Slack in Cells, Slack in Brains

• AI News Team
• March 31, 2026

​[A veridically metaphorical explanation of why you shouldn’t naïvely cram your life with local optimizations (even for noble or altruistic reasons).]TL;DR: You need Slack to be an effective agent. Slack is fragile, and it is tempting to myopically sacrifice it, and myopic sacrifice makes future myopic sacrificing more likely. Learn not to do this and cultivate slack.Slack in CellsThe smallest living mammal is the Etruscan shrew, weighing about 1.8g (“as much as a paperclip”), and ~4cm in length. When curled up, it fits on a post stamp. The largest living mammal is the blue whale, weighing ~100 tons, and about 24 meters on average. Its aorta is so large that a human newborn could fit into it.[1]Taking those two species as the lower and upper bounds of the mammalian range, we see that they are separated by  orders of magnitude in length and orders of magnitude in mass.Interestingly, this is very close to the 9 orders of magnitude that span the size of bacterial cells, as measured by volume.Here are two plots from Evolutionary tradeoffs in cellular composition across diverse bacteria by Kempes et al.[Description from the article:] (a) The volume-dependent scaling of each of the major cellular components for bacteria. (b) The total cell volume compared with the volume of all cellular components as a function of cell size.The plot on the left shows us how the volume of various cellular components—DNA, protein, ribosomes, membrane, and RNAs—scales with the total cell volume. The plot on the right shows us how the aggregate volume of all the components scales with the total cell volume. Both are modeled as power laws, inferred from available data.Two things are evident. First, the volume of all RNAs and ribosomes grows faster than the cell volume. Bigger cells are more hungry-per-cell-volume for RNA and ribosomes than smaller cells. The model predicts that a bacterial cell of about  volume would be completely filled with stuff, with zero free cytoplasmic space. This is because bigger cells have greater relative protein turnover, so they need to produce more proteins, more quickly, hence the need for more protein-producing machinery: ribosomes and RNAs.On the other hand, DNA and membrane volume grow much more slowly. Looks like bigger cells don’t really need much thicker membranes than smaller cells, and the amount of DNA needed barely changes. The two lines also intersect the line of the total cell volume on the left end, around . So the smallest possible cell “should”—again, according to the model—be completely filled with mostly DNA and membrane, with no free cytoplasmic volume.Second, the smallest observed cell sits slightly left to the first intersection of the two lines on the right plot. Does this bacterium somehow fit more into its cell than the volume of the cell allows?No. The smallest cells “cheat” the “laws” by cutting down on the most volume-occupying components. They cut down the thickness of the membrane (no cell wall) and the size of the genome. They also tend to take much more spherical shapes to minimize the relative volume of the membrane.Constraint-stretching tricks are also employed on the upper range of bacterial size. The biggest bacteria known today belong to the genus Thiomargarita and reach the volumes up to about , 3 orders of magnitude more than the limit of  predicted by the model. The simplest of the tricks is that large parts of the cell volume (generally more than half, and more than 90% in Thiomargarita namibiensis, the second-biggest known bacterium) are taken by vacuoles that don’t require much maintenance, and therefore allow for cutting down on RNA and ribosomes. So, there are certain latent constraints—specifically, regularities of relative scaling of cellular components—governing the “permitted” sizes of bacterial cells.[2] Those constraints can be stretched, by modifying the standard bacterial “body plan” (including the structure of the cell envelope, the rough size of the genome, the general cellular composition, etc.). However, there’s a reason why this bacterial body plan is the generally most common bacterial body plan.One thing that you sacrifice as you go towards the extremes of the bacterial body size is that you’re losing free cell volume. The maximum free cell volume fraction (equivalently, minimum dry volume fraction) occurs around the total cell volume of . Here’s one more plot from Kempes et al. (It’s interesting that it rises much more steeply past to the right of this point (for bigger cells), than to the left of this point (smaller cells).)[Description from the article:] The fraction of total cell volume that is occupied by the essential components.Kempes et al. write that the cell volume that maximizes the expected free cell volume is where we find “many well-studied species such as E. coli”. While a more systematic investigation would be necessary to establish this robustly, I take this as an indication that there’s a strong and common selection pressure for a lot of free cell volume. Why?The lack of physical space constraints may give those cells more flexibility. First, it allows for greater adaptivity: those cells can allow themselves to dynamically increase the number of various cellular components, depending on the environmental conditions (e.g., increase the number of ribosomes to grow more quickly when food is abundant).Second, it allows for greater robustness: the cells can accommodate toxic waste products without significant harm to the cell and excrete them slowly, rather than as quickly as viable in order to avoid increasing the concentration of those in the cell (lower free cell volume⇒greater sensitivity of concentration of substance X to the same change in the number of molecules of substance X). It seems very natural to apply to this functional free cellular volume the common in the LessWrong space term “slack”:Slack is absence of binding constraints on behavior.While we can see selection pressures occasionally pushing bacterial lineages to the extremes of the viable size, it seems that most of them stay within the region allowing some slack. Speculating, a conjecture generalizing this observation would be that slack is a naturally convergent goal for robust reproducers in a wide range of environments.Slack in Brains[OK, this is way less neuroscience-y than “Brains” might suggest (actually, it mostly isn’t neuroscience-y at all), but I decided to go with it because it’s true enough (it’s about ~minds/agents) and because it gives the title a rhythmical, rhyming structure.]It seems rather obvious that you shouldn’t just plan your entire schedule in the greatest amount of detail available to a human.First, you need to be adaptive: you don’t know the future contexts that you may face, so you need to allow yourself to determine what to do on the spot. This is the central idea behind P₂B: Plan to P₂B Better: since you don’t know everything that would allow you to plan everything in advance, you need to instead plan to make a better plan, once more information is available.[3]Second, you need to be robust: some random stuff is likely to happen, and you will need to react appropriately. For an important call, you join your important call early to check that your mic and camera work appropriately. You leave early, in case traffic slows you down, or there is some issue at the airport that makes things move much more slowly.We can think of slack as a space that an agent gives to their future self to handle hard-to-predict things that life might throw at them: filling in the gaps in one’s plans (adaptivity) and adjusting for various perturbations (robustness).[4]Slack is FragileI’ve witnessed both people around me and myself gradually have their Slack eaten. Each step is small. It may seem big in the scale of the agent-episode that you are, but inconsequential in the grand scheme of things. The frog is being boiled slowly, and the elbow room you have available to manage your projects gradually deteriorates closer and closer to zero.Each time you allow this unreflective process to eat a bit of your Slack, the process gains Steam. It acquires strength. You, instead, acquire inertia: the more things you have going on, the harder it generally is to find the time to think about how to delegate any single one of them (especially if you haven’t had the Slack to write a documentation that would make graceful delegation easy). Also, it is a human default to just keep doing what they’ve been doing—including what heuristics they’ve been applying to decide how to change what they’re doing—and humans defere more to their default settings when they don’t have the Slack to reflect. Caring about your future selves and the fate of your endeavors demands that you don’t let yourself get eaten, as does caring about people who might mimic your behavior and their endeavors.[5]Hofstadter’s Law says that “it always takes longer than you expect, even when you take into account Hofstadter’s law”. One could view it as a justification of the (non-literally true, but directionally correct) maxim “plans are useless, but planning is indispensable”. Time is one sort of “space” that one can afford oneself to use in order to accomplish some endeavor. Slack is another sort of “space”. They actually seem closely connected. If you have more time, but the amount of things you have to do is kept constant, then you have more Slack. The more Slack you have, the more of this Slack you can use to pursue some goals, so you effectively spend more time on pursuing those goals. All of this is to say that, having already accepted Hofstadter’s Law as a valid heuristic/regularity, we should not be too surprised that we systematically neglect Slack.It seems like the naive solution is to train oneself to have a better assessment of how much Slack one needs. Until then, make it your default that you have a bit more Slack than you can reasonably expect to need.[Obligatory disclaimer that the Law of Equal and Opposite Advice applies, as always. Please don’t use it to rationalize succumbing to your tendency to excessively deprioritize Slack.] ^Obviously, I can only think about smallest and biggest animals that we know of. But, it seems extremely unlikely that there are bigger extant mammals than whales that we wouldn’t have seen by now. Also, as far as I remember from reading Geoffrey West’s Scale, the Etruscan shrew hits some limits of what can be achieved with the mammalian metabolism, especially including the circulatory system. (Admittedly, mole-rats stretch the metabolism part quite a bit.) ^And organisms in general, but here we’re talking bacteria.^In particular, what you’ll need to do will often depend on what you’ll have done, but the more complex the domain you’re acting in, the more difficult it is to predict what you’ll have done.^I’m not claiming that this is all that slack is and definitely not that this is the best way to conceptualize all that slack is. See, for example, Slack gives you space to notice/reflect on subtle things. ^Association: https://www.lesswrong.com/posts/RrL7xqdPycGNHQkXR/the-lethal-reality-hypothesis Discuss ​Read More