Daemons and the Mindful Learning Curve

Humans naturally think about their own behaviors in terms of peak and trough performance levels, rather than means or medians. Without any performance tracking, we know our limits in a variety of domains. Each time we attempt a performance episode in any skilled domain, these limits change, yielding a learning curve. It has a characteristic tempo, depending on the episodic performance quality, as well as a tempo across episodes, that is correlated with quality. I prefer this model of a learning curve that I made up to the usual smooth, S-shaped one with a plateau at the end. I’ll explain why in a bit. This picture is related to what I called the Freytag Staircase in the book, but is not quite the same.

Peaks and Troughs

The fastest I’ve ever run is an 8-minute mile. On my worst days, if I can get off the couch at all, I struggle to finish a 15-minute mile. In terms of endurance, the most I’ve ever run is 9 miles. On my worst drays I can barely run half a mile.

On my best days, I can write thousands of words of near-flawless prose (by my standards that is: stuff I am happy to post without further revision). On my worst days, I can barely squeeze out a couple of hundred words. And they tend not to be very good. I often endlessly revise the words and still end up throwing them away in disgust.

These two regimes: peak performance and trough performance, are very different psychologically. The former is about an effortless, stable tempo; what positive psychologists call flow. It is about letting go and maintaining a light and gentle control on exuberant energy. The latter is about an unstable, high-effort tempo that never seems to settle. It is the opposite of flow. It is about grit, pain and endurance, and consciously willing yourself to keep going. Since the intrinsic rhythms are unstable, you may need extrinsic ones: like music, a book or just telling yourself, “just another minute, just another tenth of a mile, and I’ll stop…. how about another 10 strides…”

Discipline and Daemons

I used to think discipline was about machine-like predictable performance. Now I believe discipline seems to be about showing up like clockwork, and accepting whatever happens: peak performance,  trough performance or anything in-between. And then showing up again. Elizabeth Gilbert made this point very neatly by conceptualizing the actual performance level as the effect of whether or not a creative “daemon” visits you on that particular day. I prefer Gilbert’s model to Woody Allen’s more cryptic line, “90% of success is just showing up,” because it clarifies the link between show-up discipline and actual outcomes.

For us mere mortals, there will always be a gap between peak performance and trough performance. So it is easy to fall into the trap of only focusing on one or the other.

Those who focus only on the peak get addicted to the highs that peaks bring. They abandon discipline and only attempt performance episodes when they feel the “flow” coming on. Unfortunately, without disciplined and regular “showing up,” these events become increasingly rare. Gilbert’s daemon is not predictable but it expects you to be.

Those who focus only on showing up, gritting their teeth, and pushing through, fail in a different way. They are so exhausted by the trough days that on days when things seem a little easier — a sign that the daemon is on its way —  they cannot let go and hit a new peak. Instead, they either lose the flow through too much control, or worse, treat the relative ease of the flow state as an opportunity to relax; a break from the unending series of trough episodes. In other words, they either strangle or ignore the daemon, leading to the same result: the daemon visits get increasingly rare.

I prefer to play it by the ear. If it seems like the daemon is present, I relinquish control and let it take me to whatever peak it can that day. If the daemon appears to be absent, I view it as a training exercise in grit and resilience.

This isn’t a deterministic decision. Sometimes I make Type I errors: trying to relinquish control when there is no daemon around to take over. Other days I make Type II errors: misreading the signs of a daemon on its way, and doing the grit thing instead.


When you manage to hit this mindset, both peaks and troughs start to rise. That’s at the heart of it.

Of course, there is more going on. You become more mindful in the activity, there is deliberate practice going on, increasing both conscious and unconscious skill levels.  You are becoming aware of dimensions of the activity you were not aware of before. In short, you are learning.

But at the heart of this is simply this management of the tempo of performance evolution through its peaks and troughs.

This is a pretty difficult meta-cognition skill to acquire.

In fact it is as hard as another activity it resembles: managing the growth rate and income inequality in an economy. This connection should not surprise you. Both growing a country’s wealth and learning a skill are creative-destruction processes. Societies start to crumble if income inequality (the Gini coefficient) increases too much. But on the other hand, artificially forcing gaps to zero (idealized communism) kills growth.  Good governance is about letting the daemon — capitalism — take control sometimes, and keeping things going via regulation at other times, and knowing when to invoke which governance behavior.

There is something like an economy of performers in your life. Every day you are a different person. Some days you are wealthy with flow, other days you are impoverished.

Many people are good at managing episodic tempo on either end of the spectrum. They might be good at peak performance or trough performance, or even both, but are bad at choosing the right tempo management behavior at the right time, and switching frictionlessly. It’s a mental trick — being able to swap out a peak game-mind for a trough game-mind very quickly.

There is a tempting alternative way of looking at this dynamic, which I found in John Eliot’s interesting, but deeply flawed book, Overachievement. Here you distinguish between “training” and “performance” mindsets and different external behaviors. One is practice with critical review, the other is flow performance with no meta-cognition.

I have concluded that this is a fatally flawed framing. Eliot considers professional sports and professions like neurosurgery. While there definitely is a practice/performance distinction, I have found personally that in most domains, you cannot separate things out that way. You are learning and performing all the time. You’re just learning different things. The peak/trough distinction is more intrinsic and robust.

Traditional Learning Curves

Traditional learning curves are typically smooth and S-shaped.

They are smooth because they typically track only peak or trough performance.

They plateau for a different reason: they use the same performance metric for all levels of performance. This may work at the Olympics, but is deeply misguided in most real-world domains. Performance quality is something you measure differently as skill levels increase.

In cricket for example, fast bowlers value raw pace early in their careers, controlled swing later in their careers, and guile and variation in their mature phase.

If you measure a bowler based purely on raw pace throughout his career, you’ll naturally get a plateau, because you’re not aware of the new levels of the game he has opened up for himself. In a way, he has abandoned diminishing marginal returns on one front for higher returns on another.

But that’s only a rough external view of what’s happening. Performance is not something you measure in terms of effects and external reference points, but in terms of the quality of your unconscious, internalized understanding of what you are doing: unconscious competence.

Growth involves internalizing the natural essence of what you are doing, rather than viewing it in instrumental ways, as a means to an end. You become a better runner as opposed to having a better 100m time.

Instrumentality must be considered here. Sure, in cricket, bowling is merely a means to an end: getting the batsmen out, preventing them from making runs, and winning the game.

But there is also an internal essence of the art of bowling. Good bowlers try to achieve the goals of bowling as an instrumental behavior. Great bowlers start to manifest the essence of bowling.

Closer to my life, far from the demands of high-end professional sports, there is exercise as a means to lost weight and better looks, and exercise as a lifelong journey to manifest the essence of your body and understanding what “healthy” really means. There is writing as a means to books and income, and there is writing as a lifelong journey to become a flow of words.


There is one valid critique of such intrinsic definitions of learning and performance. What about competitive situations. Do mindful bowlers get more wickets? Yield fewer runs? Win more games? Do mindful batsmen make more runs, win more games?

Cricket fans agonize endlessly about such questions. Often, a statistically-minded fan will attempt a Gotcha! argument: “You think player X is great, but if you actually look at the win/loss record, you’ll notice that he’s scored most of his centuries in games we’d have won anyway, it’s player Y who actually delivers to win games.”

At the level of growing the game, you can partially resolve such questions by simply applying the peak/trough performance/learning curve to the team as a whole. You could view great bowlers as innovators and pioneers, and good ones as execution types. Good bowlers win games. Great bowlers advance the game, even if they don’t win as often (though they often do both).

But in a way that’s besides the point. No human-invented game is a good representation of open performance environments with no artificial rules.

What is an open performance environment? It is one where Darwin rules. Where the only measure of success is survival, and there are no rules.

Does mindful performance make you more likely to survive a little longer?

I am not sure. It is one of the questions I am pondering for future editions of the book. There are tricky issues concerning entropy, knowledge and the nature of risk that I haven’t yet worked through.

But I think it is what makes survival worthwhile. Ten years of mindful living is worth twenty years of half-dead living.



Get Ribbonfarm in your inbox

Get new post updates by email

New post updates are sent out once a week

About Tempo


  1. I am impressed! This very clearly elaborates on a number of frustrations I have been experiencing recently. One of those frustrations I have taken to calling “the cult of results”. By that I mean all the people throughout the tech, blogging, design (etc) communities who are blindly obsessed with results and optimization. Yesterday I came across the following and nearly jumped out a window:

    “Never run a marketing campaign you can’t measure!”

    The insight that I hadn’t quite formulated yet is exactly what you describe here – all these people are mindlessly focused on troughs to the point that they optimize away any opportunity to increase peaks. I had been thinking of this in terms of measuring proxies instead of actual ends, but your explanation offers a much more robust framing – they are accepting an episodic peak as an ultimate end.

    I think that leads me to a tie-in to the refinement concept. If we consider your model, not in terms of energy levels, but instead in terms of output quality over time…then movement out of a trough is unrefined work. It is the foundational, big picture stuff that pushes you towards a new episode with greater potential. Once you rise above the level of the next trough you are doing refined work. In other words, you are doing work that increases the quality of output in that specific episode but does not contribute to the foundation of the next episode.

    I think this comports with the paths that various life strategies tend to produce. People who do highly refined work are perpetually pursuing episodic peaks. They look like winners as they are rising up one of those peaks, but they are also the types who tend to experience mid-life crises, crashing the hardest when those episodic peaks evaporate. People doing unrefined work often seem to be perpetually living in troughs, and yet their comparative avoidance of peaks allows them to dedicate more time to activities that contribute to the foundations of future episodes.

    The unrefined path therefore provides a steadier (more resilient?) climb but lacks the high highs. Which one is more successful surely depends on external circumstances and the degree to which peaks pull up the troughs, or vice versa, the degree to which deep troughs pull down the peaks.

    • I think the refined/unrefined mapping will work. It may be the connection I am thinking of to the Freytag staircase (via entropy) maps to refinement.

      But I am curious about why you seem to prefer troughs to peaks in some sense. I think you need both. Without the peaks, the troughs won’t rise (i.e. you need the transient peak between troughs… it’s more than refinement, it is needed to create the next trough…. I don’t think you could have a peak-less trajectory, like a true staircase)

  2. Once you “see” the next trough on your way up, anything too much past it is a purely temporal benefit. Should I perfect combustion cylinders and fuel injection or start designing electric cars?

    Too much of the former and you’re an expert in dead technology. Too much of the latter and you’re chronically unable to cash in on cheap tricks, many of which are valuable for only a short while.

    My interpretation of what Gregory is saying is: “playing not to lose and playing to win big are only opposites on one axis; they’re flip sides of the same coin. Playing to win small and often, continually re-establishing slightly higher baseline floors, often unexpectedly outperforms both and may be more resilient over the long term.”

    Forgive me if I’ve butchered it.

    • I’m sorry, this was meant to be a reply to Venkat’s reply to Gregory.

    • Hmm… if we take this argument all the way, then it suggests that the notion of ‘flow’ is actually a dangerous subjective state to be in, since I associate it more with peak performance (and therefore, if the mapping is correct, with the diminishing marginal returns refinement in whatever your are doing). Or mathematically, you are near a local optimum: in the zone psychologically, and in the end-zone in terms of effectiveness.

      So another way to interpret this is to say that peak episodes are ‘expert’ episodes while trough episodes are ‘beginner’ episodes. Being a beginner is never fun. It feels awkward and unfulfilling. It is the opposite of flow.

      Yet, many Zen practices teach that you should try to be in a “beginner mind” all the time. Interesting paradox between that and the idea of mindfulness. One way to resolve the paradox is to think of it as ‘flow’ on the learning curve equals non-flow in each episode. If you remove the peaks and it becomes one continuous series of troughs and a smooth curve, perhaps the learning curve flow state and episodic flow state become one. Makes sense since the troughs would have to be closer together so that all practice becomes one continuous episode. Kinda like those martial arts movies where the master teaches the student to treat every moment of everyday life as practice.

      Greg, since I know you do the cross-fit thing, I can see how this fits in with the ‘muscle confusion’ idea: you never let your body adapt to one performance domain. By keeping it confused, you stay a beginner, every workout is torture and a trough, but you keep progressing much faster overall.

      Good discussion, Greg and MFH. Helped me figure out some nuances that I can work up into a more comprehensive model.

      • I agree with much in the previous comments, particularly MFH’s comment that putting too much stock in, what may be only a cheap trick, potentially leaves you as an expert in dead technology.

        Here is one more on the relationship between refinement, peaks, and troughs. The definition of refinement I have been going by is:

        “Effort invested on behalf of others, in order to make one’s work more readily consumable or more easily intelligible. ”

        If you dedicate no effort whatsoever to refinement, then you are like the person who swears he is genius but is completely inept at explaining his genius thoughts and therefore is completely ignored by everyone else. Of course, by definition we can’t know whether there is any validity to such a person’s claims.

        However, the more important point in this context is that this isolated genius loses the benefit of any feedback or collaboration. He may progress steadily through a smooth series of troughs, but that progress will be slowed significantly by his complete isolation.

        So I wouldn’t suggest a complete absence of peaks. Even if you have already developed the foundation for the next episode, some degree of refinement is likely to lead to useful feedback that may raise the potential of the next episode.

        The question is really, how ‘peaky’ should you allow an individual peak to get? Is anything you are doing relevant to next episode? Are you even aware of what the next episode might look like?

        If you are the average Taylor-ized worker, your daily tasks – constituting the skills relevant to your local peak – are so far removed from the overall sequence that you can’t possibly see the next episode. These are the people who end up unemployed for 12+ months. They are so invested in a local peak that they lose their entire identity when forced to re-engage with the long term episodic progression.

        • The nature of refinement is fractal-like. Something is refined when you can “zoom” in any dimension and find some kind of vaguely defined “smoothness”. Truly refined works are smooth even in dimensions that the author did not anticipate.

          I agree with you that an important property of refined-ness is “effort invested on the behalf of others”. However, in general I would prefer the contrapositive: “if there is insufficient work invested on the behalf of others, a work should not be considered refined”.

          This also provokes an interesting aside: who exactly is “future me”? Is he not also an “other”? Has he forgotten important context? Is he hung over, etc? Did his predilection for West coast double IPAs finally melt his brain? For those that don’t particularly care about coddling a vague and poorly-defined “other”, this is a powerful pro-refinement argument.

          Finally, I really enjoy your crystal-clear elucidation asking “how ‘peaky’ should you allow an individual peak to get?” Any trite answer is most likely garbage. In fact, what we’re looking for is a refined (fractal-like) answer.

          Borrowing loosely from Spiral Dynamic psychology, I belive the answer breaks down into two main blocs: (a) first tier, not self-aware, and (b) second tier, self-aware to a varying degree.

          A “correct” answer to a first-tier individual early in their career is likely to be a borderline pathological series of non-optimal results, with the hope that the individual becomes aware of a deeper truth and begins to actively question the path laid for them. If not, then it has likely caused no significant harm.

          A “correct” answer to a first-tier individual later in their career differs significantly. Such an individual is likely to never become second-tier. A “correct” answer is likely to be literally whatever causes them to become obsolete the least quickly.

          A “correct” answer to a second-tier individual is Zen. It’s NP-complete. Spirit. Quantum action at a distance. It’s literally tensions in every dimension balancing each other. The “correct” answer here is “follow your heart, but don’t prance through minefields”.

      • Re: Venkat –

        I was just skimming through here again and realized that your comments on ‘muscle confusion’ related to the comment I left yesterday on the newest post (http://www.tempobook.com/2011/10/25/thrust-drag-and-the-10x-effect/#comment-1392).

        The idea with crossfit is not just “muscle confusion” but also that the more you focus on one very narrow skill the more you are limited by everything you are ignoring. The more you focus exclusively on building strength, the more your (lack of) metabolic conditioning becomes a limiting factor in accomplishing anything practical.

        At some point you have to address all those things that you suck at. You enter a trough wherein you are focused on things you dislike and are not particularly good at, but your improvement in those areas forms the foundation for your next peak.

        That is essentially the same thing I am saying about spill-over effects in my comment on the other post. The meta-cognition you describe has to include the ability recognize the interplay between thrust and unrelated/non-obvious sources of drag.

  3. I don’t think you need to go all extrinsic/intrinsic about the way that metric s-curves overlap; I think there’s more subtlety in distinguishing between metrics of performance vs the performance process as being about “fitting”. In other words, as an information process involving responding to the bowling conditions, the first pre-requisite is sufficient speed. Because in this case you are fitting to the weaknesses of a human bowler, this will reduce the time allowed to process the trajectory, and require better higher levels of power in order to respond. In other words the metrics have value via their ability to tilt and shift the game theoretic space between the two players, the options and cues they have access to.

    There is only such a thing as a good bowler in an intrinsic sense to the extent that the prerequisites for advantages can be combined together. If they can’t be constructed into a single “peak” to aim for, or rather a single direction (because most intrinsic improvement methods are about heuristics of change), then the model will break down, you run the risk of a “good bowler” diverging from one who can bowl successfully, as in wushu. Of course, often many sports have a basic level of craft, before schools diverge from it, with their own specialised metrics and improvement heuristics, and after that certain forms of tactics, with their own supporting metrics.

    To be ruthless, there is no existing world where survival can be guaranteed in a purely intrinsic sense. Without a niche, there isn’t any darwin.

    Fit isn’t a perfect metaphor, as often you are attempting to creating a niche from a combination of previously unrecognised potential and your imagination, rather than fitting to a conventionally recognised opportunity. But I think you need something like it to make functional sense of these situations.

    This then opens up a further layer of intrinsic motivation within that, the relationship between coasting vs finding sufficiently sophisticated opponents or problems to make the moment of play/problem solving/info processing sufficiently difficult, and to retain the incitement to improvement motivated by this noisy feedback mechanism. Unconscious competence as experienced is pretty close to daydreaming, you always want a smidge of conscious incompetence in there too.

    • Relevant typo – weaknesses of a human batsman.

      I also notice you’re using a different definition of grit here compared to in your ribbonfarm post. There could be a paradox here that the times you have to exhibit the most grit you look a lot like everyone else, whereas the times you are closer to a flow state, your amazing grit is obvious to everyone but you!

      • Yeah, I wrote this before I wrote the grit post, and some ideas changed in my head. This stuff is coming together in a messy way.

        Good points in previous comment. Will need to let them simmer. Don’t have an immediate response.