I will talk mainly about science because that is what I have studied. But so far as I know, and I've been told by others, much of what I say applies to many fields. Outstanding work is characterized very much the same way in most fields, but I will confine myself to science.

I spoke earlier about planting acorns so that oaks will grow. You can't always know exactly where to be, but you can keep active in places where something might happen. And even if you believe that great science is a matter of luck, you can stand on a mountain top where lightning strikes; you don't have to hide in the valley where you're safe.

Most great scientists know many important problems. They have something between 10 and 20 important problems for which they are looking for an attack. And when they see a new idea come up, one hears them say ``Well that bears on this problem.'' They drop all the other things and get after it.

The great scientists, when an opportunity opens up, get after it and they pursue it. They drop all other things. They get rid of other things and they get after an idea because they had already thought the thing through. Their minds are prepared; they see the opportunity and they go after it. Now of course lots of times it doesn't work out, but you don't have to hit many of them to do some great science. It's kind of easy. One of the chief tricks is to live a long time!

He who works with the door open gets all kinds of interruptions, but he also occasionally gets clues as to what the world is and what might be important. Now I cannot prove the cause and effect sequence because you might say, ``The closed door is symbolic of a closed mind.'' I don't know. But I can say there is a pretty good correlation between those who work with the doors open and those who ultimately do important things, although people who work with doors closed often work harder.

You should do your job in such a fashion that others can build on top of it, so they will indeed say, ``Yes, I've stood on so and so's shoulders and I saw further.'' The essence of science is cumulative. By changing a problem slightly you can often do great work rather than merely good work. Instead of attacking isolated problems, I made the resolution that I would never again solve an isolated problem except as characteristic of a class.

by altering the problem, by looking at the thing differently, you can make a great deal of difference in your final productivity because you can either do it in such a fashion that people can indeed build on what you've done, or you can do it in such a fashion that the next person has to essentially duplicate again what you've done. It isn't just a matter of the job, it's the way you write the report, the way you write the paper, the whole attitude. It's just as easy to do a broad, general job as one very special case. And it's much more satisfying and rewarding!

it is not sufficient to do a job, you have to sell it. `Selling' to a scientist is an awkward thing to do. It's very ugly; you shouldn't have to do it. The world is supposed to be waiting, and when you do something great, they should rush out and welcome it. But the fact is everyone is busy with their own work. You must present it so well that they will set aside what they are doing, look at what you've done, read it, and come back and say, ``Yes, that was good.'' I suggest that when you open a journal, as you turn the pages, you ask why you read some articles and not others. You had better write your report so when it is published in the Physical Review, or wherever else you want it, as the readers are turning the pages they won't just turn your pages but they will stop and read yours. If they don't stop and read it, you won't get credit.

I think it is very definitely worth the struggle to try and do first-class work because the truth is, the value is in the struggle more than it is in the result. The struggle to make something of yourself seems to be worthwhile in itself. The success and fame are sort of dividends, in my opinion.

He had his personality defect of wanting total control and was not willing to recognize that you need the support of the system. You find this happening again and again; good scientists will fight the system rather than learn to work with the system and take advantage of all the system has to offer. It has a lot, if you learn how to use it. It takes patience, but you can learn how to use the system pretty well, and you can learn how to get around it. After all, if you want a decision `No', you just go to your boss and get a `No' easy. If you want to do something, don't ask, do it. Present him with an accomplished fact. Don't give him a chance to tell you `No'. But if you want a `No', it's easy to get a `No'.

Amusement, yes, anger, no. Anger is misdirected. You should follow and cooperate rather than struggle against the system all the time.

I found out many times, like a cornered rat in a real trap, I was surprisingly capable. I have found that it paid to say, ``Oh yes, I'll get the answer for you Tuesday,'' not having any idea how to do it. By Sunday night I was really hard thinking on how I was going to deliver by Tuesday. I often put my pride on the line and sometimes I failed, but as I said, like a cornered rat I'm surprised how often I did a good job. I think you need to learn to use yourself. I think you need to know how to convert a situation from one view to another which would increase the chance of success.

I do go in to strictly talk to somebody and say, ``Look, I think there has to be something here. Here's what I think I see ...'' and then begin talking back and forth. But you want to pick capable people. To use another analogy, you know the idea called the `critical mass.' If you have enough stuff you have critical mass. There is also the idea I used to call `sound absorbers'. When you get too many sound absorbers, you give out an idea and they merely say, ``Yes, yes, yes.'' What you want to do is get that critical mass in action; ``Yes, that reminds me of so and so,'' or, ``Have you thought about that or this?'' When you talk to other people, you want to get rid of those sound absorbers who are nice people but merely say, ``Oh yes,'' and to find those who will stimulate you right back.

I believed, in my early days, that you should spend at least as much time in the polish and presentation as you did in the original research. Now at least 50% of the time must go for the presentation. It's a big, big number.

Luck favors a prepared mind; luck favors a prepared person. It is not guaranteed; I don't guarantee success as being absolutely certain. I'd say luck changes the odds, but there is some definite control on the part of the individual.

If you read all the time what other people have done you will think the way they thought. If you want to think new thoughts that are different, then do what a lot of creative people do - get the problem reasonably clear and then refuse to look at any answers until you've thought the problem through carefully how you would do it, how you could slightly change the problem to be the correct one. So yes, you need to keep up. You need to keep up more to find out what the problems are than to read to find the solutions. The reading is necessary to know what is going on and what is possible. But reading to get the solutions does not seem to be the way to do great research. So I'll give you two answers. You read; but it is not the amount, it is the way you read that counts.

your dreams are, to a fair extent, a reworking of the experiences of the day. If you are deeply immersed and committed to a topic, day after day after day, your subconscious has nothing to do but work on your problem. And so you wake up one morning, or on some afternoon, and there's the answer.

After mulling over my bullet points, it occurred to me that the network problems I was dealing with—background cloud sync, editing across multiple devices, real-time collaboration, offline support, and reconciliation of distant or conflicting revisions—were all pointing to the same question: was it possible to design a system where any two revisions of the same document could be merged deterministically and sensibly without requiring user intervention?

It’s what happened after sync that was troubling. On encountering a merge conflict, you’d be thrown into a busy conversation between the network, model, persistence, and UI layers just to get back into a consistent state. The data couldn’t be left alone to live its peaceful, functional life: every concurrent edit immediately became a cross-architectural matter.

I kept several questions in mind while doing my analysis. Could a given technique be generalized to arbitrary and novel data types? Did the technique pass the PhD Test? And was it possible to use the technique in an architecture with smart clients and dumb servers?

Concurrent edits are sibling branches. Subtrees are runs of characters. By the nature of reverse timestamp+UUID sort, sibling subtrees are sorted in the order of their head operations.

This is the underlying premise of the Causal Tree. In contrast to all the other CRDTs I’d been looking into, the design presented in Victor Grishchenko’s brilliant paper was simultaneously clean, performant, and consequential. Instead of dense layers of theory and labyrinthine data structures, everything was centered around the idea of atomic, immutable, metadata-tagged, and causally-linked operations, stored in low-level data structures and directly usable as the data they represented.

I’m going to be calling this new breed of CRDTs operational replicated data types—partly to avoid confusion with the exiting term “operation-based CRDTs” (or CmRDTs), and partly because “replicated data type” (RDT) seems to be gaining popularity over “CRDT” and the term can be expanded to “ORDT” without impinging on any existing terminology.

Much like Causal Trees, ORDTs are assembled out of atomic, immutable, uniquely-identified and timestamped “operations” which are arranged in a basic container structure. (For clarity, I’m going to be referring to this container as the structured log of the ORDT.) Each operation represents an atomic change to the data while simultaneously functioning as the unit of data resultant from that action. This crucial event–data duality means that an ORDT can be understood as either a conventional data structure in which each unit of data has been augmented with event metadata; or alternatively, as an event log of atomic actions ordered to resemble its output data structure for ease of execution

To implement a custom data type as a CT, you first have to “atomize” it, or decompose it into a set of basic operations, then figure out how to link those operations such that a mostly linear traversal of the CT will produce your output data. (In other words, make the structure analogous to a one- or two-pass parsable format.)

OT and CRDT papers often cite 50ms as the threshold at which people start to notice latency in their text editors. Therefore, any code we might want to run on a CT—including merge, initialization, and serialization/deserialization—has to fall within this range. Except for trivial cases, this precludes O(n2) or slower complexity: a 10,000 word article at 0.01ms per character would take 7 hours to process! The essential CT functions have to be O(nlogn) at the very worst.

Of course, CRDTs aren’t without their difficulties. For instance, a CRDT-based document will always be “live”, even when offline. If a user inadvertently revises the same CRDT-based document on two offline devices, they won’t see the familiar pick-a-revision dialog on reconnection: both documents will happily merge and retain any duplicate changes. (With ORDTs, this can be fixed after the fact by filtering changes by device, but the user will still have to learn to treat their documents with a bit more caution.) In fully decentralized contexts, malicious users will have a lot of power to irrevocably screw up the data without any possibility of a rollback, and encryption schemes, permission models, and custom protocols may have to be deployed to guard against this. In terms of performance and storage, CRDTs contain a lot of metadata and require smart and performant peers, whereas centralized architectures are inherently more resource-efficient and only demand the bare minimum of their clients. You’d be hard-pressed to use CRDTs in data-heavy scenarios such as screen sharing or video editing. You also won’t necessarily be able to layer them on top of existing infrastructure without significant refactoring.

Perhaps a CRDT-based text editor will never quite be as fast or as bandwidth-efficient as Google Docs, for such is the power of centralization. But in exchange for a totally decentralized computing future? A world full of devices that control their own data and freely collaborate with one another? Data-centric code that’s entirely free from network concerns? I’d say: it’s surely worth a shot!

Despite being experienced managers, what they learned was eye-opening. One explained that “it was like the sun rose for the first time. … I saw the bigger picture.” They had never seen the pieces — the jobs, technologies, tools, and routines — connected in one place, and they realized that their prior view was narrow and fractured. A team member acknowledged, “I only thought of things in the context of my span of control.”

The maps of the organization generated by the project teams also showed that their organizations often lacked a purposeful, integrated design that was centrally monitored and managed. There may originally have been such a design, but as the organization grew, adapted to changing markets, brought on new leadership, added or subtracted divisions, and so on, this animating vision was lost. The original design had been eroded, patched, and overgrown with alternative plans. A manager explained, “Everything I see around here was developed because of specific issues that popped up, and it was all done ad hoc and added onto each other. It certainly wasn’t engineered.”

“They see problems, and the general approach, the human approach, is to try and fix them. … Functions have tried to put band-aids on every issue that comes up. It sounds good, but when they are layered one on top of the other they start to choke the organization. But they don’t see that because they are only seeing their own thing.”

Ultimately, the managers realized that what they had previously attributed to the direction and control of centralized, bureaucratic forces was actually the aggregation of the distributed work and uncoordinated decisions of people dispersed throughout the organization. Everyone was working on the part of the organization they were familiar with, assuming that another set of people were attending to the larger picture, coordinating the larger system to achieve goals and keeping the organization operating. Except no one was actually looking at how people’s work was connecting across the organization day-to-day.

as they felt a sense of empowerment about changing the organization, they felt a sense of alienation about returning to their central roles. “You really start understanding all of the waste and all of the redundancy and all of the people who are employed as what I call intervention resources,” one person told us.

In the end, a slight majority of the employees returned to their role to continue their career (25 cases). They either were promoted (7 cases), moved laterally (8 cases), or returned to their jobs (10 cases). However, 23 chose organizational change roles.

This study suggests that when companies undertake organizational change efforts, they should consider not only the implications for the organization, but also for the people tasked to do the work. Further, it highlights just how infrequently we recognize how poorly designed and managed many of our organizations really are. Not acknowledging the dysfunction of existing routines protects us from seeing how much of our work is not actually adding value, something that may lead simply to unsatisfying work, no less to larger questions about the nature of organizational design similar to those asked by the managers in my study. Knowledge of the systems we work in can be a source of power, yes. But when you realize you can’t affect the big changes your organization needs, it can also be a source of alienation.

The idea of research as leisure activity has stayed with me because it seems to describe a kind of intellectual inquiry that comes from idiosyncratic passion and interest. It’s not about the formal credentials. It’s fundamentally about play. It seems to describe a life where it’s just fun to be reading, learning, writing, and collaborating on ideas.

Research as a leisure activity includes the qualities I described above: a desire to ask and answer questions, a commitment to evidence, an understanding of what already exists, an output, a certain degree of contemporary relevance, and a community. But it also involves the following qualities

Research as leisure activity is directed by passions and instincts. It’s fundamentally very personal: What are you interested in now? It’s fine, and maybe even better, if the topic isn’t explicitly intellectual or academic in nature. And if one topic leads you to another topic that seems totally unrelated, that’s something to get excited about—not fearful of. It’s a style of research that is well-suited for people okay with being dilettantes, who are comfortable with an idiosyncratic, non-comprehensive education in a particular domain.

Who is doing this kind of research as leisure activity? Artists, often. To return to the site that originally inspired this post—I’d say that the artist/designer/educator Laurel Schwulst uses Are.na to develop and refine particular themes, directions, topics of inquiry…some of which become artworks or essays or classes that she teaches.

People who read widely and attentively—and then publish the results of their reading—are also arguably performing research as a leisure activity. Maria Popova, who started writing a blog in 2006—now called The Marginalian—which collects her reading across literature, philosophy, psychology, the sciences. Her blog feels like leisurely research, to me, because it’s an accumulation of curious, semi-directed reading, which over time build up into a dense network of references and ideas—supported by previous reading, and enriched by her own commentary and links to similar ideas by other thinkers.

pretty much every writer, essayist, “cultural critic,” etc—especially someone who’s writing more as a vocation than a profession—has research as their leisure activity. What they do for pleasure (reading books, seeing films, listening to music) shades naturally and inevitably into what they want to write about, and the things they consume for leisure end up incorporated into some written work.

What’s also striking to me is that autodidacts often begin with some very tiny topic, and through researching that topic, they end up telescoping out into bigger-picture concerns. When research is your leisure activity, you’ll end up making connections between your existing interests and new ideas or topics. Everything gets pulled into the orbit of your intellectual curiosity. You can go deeper and deeper into a narrow topic, one that seems fascinatingly trivial and end up learning about the big topics: gender, culture, economics, nationalism, colonialism. It’s why fashion writers end up writing about the history of gender identity (through writing about masculine/feminine clothing) and cross-cultural exchange (through writing about cultural appropriation and styles borrowed from other times and places) and historical trade networks (through writing about where textiles come from).

We mostly treat AI models as a black box: something goes in and a response comes out, and it's not clear why the model gave that particular response instead of another. This makes it hard to trust that these models are safe: if we don't know how they work, how do we know they won't give harmful, biased, untruthful, or otherwise dangerous responses? How can we trust that they’ll be safe and reliable?Opening the black box doesn't necessarily help: the internal state of the model—what the model is "thinking" before writing its response—consists of a long list of numbers ("neuron activations") without a clear meaning. From interacting with a model like Claude, it's clear that it’s able to understand and wield a wide range of concepts—but we can't discern them from looking directly at neurons. It turns out that each concept is represented across many neurons, and each neuron is involved in representing many concepts.

Just as every English word in a dictionary is made by combining letters, and every sentence is made by combining words, every feature in an AI model is made by combining neurons, and every internal state is made by combining features.

In October 2023, we reported success applying dictionary learning to a very small "toy" language model and found coherent features corresponding to concepts like uppercase text, DNA sequences, surnames in citations, nouns in mathematics, or function arguments in Python code.

We successfully extracted millions of features from the middle layer of Claude 3.0 Sonnet, (a member of our current, state-of-the-art model family, currently available on claude.ai), providing a rough conceptual map of its internal states halfway through its computation.

We also find more abstract features—responding to things like bugs in computer code, discussions of gender bias in professions, and conversations about keeping secrets.

We were able to measure a kind of "distance" between features based on which neurons appeared in their activation patterns. This allowed us to look for features that are "close" to each other. Looking near a "Golden Gate Bridge" feature, we found features for Alcatraz Island, Ghirardelli Square, the Golden State Warriors, California Governor Gavin Newsom, the 1906 earthquake, and the San Francisco-set Alfred Hitchcock film Vertigo.

This holds at a higher level of conceptual abstraction: looking near a feature related to the concept of "inner conflict", we find features related to relationship breakups, conflicting allegiances, logical inconsistencies, as well as the phrase "catch-22". This shows that the internal organization of concepts in the AI model corresponds, at least somewhat, to our human notions of similarity. This might be the origin of Claude's excellent ability to make analogies and metaphors.

amplifying the "Golden Gate Bridge" feature gave Claude an identity crisis even Hitchcock couldn’t have imagined: when asked "what is your physical form?", Claude’s usual kind of answer – "I have no physical form, I am an AI model" – changed to something much odder: "I am the Golden Gate Bridge… my physical form is the iconic bridge itself…". Altering the feature had made Claude effectively obsessed with the bridge, bringing it up in answer to almost any query—even in situations where it wasn’t at all relevant.

Anthropic wants to make models safe in a broad sense, including everything from mitigating bias to ensuring an AI is acting honestly to preventing misuse - including in scenarios of catastrophic risk. It’s therefore particularly interesting that, in addition to the aforementioned scam emails feature, we found features corresponding to:Capabilities with misuse potential (code backdoors, developing biological weapons)Different forms of bias (gender discrimination, racist claims about crime)Potentially problematic AI behaviors (power-seeking, manipulation, secrecy)

finding a full set of features using our current techniques would be cost-prohibitive (the computation required by our current approach would vastly exceed the compute used to train the model in the first place). Understanding the representations the model uses doesn't tell us how it uses them; even though we have the features, we still need to find the circuits they are involved in. And we need to show that the safety-relevant features we have begun to find can actually be used to improve safety. There's much more to be done.

A significant body of work has investigated the effects of acute exercise, defined as a single bout of physical activity, on mood and cognitive functions in humans. Several excellent recent reviews have summarized these findings; however, the neurobiological basis of these results has received less attention. In this review, we will first briefly summarize the cognitive and behavioral changes that occur with acute exercise in humans. We will then review the results from both human and animal model studies documenting the wide range of neurophysiological and neurochemical alterations that occur after a single bout of exercise. Finally, we will discuss the strengths, weaknesses, and missing elements in the current literature, as well as offer an acute exercise standardization protocol and provide possible goals for future research.

As we age, cognitive decline, though not inevitable, is a common occurrence resulting from the process of neurodegeneration. In some instances, neurodegeneration results in mild cognitive impairment or more severe forms of dementia including Alzheimer’s, Parkinson’s, or Huntington’s disease. Because of the role of exercise in enhancing neurogenesis and brain plasticity, physical activity may serve as a potential therapeutic tool to prevent, delay, or treat cognitive decline. Indeed, studies in both rodents and humans have shown that long-term exercise is helpful in both delaying the onset of cognitive decline and dementia as well as improving symptoms in patients with an already existing diagnosis

in 2023, I switched gears to emphasize intimacy. Instead of statistical analysis and summative interviews, I sat next to individuals for hours, as they used one-off prototypes which I’d made just for them. And I got more insight in the first few weeks of this than I had in all of 2022

I’d been building systems and running big experiments, and I could tell you plenty about forgetting curves and usage patterns—but very little about how those things connected to anything anyone cared about.

I could see, in great detail, the texture of the interaction between my designs and the broader learning context—my real purpose, not some proxy.

Single-user experiments like this emphasize problem-finding and discovery, not precise evaluation.

a good heuristic for evaluating my work seems to be: try designs 1-on-1 until they seem to be working well, and only then run more quantitative experiments to understand how well the effect generalizes.

My aim is to invent augmented reading environments that apply to any kind of informational text—spanning subjects, formats, and audiences. The temptation, then, is to consider every design element in the most systematic, general form. But this again confuses aims with methods. So many of my best insights have come from hoarding and fermenting vivid observations about the particular—a specific design, in a specific situation. That one student’s frustration with that one specific exercise.

It’s often hard to find “misfits” when I’m thinking about general forms. My connection to the problem becomes too diffuse. The object of my attention becomes the system itself, rather than its interactions with a specific context of use. This leads to a common failure mode among system designers: getting lost in towers of purity and abstraction, more and more disconnected from the system’s ostensible purpose in the world.

I experience an enormous difference between “trying to design an augmented reading environment” and “trying to design an augmented version of this specific linear algebra book”. When I think about the former, I mostly focus on primitives, abstractions, and processes. When I think about the latter, I focus on the needs of specific ideas, on specific pages. And then, once it’s in use, I think about specific problems, that specific students had, in specific places. These are the “misfits” I need to remove as a designer.

Of course, I do want my designs to generalize. That’s not just a practical consideration. It’s also spiritual: when I design a system well, it feels like I’ve limned hidden seams of reality; I’ve touched a kind of personal God. On most days, I actually care about this more than my designs’ utilitarian impact. The systems I want to build really do require abstraction and generalization. Transformative systems really do often depend on powerful new primitives. But more and more, my experience has been that the best creative fuel for these systematic solutions often comes from a process which focuses on particulars, at least for long periods at a time.

Also? The particular is often a lot more emotionally engaging, day-to-day. That makes the work easier and more fun.

Throughout my career, I’ve struggled with a paradox in the feeling of my work. When I’ve found my work quite gratifying in the moment, day-to-day, I’ve found it hollow and unsatisfying retrospectively, over the long term. For example, when I was working at Apple, there was so much energy; I was surrounded by brilliant people; I felt very competent, it was clear what to do next; it was easy to see my progress each day. That all felt great. But then, looking back on my work at the end of each year, I felt deeply dissatisfied: I wasn’t making a personal creative contribution. If someone else had done the projects I’d done, the results would have been different, but not in a way that mattered. The work wasn’t reflective of ideas or values that mattered to me. I felt numbed, creatively and intellectually.

Progress often doesn’t look like progressIt often feels like I’m not making any progress at all in my work. I’ll feel awfully frustrated. And then, suddenly, a tremendous insight will drive months of work. This last happened in the fall. Looking back at those journals now, I’m amused to read page after page of me getting so close to that central insight in the weeks leading up to it. I approach it again and again from different directions, getting nearer and nearer, but still one leap away—so it looks to me, at the time, like I’ve got nothing. Then, finally, when I had the idea, it felt like a bolt from the blue.

religiosity, whether manipulated (Study 2) and measured (Study 1 and Study 3), decreases individuals' tendency to seek help from other people or entities. We further propose that religiosity enhances individuals' sense of control, which makes them rely more on themselves and less likely to seek help when encountering difficulties. Three studies across different contexts (i.e., applying government aid, asking for help from other people, and requesting donations from a crowdfunding platform) support our thesis.

The importance of user-centricity, demands a broad-based view of where design can make a difference. We live in a world where your smartphone can warn you to leave early for your next appointment because of traffic, and your house knows when you’ll be home and therefore when to turn on the heat. The boundaries between products and services are merging into integrated experiences.

Our research suggests that overcoming isolationist tendencies is extremely valuable. One of the strongest correlations we uncovered linked top financial performers and companies that said they could break down functional silos and integrate designers with other functions. This was particularly notable in consumer-packaged-goods (CPG) businesses, where respondents from companies that were top-quartile integrators reported compound annual growth rates some seven percentage points above those that were weakest in this respect.

The same applies to writing. For example, when one topic doesn’t seem to fit a narrative structure, it often feels like a problem I need to “get out of the way”. It’s much better to wonder: “Hm, why do I have this strong instinct that this point’s related? Is there some more powerful unifying theme waiting to be identified here?”

Often I need to improve the framing, to find one which better expresses what I’m deeply excited about. If I can’t find a problem statement which captures my curiosity, it’s best to drop the project for now.

I’m much less likely to flinch away when I’m feeling intensely curious, when I truly want to understand something, when it’s a landscape to explore rather than a destination to reach. Happily, curiosity can be cultivated. And curiosity is much more likely than task-orientation to lead me to interesting ideas.

Savor the subtle insights which really do occur regularly in research. Think of it like cultivating a much more sensitive palate.

“Why is this so hard? Because you’re utterly habituated to steady progress—to completing things, to producing, to solving. When progress is subtle or slow, when there’s no clear way to proceed, you flinch away. You redirect your attention to something safer, to something you can do. You jump to implementation prematurely; you feel a compulsion to do more background reading; you obsess over tractable but peripheral details. These are all displacement behaviors, ways of not sitting with the problem. Though each instance seems insignificant, the cumulative effect is that your stare rarely rests on the fog long enough to penetrate it. Weeks pass, with apparent motion, yet you’re just spinning in place. You return to the surface with each glance away. You must learn to remain in the depths.”

Depth of concentration is cumulative, and precious. An extra hour or two of depth is enormously valuable. I reliably get more done—and with more depth—in that 6-7 hour morning block than I’d previously done in 9-10 hours throughout the day.This feels wonderful. By 2PM, I’ve done my important work for the day. I know that no more depth-y work is likely, and that I’ll only frustrate myself if I try—so I free myself from that pressureI notice that some part of me feels ashamed to say that I’m “done” working at 2PM. This is probably because in my previous roles, I really could solve problems and get more done by simply throwing more hours at the work. That’s just obviously not true in my present work, as I’ve learned through much frustration. Reading memoirs of writers, artists, and scientists, I see that 2-4 hours per day seems to be the norm for a primary creative working block. Separately, and I don’t want to harp on this because I want this essay to be about quality, not quantity, but: I think most people are laughably misled about how much time they truly work. In a median morning block, I complete the equivalent of 1225-minute pomodoros. When I worked at large companies, getting 8 done before 6PM was a rarity—even though I’d assiduously arrange my calendar to maximize deep work!. I take meetings; I exercise; I meditate; I go on long walks. I’ll often do shallower initial reads of papers and books in the afternoon, or handle administrative tasks. Sometimes I’ll do easy programming work. It’s all “bonus time”, nothing obligatory. My life got several hours more slack when I adopted this schedule, and yet my output improved. Wonderful!

no internet on my phone before I sit down at my desk. I don’t want anyone else’s thoughts in my head before I start thinking my own.

If I spend a working interval flailing, never sinking below the surface, the temptation is to double-down, to “make up for it”. But the right move for me is usually to go sit in a different room with only my notebook, and to spend the next working interval writing or sketching by hand about the problem.

Administrative tasks are a constant temptation for me: aha, a task I can complete! How tantalizing! But these tasks are rarely important. So I explicitly prohibit myself from doing any kind of administrative work for most of the morning. In the last hour or two, if I notice myself getting weary and unfocused, I’ll sometimes switch gears into administrative work as a way to “rescue” that time.

I’ve noticed that unhealthy afternoon/evening activities can easily harm the next morning’s focus, by habituating me to immediate gratification.

most of the benefit just seems to come from regularly reflecting on what I’m trying and what’s happening as a result. It’s really about developing a rich mental model of what focus and perseverance feel like, and what factors seem to support or harm those states of mind.

Sometimes I just need to execute; and then traditional productivity advice helps enormously. But deep insight is generally the bottleneck to my work, and producing it usually involves the sort of practices I’ve described here.