Tag: efficiency

Categories
Menlo Park Serendipity

Two Kinds of Efficiency

The fog hadn’t lifted yet over Sharon Park, the kind of gray that Menlo Park wears many June mornings like it’s embarrassed to admit the sun is up there somewhere, and I was on my usual loop around the pond when I noticed in the distance that the goats were back. And one more thing too. I stopped.

On one side: forty, maybe fifty goats, heads down, working a hillside of dry summer grass like a crew that had done this job a thousand times, because they had. The city brings them in every year around now, before fire season, to eat down the fuel load that nobody wants to mow. White ones, brown ones, a few with horns curling back like something out of a hieroglyph. They don’t look up much. A goat eating is a goat with one job and no curiosity about yours.

On the other side, maybe forty yards past them, through the wire: a Waymo. White, sensor pod spinning slow on the roof like a lighthouse that had wandered inland and gotten confused about its purpose, parked at the curb with nobody in it. Just sitting there. Idling, if a thing with no engine can idle. Waiting on a fare, or waiting on nothing, the way these cars do now, patient in a way that doesn’t read as patience because there’s no face attached to it.

I stood looking for longer than the moment deserved, the way you do when something hands you a thought before you’ve earned it. I remembered I should take a photograph.

Here is what struck me, eventually: both of them were efficient. That’s the word that kept showing up, uninvited. The goats are efficient in the oldest way there is — they convert a problem (too much dry brush, a fire waiting to happen) into a solved problem, using nothing but appetite and stomachs and several thousand years of being bred for exactly this. Nobody programmed a goat. A goat doesn’t have a model. A goat has a memory that goes back to whatever the last hillside tasted like, and an instinct that says eat that one next, and that’s the whole operating system.

The Waymo is efficient in the newest way there is. Lidar instead of appetite. A map instead of memory. It doesn’t get bred for the job, it gets trained for it, mile after simulated mile, until eventually you can park it at a curb in a quiet park and trust it not to do anything stupid. It was, in its way, doing the same thing the goats were doing — converting a hard, slightly dangerous task that used to require a person’s full attention into something that just sort of happens now, off to the side, while everyone gets on with their morning.

I’ve spent a fair amount of my working life around payments systems and fraud models, which is its own quiet machinery — systems built to notice the thing before the thing becomes a problem, the same job the goats were doing on that hillside, eating the grass before it becomes a fire. So maybe that’s why I stood looking longer than I meant to. I recognized the shape of it, even though one side of the fence had hooves and the other side had a sensor array worth more than my first house.

What I didn’t expect was how unbothered each side seemed by the other. The goats did not care that there was an expensive autonomous vehicle parked within sight of their breakfast. The Waymo, for its part, did not care about anything, which I suppose is the whole point of it — it isn’t built to care, only to notice, and the goats had registered exactly zero on whatever sensor suite decides what’s worth noticing. Two systems, separated by maybe forty yards and several thousand years of technological distance, each one going about its business with total indifference to the other’s existence.

I used to think the line between old world and new world would announce itself — some clean morning where you’d wake up and the future would have visibly arrived, banners out, the old thing retired with a gold watch. It doesn’t work that way, it turns out. It works like this: a fence, some goats, a car with nobody driving it, and a guy on his usual walk who happens to notice that both of them are quietly, competently doing a job that fire season requires somebody — or something — to do.

I kept walking. The goats kept eating. The Waymo, as far as I know, was dispatched somewhere, picked up whoever needs a ride, sensor pod turning over the same hill the goats had already half cleared. Two kinds of efficiency, on either side of an electrified wire fence, neither one impressed by the other, both of them right.

I don’t know what to do with that, exactly, except to write it down and remember it. Some mornings my walk gives me exercise. Some mornings it gives me a simple memory I didn’t ask for, standing there looking.

Categories
AI AI: Large Language Models

The Echo Effect: Why Prompt Repetition is AI’s Best Kept Secret

In our relentless pursuit of complexity, we often overlook the elegant simplicity of a fundamental human habit: repeating ourselves.

We build colossal architectures, weave intricate neural networks, and throw mountains of computational power at our artificial intelligence systems, hoping to squeeze out a few more drops of reasoning and logic. Yet, sometimes the most profound breakthroughs require no new code, no additional latency, and no extra training data.

Sometimes, you just have to say it twice.

In a fascinating December 2025 paper titled “Prompt Repetition Improves Non-Reasoning LLMs,” researchers Yaniv Leviathan, Matan Kalman, and Yossi Matias uncovered an almost absurdly simple “free lunch” in AI optimization.

Their premise is straightforward: when you aren’t using a heavy reasoning model, simply copying and pasting your input prompt multiple times significantly boosts the model’s performance.

“When not using reasoning, repeating the input prompt improves performance for popular models (Gemini, GPT, Claude, and Deepseek) without increasing the number of generated tokens or latency.”

The mechanics behind this are elegantly pragmatic.

By repeating the prompt, you are moving the heavy computational lifting to the parallelizable “pre-fill” stage of the model’s processing. The AI’s causal attention mechanism gets to process the same tokens again, allowing the later iterations of the prompt to attend to the earlier ones. It effectively acts as a hack to simulate bidirectional attention in a decoder-only architecture.

What’s even more telling is the paper’s observation on why this works so well.

The researchers noted that models trained with Reinforcement Learning (like OpenAI’s deep-thinking variants) naturally learn to “restate the problem” in their internal monologue. They figured out on their own what these researchers are suggesting we do manually: repeat the question to focus the mind.

Reading this paper, I couldn’t help but draw a parallel to the human condition and the nature of listening.

How often do we assume that because we have articulated a thought once, it has been fully absorbed? We fire off a single, dense instruction to a colleague, a partner, or a friend, and then marvel when the nuance is lost in translation.

We suffer from our own attention bottlenecks.

Like a non-reasoning LLM trying to parse a complex query in a single pass, we are constantly bombarded with a stream of tokens—emails, notifications, conversations, fleeting thoughts. To truly understand, to truly digest and synthesize information, we need the grace of repetition.

There is a strange poetry in the fact that to make our most advanced digital minds smarter, we have to talk to them the way we talk to a distracted child or a busy spouse. The “microscope effect” highlighted in the study—where repeating a prompt drastically improved extraction tasks—shows that the failure wasn’t in the model’s capacity to know, but in its capacity to focus. Repetition forces focus. It creates a resonant echo in the context window, a digital highlighter that screams, “This matters. Look here again.”

As we continue to navigate a world increasingly augmented by artificial intelligence, this paper serves as a humbling reminder. The bleeding edge of technology isn’t always found in the most complex equation; sometimes, it’s hidden in the most basic principles of communication.

Whether you’re prompting a billion-parameter language model or trying to connect with the human sitting across from you, the lesson is clear.

Clarity isn’t just about the words you choose. It’s about giving those words the space, the resonance, and the repetition they need to be truly understood.

Say it once to be heard; say it twice to be understood.

Categories
Computers FORTH IBM Programming

The Architecture of the Stack

Back in the early 1980’s when I worked for IBM, I was able to acquire my own IBM PC and experience my own form digital frontierism. Today I really wish I had a logbook at hand with a record of everything I did as my ability to recall those details has faded with age. A couple of those memories that still do remain with me involve two obscure languages: APL and FORTH. And then there was Borland Turbo Pascal.

In those early days of the 1980’s, memory wasn’t an infinite field; it was a precious, finite resource. While most of us were content living with the structured guardrails of BASIC, there was a subset of us drawn to the elegant, stripped-back world of FORTH.

Learning FORTH felt less like coding and more like learning a new way to breathe. It was lean. It was efficient. It stripped away the overhead of high-level syntax until it was just you, the dictionary, and the stack. There was an honesty to it—no hidden abstractions, just a direct conversation with the hardware.

Then, of course, there was the hurdle of Reverse Polish Notation (RPN). Grokking the stack meant rewiring your brain. You couldn’t just state an operation; you had to prepare the world for it first. You pushed your data onto the stack, one piece at a time, and only then did you call the action. It was a rhythmic, almost percussive way of thinking: Input, input, act.

“In FORTH, you don’t just write programs; you build a language to solve the problem.”

This “bottom-up” philosophy changed the relationship between the creator and the machine. You weren’t just a user; you were an architect of your own vocabulary. To define a new “word” in FORTH was to permanently expand the capabilities of your environment. It was a recursive journey where every small success became a building block for the next complexity.

Looking back, those days with the IBM PC and the stack weren’t just about efficiency. They were about the discipline of clarity. When resources are limited, your thinking must be precise. The difficulty of RPN wasn’t a bug—it was a feature that forced you to understand the flow of data at its most fundamental level.

Categories
Business

The Geometry of Focus: Finding the Limiting Factor

In the modern landscape of high-stakes management, there is a recurring temptation to solve everything at once. We are taught to optimize across the board—to improve efficiency by 2% here, 5% there—until the entire machine hums. But in a recent conversation with John Collison and Dwarkesh Patel, Elon Musk repeatedly returned to a single, almost obsessive mantra: the “limiting factor.”

It is a deceptively simple phrase. It suggests that at any given moment, there is one specific bottleneck that dictates the speed of the entire enterprise. If you aren’t working on that, you aren’t really moving the needle. You are merely polishing stuff.

“I think people are going to have real trouble turning on like the chip output will exceed the ability to turn chips on… the current limiting factor that I see… in the one-year time frame it’s energy power production.”

Musk’s management technique is not about broad oversight; it is about a radical, almost violent prioritization. He looks at the timeline—one year, three years, ten years—and asks: What is the wall we are about to hit? Right now, it might be the availability of GPUs. In twelve months, it might be the physical gigawatts of electricity required to plug them in. In thirty-six months, it might be the thermal constraints of Earth’s atmosphere, necessitating a move to space.

This approach requires a high “pain threshold.” To solve a limiting factor, you often have to lean into acute, short-term struggle to avoid the chronic, slow death of stagnation. John Collison noted this during the interview:

“Most people are willing to endure any amount of chronic pain to avoid acute pain… it feels like a lot of the cases we’re talking about are just leaning into the acute pain… to actually solve the bottleneck.”

For many leaders, the “limiting factor” is often something they aren’t even looking at because it lies outside their perceived domain. A software CEO might think their limit is talent, when it’s actually the speed of their internal decision-making. A manufacturer might think it’s raw materials, when it’s actually the morale of the factory floor.

To manage by the limiting factor is to admit that 90% of what you could be doing is a distraction. It is a philosophy of subtraction and focus. It demands that we stop asking “What can we improve?” and start asking “What is stopping us from being ten times larger?” Once you identify that wall, you throw every resource you have at it until it crumbles. And then—and this is the part that requires true stamina—you immediately go looking for the next wall.

By focusing on the one thing that matters, we stop being busy and start being effective. We stop managing the status quo and start engineering what may feel like the impossible.

Categories
Living Productivity

The Ghost in the Calendar

We have become architects of our own incarceration, building prisons out of thirty-minute blocks and color-coded labels. We operate under a modern delusion: that a gap in the schedule is a leak in the ship. If we aren’t “doing,” we must be failing.

We treat our minds like high-performance engines that must never idle, forgetting that an engine constantly redlining eventually catches fire. Morgan Housel captures this paradox perfectly in Same as Ever:

“The most efficient calendar in the world—one where every minute is packed with productivity—comes at the expense of curious wandering and uninterrupted thinking, which eventually become the biggest contributors to success.”

The tragedy of the “most efficient calendar” is that it optimizes for the visible while starving the invisible. Productivity, in its most common definition, is about throughput—how many emails were sent, how many tickets were closed, how many boxes were checked. But these are administrative victories, not intellectual ones.

When we eliminate “curious wandering,” we eliminate the serendipity required for breakthrough. A breakthrough is rarely the result of a scheduled task; it is the byproduct of a mind allowed to roam until it trips over a connection it wasn’t looking for. By packing every minute, we ensure we are always busy, but we also ensure we are never surprised.

Uninterrupted thinking requires a certain level of inefficiency. It looks like staring out a window, taking a walk without a podcast, or sitting with a problem long after the “allocated” time has expired. In the eyes of a traditional manager—or our own internal critic—this looks like waste. Yet, this “waste” is the soil in which high-leverage ideas grow.

If we lose the ability to wander, we lose our edge. We become mere processors of information rather than creators of value. Real success isn’t found in the frantic filling of space, but in the courage to leave space empty, trusting that the silence will eventually speak.

Categories
Books Productivity

Slow Productivity – in Retirement

Computer science professor Cal Newport has authored a new book titled “Slow Productivity: The Lost Art of Accomplishment without Burnout“.

In a New Yorker article a couple of months ago, Newport wrote of a climate of “Great Exhaustion” in the workplace – a “vibe of weary disappointment”. He talks about a side effect of the covid pandemic being a greater quantity of digital communication (email, chat, video, etc.) and goes on to relate how recent research correlates an increase in this kind of digital communication with decreased satisfaction.

A big part of the problem seems to be the endless context switching these digital tools can create generating significant fatigue and feelings of overload. Our “work becomes inescapable” to use his words.

Of course, this has its worst impact on us while we’re still working and has its least impact after we are retired. Retirement may introduce other stresses – often creating feelings of guilt from not bringing in an income or grief from a work life left behind. But even in retirement we can become overwhelmed and might benefit from his recommendations on dealing with the issue.

His primary prescription in this article for this problem is to take steps to dramatically reduce these digital communication intrusions into our lives by using techniques as simple as scheduling office hours for dealing in bulk with needed interactions between individuals or small teams. In other words, slow down the impact of the interruptions and batch them up for efficiency. In other words, “slow productivity”. That slowness helping to improve our efficiency thereby producing better outcomes and increase satisfaction.

Of course, even small steps like these require cooperation and common understanding among the participants to make an impact. One of the challenges with email in particular is how someone else can insert themselves into your train of thought completely without your agreement.

As Newport reminds us, “Achievement is most satisfying when it arises from a sense of control and purpose, not from the pressure of constant busyness.” It’s time to reclaim our time and rediscover the joy of getting things done, not just being busy doing them or, worse, constantly having to shift our focus in reaction to the tools in our digital lives.