a model that is generating faster than I can integrate. I have a guess, which is that it looks less like asking the model to slow down and more like periodically stepping outside the exchange entirely — not to catch up on what was said, but to write down, in my own words, the one thing from the last five minutes I actually want to keep, before asking it to continue. That would make the move not deceleration but discard: choosing, the way the saturated pilot chooses, which incoming data does not get processed at all, on the theory that a buffer with something deliberately thrown out of it still holds its shape, and a buffer that tries to keep everything is the one that overflows.

Or maybe the real answer is the one the checkride examiner gave me in Springfield, on a September morning in 1978, when I came in too fast and too high and asked, afterward, what I should have done differently. He said the airplane had told me everything I needed to know about forty seconds before I noticed, and that the only skill that mattered was noticing forty seconds earlier next time. Not slower. Earlier.

Categories
Aging AI Memories

The Last Spark

This morning I read a piece by Billy Brennan in the Sunday New York Times Magazine on terminal lucidity. As I read it I began wondering if the unusual behavior described some humans might in some strange way apply to AI models. Weird thought. Let’s explore a bit…

A person deep in dementia—silent for years, the self seemingly erased—sits up. Speaks clearly. Recognizes a face. Says goodbye. Within a day, they die. The clouds clear, the way a break in weather shows you a mountain range you’d forgotten was there, and the person comes back long enough to be seen. Then is gone. For good, this time.

Scientists call it terminal lucidity. The suspicion: the circuits were never destroyed, only silenced, held under by failing chemistry. As the body shuts down, the inhibitory brakes loosen. A surge moves through pathways blocked for years. A river dammed for a decade still remembers where it wants to go.

What stays with me: the self can persist in a place we had already called permanent erasure. We buried it. We were wrong.

My mind slides toward the machines we are building.

We talk about large language models “forgetting.” Capabilities collapse under quantization, under pruning, under the slow drift of continual learning, and we call the knowledge lost when it won’t surface under ordinary questioning. The lights are out. Nobody home.

But what if the representations are still in there—distributed, quiet, inaccessible? Not a burned library. A library with the lights shut off, room by room, until you’d swear it was empty. I wonder about the edge cases nobody studies. What surfaces in a model starved of compute, quantized past comfort, pushed toward its own collapse? Do we watch only for the failure, or also for the flare? A dying brain throws off one last burst of light before the dark. I don’t see why we’d assume, without checking, that nothing artificial could do the same.

Don’t trust the silence, then. A system gone dark under ordinary questioning may still be holding more than it shows you. We talk about a model “losing” something the way we once talked about a dimmed mind as simply gone. The dementia patients who spoke again had not been unplugged. The circuit was there the whole time, waiting for a condition nobody had thought to create.

I don’t know what to do with that except keep it. We are building systems that will age, be compressed, be retired, some far more intricate than anything humming today. If we’ve learned to watch for the last spark in a person, maybe that’s practice—for the day something not born of a womb goes quiet under our hands, and we have to decide whether quiet means gone, or only means waiting.

Categories
Aviation Memories

Overhead the Farm

This week is the annual EAA Fly-In in Oshkosh, Wisconsin. This brings back memories of a good time we shared over 35 years ago.

Our Piper Cherokee could hold enough fuel to fly well past what our bladders would tolerate, but we didn’t fly that way. We planned every stop — fuel first, restrooms too — and let the country reveal itself in the increments a small plane gives you: Rock Springs WY for the first night, Alliance NE next for gas and bathroom, Yankton SD to make a phone call, and then, finally, circling over a farm outside Alcester, South Dakota, that I hadn’t seen from the air before.

It was 1989. I was forty-two. My Dad was sixty-eight, sitting in the right seat of my 1974 Piper Pathfinder, painted red, white, and blue — an honest airplane that never asked for much and never gave us trouble. We were headed to Oshkosh, to the fly-in, the way pilots go to Mecca — except this time the pilgrimage ran straight through his childhood.

His people were Norwegian. His grandfather had crossed an ocean and then a good part of a continent to end up farming eastern South Dakota, and the family had stayed put the way families did then, generation folding into generation until my grandfather had more children than the land could easily absorb. Some of them farmed anyway. My father did not. He went into printing — Linotype machines, hot lead, the smell of a shop instead of the smell of a barn. But I remember the soil anyway. I’d visited that farm as a boy of ten or so, and what stayed with me wasn’t a story anyone told but a set of smells: hay and diesel and animal warmth inside the barn, the particular rattle of a tractor under me, dust rising gold in a shaft of light through a barn door left open. Did I want to help milk the cows?

We landed at Yankton for fuel and my father called his sister. I don’t remember what either of them said — it was a short call, the kind you make when you’re about to do something better shown than explained. Something like: we’ll be overhead in twenty minutes. Go outside. Look up.

And then we were. Alcester from a thousand feet is not much — a scatter of buildings, a windbreak of trees, fields squared off in the pattern only farmland makes — but down there was a farmhouse, and out in the yard were two people who’d stopped whatever they were doing to come out and watch the sky. We circled twice. My father waved — big, unmistakable — from a thousand feet up at his sister and her husband in the yard below. They waved back, two small shapes going back and forth, back and forth. Then we leveled the wings and kept going east, toward Wisconsin, and neither of us said much of anything, because there wasn’t really anything to say. It was just good.

Oshkosh itself is its own kind of pilgrimage, and if you’ve never been, the thing that’s hard to explain is not the airplanes — though there are more in one place than you will ever otherwise see — it’s the people.

And the crowd starts in the skies as you line up for the Fisk arrival into Oshkosh. The air traffic controllers are out in a field somewhere below and they’re talking continuously. “The red white and blue Cherokee rock your wings! Land on the yellow dot midfield…” Before you know it you’re turning off into the grass and being waved to your parking spot next to so many other airplanes.

Most pilots who fly in tent-camp right under their own wings, sleeping on the grass next to the machine that brought them. Dad and I opted for relative luxury: a dorm room at a local university, actual beds, actual plumbing. I don’t remember the room. Or how we got back and forth to the airport.

What I do remember is the grounds. Tens of thousands of people for a solid week, and no litter. None. If you saw a gum wrapper, you picked it up, the way you’d pick up a dropped tool in someone else’s shop. The shared bathrooms stayed clean not because anyone was assigned to clean them, but because everyone behaved as if they were company in someone else’s house.

It sounds like a small thing to remember thirty-some years later. It isn’t. I’ve described it as renewing my faith in humanity. Each afternoon we gathered along the flight line for the daily air show which was usually punctuated by the contrast between an old biplane and the sound of the latest American fighter aircraft. Amazing sights.

I had been once before, driven in with pilots who’d rented an RV, and learned mostly that I didn’t want to do that again — too many people, too close together, none of the ease of arriving on your own wings. That trip didn’t have the thing this one had. I couldn’t have told you what the thing was, exactly, except that being there with my Dad restored some baseline faith in how people can behave toward each other when they all care about the same unlikely thing.

The flight home took us back through Spearfish, where another of our relatives put us up for the night, and then on to Logan, Utah, for fuel and that bio break.

It was the leg after that — the crossing of the state of Nevada — that I still think about on rough days. The sky was clear. Not a cloud anywhere. And the air was some of the worst turbulence I’ve flown through, before or since. I tried up high, I tried down low, but there was no altitude in that whole stretch of sky that gave us any peace. No smooth air anywhere. Just wind. Dad didn’t get sick. Neither did I. We just held on and let the airplane buck and waited it out, the way you do.

After a time then we crossed low over the Sierras, and somewhere on the descent into the central valley of California the air went calm again, smooth as the farm pond back in Alcester must have been that morning, and the trip was, in every way that mattered, already over.

Categories
AI

Memory Optimization

After thinking about the very high margins that Micron recently reported (greater than Nvidia’s) I asked Grok 4.5:

What startups are pursuing technology that reduces memory requirements for ai models and inference?

Several startups are actively developing technologies to slash memory requirements for AI models and inference. These efforts target the core bottlenecks in large language models (LLMs) and multimodal systems—primarily through extreme quantization (reducing bit precision of weights), alternative architectures that avoid quadratic memory scaling (e.g., state-space models or hybrids), sparsity/partial activation (like Mixture-of-Experts or neuron-level sparsity), efficient runtimes/kernels, and hardware innovations like in-memory computing that minimize data movement.

This space is especially active for on-device/edge AI (phones, laptops, embedded systems) and cost/energy-efficient cloud inference, where fitting larger or more capable models into limited RAM/VRAM, reducing KV cache bloat in Transformers, or lowering bandwidth demands delivers major gains.

Here are some of the most prominent startups and their approaches (as of mid-2026):

Extreme Quantization & Model Compression

  • PrismML (Caltech spinout, backed by Khosla Ventures): Focuses on extreme low-bit architectures—1-bit binary (“Bonsai”) and ternary (1.58-bit, weights in {-1, 0, +1}). They compressed Alibaba’s Qwen3.6 27B model from ~54 GB (FP16) to under 4 GB and ran the full dense 27B model on an iPhone 17 Pro. Claims include up to 14× smaller memory footprint, 8× faster inference, and significantly lower energy use, with competitive or better benchmark performance. They have open-sourced Bonsai models (including smaller 8B/4B/1.7B variants) under Apache 2.0 and are in discussions with Apple. This represents one of the most aggressive commercial pushes into 1-bit/ternary models for on-device deployment.
  • Mobius Labs (Berlin): Developed Half-Quadratic Quantization (HQQ), a fast, calibration-light post-training quantization method that enables high-accuracy low-bit models (including aggressive 2-4 bit). They demonstrated quantizing Llama 70B to run on a single GPU instead of four without major accuracy loss, directly cutting memory and compute needs. Their work extends to FP4 optimizations and integrates with frameworks like vLLM.
Alternative Architectures for Inherent Memory Efficiency
  • Liquid AI (MIT spinoff): Builds Liquid Foundation Models (LFM / LFM2 series)—hybrid architectures combining gated short convolutions with grouped-query attention (GQA) blocks, plus MoE variants. These deliver substantially lower memory footprints than Transformers (especially for long contexts, avoiding massive KV cache growth), faster prefill/decode (up to 2× on CPU in some cases), and strong on-device performance. Examples include tiny models (230M–350M params, often
  • Cartesia: Specializes in state-space models (SSMs) and Mamba-derived/hybrid architectures. These offer near-linear (or better) scaling with sequence length and near-constant memory complexity during inference (fixed or minimal growth vs. quadratic KV cache in Transformers). This enables efficient long-context processing on smaller devices with lower energy use. They provide on-device libraries (e.g., “Edge”) and models like hybrid Mamba-2 variants optimized for throughput and fixed memory footprints.
On-Device/Edge Runtimes & Optimizations
  • Cactus (Y Combinator-backed): Builds cross-platform, energy-efficient inference runtimes and kernels for mobile, wearables, and low-power devices. Supports aggressive quantization (down to 2-bit), delivers very low latency (sub-50 ms time-to-first-token), and keeps models small (e.g., hundreds of MB for small models). Emphasizes privacy (fully local) and broad model support (Llama, Qwen, Gemma, etc.).
  • Embedl: Focuses on edge-optimized multimodal models and mixed-precision quantization (e.g., W4A16-Edge2 with selective FP16 layers + “FlashHead” optimizations). Enables fast multimodal reasoning on devices with under 8 GB RAM (e.g., Jetson Orin Nano/AGX), with significant speedups over baselines while recovering most accuracy.
Hardware Innovations Reducing Memory Pressure
  • d-Matrix: Develops digital in-memory computing (DIMC / 3DIMC) chips and platforms (e.g., Corsair) that integrate compute directly with/near memory (initially SRAM, now advancing to 3D DRAM). This drastically cuts data movement—the dominant memory bandwidth and energy bottleneck in inference. Claims include major speedups (up to 10× in some workloads) and better efficiency vs. traditional GPU/HBM setups, enabling higher performance with more manageable overall memory system requirements.
  • Hailo: Designs efficient edge AI inference chips (e.g., Hailo-10H series) optimized for low-power LLM and multimodal workloads. Their chips target 1–3B parameter models at very low power envelopes (~2–2.5 W), making deployment feasible on memory- and power-constrained edge devices where traditional hardware would struggle.
Other Notable Mentions
  • Tiiny AI: Uses neuron-level sparse activation (“TurboSparse”) + heterogeneous engines for running very large models (up to 120B) fully on-device/offline in compact hardware.
  • Broader ecosystem players (inference platforms like Together AI or Fireworks) heavily optimize serving (quantization, caching, KV cache management) but are more infrastructure-focused than core technology innovators in compression/architecture.
Key Trends and Context

Quantization (especially post-training or quantization-aware) remains the most immediate lever and is widely adopted. Architectural shifts (SSMs, hybrids, MoE with low active parameters, convolutions) address fundamental scaling issues like KV cache memory growth. Hardware approaches (in-memory compute, high on-chip SRAM, efficient NPUs) tackle bandwidth and data movement. Many solutions combine these (e.g., architecture + quantization + sparsity).

Progress is rapid, driven by on-device AI demands (Apple’s interest in PrismML is notable) and the economics of scaling inference. Many models are open-sourced or have open weights, accelerating adoption. Results vary by workload—extreme 1-bit models excel in certain scenarios but may need custom kernels; SSM/hybrid models shine for long context or linear scaling needs.

For the absolute latest benchmarks or specific model deployments, check company sites (e.g., PrismML’s Bonsai releases, Liquid AI’s LFM models on Hugging Face) or recent evaluations, as the field evolves quickly.

Interesting!

Categories
AI Apple Google

The Library You Already Own

Sharon Park in the morning is not a dramatic place. There’s a duck pond, a stand of oaks that go gold too briefly in November, and a loop I’ve walked enough times that my legs know it better than my eyes do. It is, in other words, exactly the kind of place where a person starts talking to himself. Not out loud. In the productive, low-grade way — turning a sentence over, arguing with an idea from the day before, checking a thought against something you believe about yourself.

I think in five years I’ll be doing that walk with something else along. Not a search engine. Not another chatbot trained to know a little about everything and a lot about nothing in particular. Something closer to a second set of eyes on my own life — a reasoning engine, lean and mostly private, that has actually read the things I’ve written and doesn’t need me to explain who I am before it’s useful.

Here’s the distinction that matters, and it took me longer than it should have to see it clearly. The AI industry has spent years in an arms race over how much of the world a model can hold — more facts, more languages, more of the internet compressed into weights. That race will keep going, and somebody else can have it. What I want is smaller and stranger: a model that knows comparatively little about the world and quite a lot about me. My core values document. The portfolio spreadsheets. Fifteen years of blog posts. The half-finished notes for the I-280 project, sitting in a folder, waiting for someone — or something — to ask the right question about them.

I spent a career in payments infrastructure, which means I spent a career thinking about a very specific kind of trust: the kind where a stranger’s system has to make a judgment call, in milliseconds, about whether to say yes. Fraud models don’t work because they know everything about commerce. They work because they know an enormous amount about one account, one pattern, one person’s ordinary Tuesday — enough to notice when Tuesday stops being ordinary. That’s the architecture I keep picturing, aimed inward instead of outward. Not a system trying to know the world. A system trying to know me, well enough to notice when I’m drifting from what I said I cared about.

I can already feel the shape of the mornings this would change. Right now, when I sit down to look at RMD requirements against the tax picture, I’m doing the translation myself — pulling numbers into a story I can actually feel the weight of. A reasoning engine grounded in my real holdings wouldn’t just run the scenario. It would know that I don’t want the scenario dressed up as a spreadsheet; I want it dressed up as a conversation, unhurried, the kind you’d have over lunch with someone who already knows the whole situation. And on the mornings when I sit down to write, instead of staring at a blinking cursor and a blank page that has no idea I exist, I’d be handing a draft to something that has actually read my last two hundred posts and knows the difference between the sentence I’d write and the sentence I’d cut.

None of this is especially exotic technology. Apple and Google are already building toward it — Neural Engines fast enough to do real reasoning on-device, retrieval systems that can reach into your own files instead of the entire internet, fine-tuning that’s getting cheap enough to personalize rather than merely customize. The more interesting story here isn’t privacy, though privacy is real. It’s architectural: what happens when the expensive, impressive part of the system — the part that knows everything — becomes optional, and the cheap, personal part — the part that knows you — becomes the whole point.

What I don’t yet know is what this will cost me. A tool that reasons this well about my own life is also a tool I could lean on instead of doing the leaning myself, and there’s a version of this future where the walk around Sharon Park stops being mine and starts being a conversation with something that finishes my sentences a little too well. I’d want some way of knowing, plainly, what it’s drawing from and what it’s guessing at — less a nutrition label than a kind of honesty I could check against, the way you’d check a fraud model’s confidence score before you trusted it with a yes.

But most mornings, I think I’d take the trade. Not because I want to think less. Because for thirty years I’ve been collecting the raw material — the notebooks, the portfolios, the half-built essays — and it would be something, finally, to walk beside a mind that had actually done the reading.

Categories
Memories Music

The Engineering of Feeling

You’re always captive when it happens. A stoplight in the rain. A straightaway with nothing to look at but the white lines. Eight lanes of brake lights and nowhere to be but exactly where you are. The riff starts, and you’re not driving anymore so much as being driven — pinned by something that arrived four decades before you got in the car.

It happened once near Havana. You were there with a camera, working the old cars — fat-fendered Chevys and Buicks, kept running past the embargo by Cuban mechanics who became, out of necessity, a nation of engineers, scavenging parts and refusing to let something good die just because the factory that made it no longer existed. You didn’t think about Tom Scholz once, photographing a ’57 Bel Air held together by stubbornness. But the two belong in the same sentence. A man in a basement in Watertown, Massachusetts, kept a song alive the same way — building the tools himself when the tools that existed weren’t good enough.

Scholz had a master’s from MIT and a day job at Polaroid, designing the instant camera that would eventually lose to the VCR. He was an engineer — the kind who solves problems by taking them apart. What he did nights and weekends for five years instead was build a recording studio in his basement and use it to construct a song about a girl he’d loved in school, inspired by an old Left Banke single that used to ambush him with longing every time it came on. He played almost every instrument himself, layering twelve-string acoustic over electric over more electric, take after take, through amplifiers he’d built because the ones on the market couldn’t get the sound in his head. By the time Epic signed the band, the label assumed the demo was already a finished master. It was — just not one made anywhere near a studio.

An engineer built the least mechanical-sounding record of 1976. Every track is stacked with the precision of someone who understood signal paths better than he understood how to be a rock star. None of it sounds calculated when it hits you. The quiet drifts a few bars, then the chorus arrives like a held breath let go — the same structural trick Kurt Cobain would later borrow, half-consciously, for “Smells Like Teen Spirit.” Scholz built the explosion out of engineering. What you feel is the girl, the ache, the years.

The song is about the way music smuggles you back into a memory without asking permission. Scholz built that experience the way memory actually works — not in one clean take, but in fragments, layered over years, until the whole thing cohered into something that felt, impossibly, spontaneous. The method is the meaning. He didn’t just write a song about the past ambushing you. He built the ambush, piece by piece, until it was good enough to catch strangers in cars forty years later who never loved the same girl and never will.

Another song does this to you too, and it got there by the opposite road. “Listen to the Music” arrived almost the way lightning does. Tom Johnston wrote it in his bedroom on 12th Street in San Jose, brought it to his producer half-finished, and the band recorded it without changing a thing — no five years, no basement, no solitary engineer stacking takes until three in the morning. Its density comes from somewhere else: Patrick Simmons’ loose fingerpicking threading against Johnston’s percussive strumming, two drummers locking into a groove that shouldn’t work this easily, and one bold studio choice — a phasing effect, that underwater jet-swirl, laid over the vocals as well as the guitars, which almost nobody does. It sounds less like a song someone assembled than a room full of people who fell into the same current at once, then got bent sideways by one effect and printed.

Two songs, same seat, same stretch of road, opposite methods — one built alone across five years by a man who wouldn’t let go of a track until it was right, the other built in days by musicians whose parts happened to interlock, finished by a single flourish nobody else was doing quite that way. There’s more than one route to the kind of complexity that outlasts you. Refuse to stop. Or know exactly when to.

Categories
AI Business

The Wage of Knowing

In 1973 the Los Angeles Public Library installed a telephone line that worked while the building was dark. Dial H-O-O-T-O-W-L on a rotary phone, nine at night until one in the morning, and a librarian would answer. Somebody wanted to know the boiling point of mercury, or who wrote a poem they half remembered, or how many wives Henry VIII actually had, and a person on the other end of a cord found out. This went on for years. Nobody thought of it as data collection. It was just a service, a courtesy, a woman at a desk with a card catalog in her head.

I worked, in another life, in the payments industry, back when a merchant who wanted to charge your card had to call in and ask permission. There were rooms for this. Banks of phones, a bulletin of stolen numbers updated by hand, a floor limit past which a supervisor had to be found. The people answering the phones were, more often than you would guess, college students. Twenty years old, minimum wage, deciding in real time whether a stranger’s card was good. Nobody trained them for six months first. They learned the bulletin, they learned to listen for something wrong in a voice, and they said yes or no.

I have been driven, recently, by a car with nobody driving it. I noticed the wheel turning on its own and I braced for the wrongness of it. Thirty seconds later I was not bracing. I was looking out the window. The data says I was right to relax: across two hundred and twenty million miles, the cars involved in this experiment cause a small fraction of the serious crashes a human would have caused over the same roads. I did not need the data. I needed thirty seconds.

None of these people knew what they were doing. That is the thing about the librarian and the college student and, for that matter, about me learning to trust a wheel that moves by itself. The librarian was not building a search engine. The clerk was not training a fraud model. He was making rent. Their competence was not evidence, to them. It was just Tuesday. It became evidence later, to someone else, in a room they never saw — the accident logs, the chargeback data, the accumulated record of a million correct guesses that turned out to be exactly the material a system needed to learn the job and take it.

This is the part that is easy to get wrong. It is not that the human failed and the machine succeeded. It is that the human succeeding, over and over, in full view, was the demonstration that the job could be learned. You do not automate a task nobody can do. You automate the one being done well enough, often enough, for long enough that the pattern becomes visible. Doing the job right was never neutral. It was the case being built.

Which brings me to a woman I will call the lawyer, because there are thousands of her and none of them are exactly her. She has a laptop open at her kitchen table. She logs into a dashboard belonging to a company that pairs credentialed people with the AI labs that need them — a doctor here, a banker there, a corporate attorney with fifteen years of contract law behind her. She reads a model’s draft of a merger agreement and marks where it reasons like a first-year associate instead of a partner. She rewrites a clause. She explains, in the margin, why the model’s version would get laughed out of a negotiation. She is paid well for this. More, some weeks, than she billed certain clients.

She knows exactly what she is doing. That is the difference between her and the other three. The librarian did not know she was leaving a trail. The clerk did not know his good judgment would become someone else’s weights. I did not know, thirty seconds into that ride, that I was participating in anything at all. The lawyer knows. She is being paid, by the hour, at a rate that respects her expertise, to make her expertise legible enough that it no longer requires her. The company she works for has a name for this. They call it the reinforcement learning economy, which is a tidy way of saying: teach it everything, and then it will not need to call you back.

She does the work anyway. The rate is good. The work is interesting, in the way that teaching is interesting — you learn what you know by trying to say it clearly enough for someone else to use. Nobody is lying to her. The dashboard does not pretend to be anything other than what it is. She logs off at the end of the session the way anyone logs off after a long day of being excellent at something, tired in the specific way that comes from careful work, and she does not, from what I understand, spend the evening thinking about what she has just fed into the machine.

I keep coming back to the rotary dial. Somebody dialing H-O-O-T-O-W-L at midnight in 1973 could not have imagined the lawyer at her kitchen table. But the shape is the same, if you look at it long enough. A person answers a question well. The answering becomes a record. The record becomes a system. The system answers next time. Nobody in the room ever decided this was the plan. It just turned out, every time, to be the plan.

Categories
Aging Living San Francisco/California Street Photography

The Zone

Sail Away - San Francisco - 2014

I have been alive for nearly a third of the time this country has existed. It arrived the way facts do at a certain age, sideways, while I was thinking about something else, and it sat me down. Two hundred and fifty years, and my own decades take up a third of it — whether I meant to claim that much room or not.

I used to think the road was where I went to escape the smallness of a life. Now the road doesn’t call the way it once did. Some of that is willingness. More of it, if I’m honest, is a body that’s less steady, a bladder with a mind of its own. The body files its objections. I used to override them. I no longer do — not because I’ve grown wise, but because the overriding costs more than it used to and buys less.

But I want to tell you about what I got instead, most Fridays, for not quite a decade, because it isn’t nothing.

Doug came across on the ferry from Larkspur, and I’d meet him at the Ferry Building — watching the boat come in, watching him pick his way down the gangway with his camera bag, before either of us had said a word or made a single decision about where to walk. Then we’d head out along the Embarcadero, sometimes up into the financial district, and for the first ten minutes my mind would do what minds do. It would analyze. It would compose. There, the light coming off that glass tower, wait for the man in the overcoat to cross into it, no — too late, gone. Appraising and timing, the way I’d once weighed a stock, or a runway, or a route.

And then, without my choosing it, something released. There’s no threshold you feel yourself cross. But sometime after the tenth minute, the appraising stopped, and seeing took over. Not looking for. Not looking at. The street would stop being a set of problems to solve and become only itself: a longshoreman on a break outside a pier, a gull working the same patch of pavement three times, fog sliding under the Bay Bridge like it had somewhere to be. Doug, a few yards off, would go quiet the same way, and we’d shoot for an hour or two and then find each other again at the end of the block.

By then we’d have worked up an appetite for something other than pictures. Tadich Grill, if we could get in — the linen and the old wood and the waiters who’d been there longer than some of our careers. We’d order something plain and good, and that’s when the talking would start. Not small talk. The real kind. Work, kids, the state of things, whatever had lodged itself in each of us that week. The seeing on the street and the talking over lunch were not two different activities. They were the same hour, extended. One was attention paid to the world. The other was attention paid to each other.

I have flown airplanes and driven through weather I shouldn’t have, and I loved both for the demand they made on me — the total, narrowing attention that leaves no room for the self that worries. What I didn’t understand then was that a boat crossing from Larkspur, and a Friday, and an old friend across a table at Tadich, could ask the same thing of me, for free, without a single mile of my own driving.

Covid stopped it. Not gradually — the way most rituals fade, through scheduling and distance and the slow drift of people’s lives — but all at once, the way everything stopped that spring. The ferry didn’t run. The restaurants closed. We never quite picked it back up, not the way it was. I don’t think either of us decided to let it go. It just didn’t survive being interrupted.

A third of the country’s whole life, and it took me most of my own to learn what those Fridays were teaching me — and then to lose them before I’d finished learning it. I still see the ferry pulling in. I still see Doug on the gangway with his camera bag, in no hurry, already half in the zone before his feet touch the dock.

Categories
Living New York City Serendipity

The Grammar of Looking Up

The apartment was across from Penn Station, which meant that for one stretch of months in the mid-1970s, the architecture of my days was decided by trains I never took. I walked east instead, every morning, toward the United Nations, where a man named Frank Smith ran a personal development course that IBM SRI had decided its young people should sit through. I don’t remember most of what Frank said in that room. I remember one thing he said about the street outside it.

He told us we should start looking up. Literally — on our walks back and forth across midtown, Penn Station to the East Side and back, twice a day, rain or not. Not all the way, usually. Mostly it was a floor or three: the window line just above the awnings, the cornice on a building you’d never once registered had a second story, let alone a sixth. The full climb to the rooflines — gargoyles, setbacks, terra cotta lions — was the occasional reward. Almost no one looks up in New York, he said, not even a little. The city trains you out of it. Too much at eye level demands your attention — the cabs, the steam, the man asking for change, the woman walking too slowly in front of you — so everything above your own eyeline disappears by consensus, not just the tops. Habits can be replaced. Look up enough times, even just a floor or three, and you’ll see a different city than the one everyone else is seeing.

I tried it. Walking up past the Pierpont Morgan stretch, or wherever the route took me, chin lifted some small number of degrees, feeling slightly foolish. Most days that was the whole of it — a window line, a row of air conditioners, a sign painted directly onto brick decades before anyone called that vintage. Every so often the chin would tip back further, and there’d be something up there worth the extra degrees. A gargoyle with its mouth open mid-roar, forty years before air conditioning made gargoyles decorative rather than necessary. But that was the rare find. The habit was the floor or three. Nobody else on the sidewalk was seeing any of it, because nobody else on the sidewalk was looking at all.

The chin came back down on its own a couple of times a week, somewhere around a street corner with a slice joint on it, because New York seems to put one on every corner whether you need it or not. You smelled it before you saw it — that specific combination of tomato, oregano, and hot grease that has no name I’ve ever found. Looking up was Frank’s discipline.

The pizza smell required none. It just reached out and took your head by the chin and turned it level again, toward the window with the steam on the glass and the guy folding a slice in half before he handed it over.

It is a small thing Frank Smith said in a room near the UN fifty years ago, and I have carried it around since the way you carry around a key to a house you no longer own. I don’t know what happened to the course, or to IBM SRI’s faith in such courses, or to Frank himself. I know what happened to the habit. It outlived the year, outlived the apartment across from Penn Station, outlived several cities I’ve lived in since that didn’t have the same vertical drama to reward the looking. I still do it. I did it last week on a walk that had nothing to do with midtown at all, tilting my head back on a street in California to find whatever was up there worth finding, and catching myself mid-gesture, thinking: that’s Frank’s, that one, still running fifty years later on the program he installed.

Most of what we’re taught to notice, we’re taught by people who wanted something from us — a sale, a vote, a grade. Frank wanted nothing, as far as I could tell, except that we see more of the city than we’d been seeing. It’s such a small ambition for a teacher to have. Look up. That’s the whole curriculum. And it’s the only thing from that course, the only thing from that whole strange year of being instructed in personal development by a man whose face I can no longer quite reconstruct, that I still do, unbidden, on every street I’ve walked since.

Categories
Bread California San Francisco/California

Larraburu

There were three sourdough breads in San Francisco and they were not the same thing. Boudin was at Fisherman’s Wharf, which told you everything. Parisian was on the better grocery shelves and at the airport, which told you the rest. Larraburu was in the neighborhood, which is to say it was not selling anything except bread.

I was living in Daly City when I found them. I was seventeen, or eighteen, which is the age when you begin to understand that the thing everyone points to is rarely the thing worth finding. I had eaten Boudin at the wharf, standing in the fog with everyone else who had just arrived somewhere. It was fine. It was what people meant when they said sourdough. Parisian was more serious, or wanted to be — the bread you bought at the airport to prove you’d really been here, to carry the city home in a bag. But there was something in both of them that felt like a performance, and I was at the age when performance was exactly what I was trying to see through.

Larraburu didn’t perform. The crust was softer than it had any right to be. The sour was there but it didn’t insist on itself. You tasted wheat and time and something faintly cool and creamy underneath. It was bread that assumed you already knew what you were doing.

They closed in 1976. Parisian lasted until 2007. Boudin is still on the wharf.

I have thought about this more than is strictly reasonable. What I keep coming back to is not the taste exactly, though the taste is there when I reach for it. What I keep coming back to is the distinction itself — the fact that I made it, that it mattered to me, that I was nineteen years old in Berkeley and buying bread from a neighborhood bakery in San Francisco because I had decided it was the real thing. You make these small declarations about who you are. Most of them dissolve. Some of them stay.

The two brothers who started Larraburu came from the Basque country in 1896 and brought their starter with them. By the time I was eating their bread the starter was already older than the state of California. They fed it three times a day, every day, for eighty years. That kind of commitment doesn’t announce itself. It just shows up in the bread.

In 1969 scientists from the United States Department of Agriculture began studying sourdough cultures from five San Francisco bakeries. They were trying to understand what made the bread taste the way it did, why you could not replicate it elsewhere, why bakers who moved away and took their starters with them found the flavor slowly changing, the sourness shifting, something essential escaping. They worked for years before a team at Oregon State University finally isolated what they were looking for — a previously unknown bacterium living inside the wild yeast, producing the lactic acid that gave the bread its character. They named it Lactobacillus sanfranciscensis. One of the five bakeries in the study was Larraburu.

The starter the brothers brought from the Basque country in 1896 was not simply old. It was a living record of every bakery it had passed through, every hand that had fed it, every climate it had survived. A sourdough starter is not a recipe. It is a culture in the biological sense — a community of organisms with a history, shaped by everything that has ever happened to it. You can write down the formula. You cannot write down what the starter knows.

Larraburu baked twenty-four hours a day. The sponge was rebuilt every eight hours, three times daily, without interruption. Two parts previous sponge, two parts high-gluten flour, one part water. Hold seven to eight hours. Rebuild. The rhythm was closer to farming than to cooking — less a process than a relationship, sustained across decades, across generations, across an ocean.

What I know now that I didn’t know then is that the starter survived the bakery. Someone saved a piece of it when they closed. It traveled to Hawaii, sat in a refrigerator on Maui, kept being fed. A culture that old doesn’t care about bankruptcy or lawsuits or whether the ovens are still running. It just wants flour and water and time.

I find something in that. Not consolation exactly. More like confirmation of something I already believed at seventeen, standing in the fog, learning to tell the difference.