The 60-Second Exercise to Recover Your Memory - Leonard Susskind Explains

Added: 2026-05-23

[00:00:00]Your memory is not stored in your brain.

[00:00:03]I know that sounds like the kind of thing a confused person says at a party, but I mean it literally. And by the end of this, I'm going to prove it to you.

[00:00:11]Now, before you start shifting in your seats and deciding, I've lost my mind.

[00:00:15]Stay with me because the version of that sentence that is true is far more interesting than either the version that sounds crazy or the version you already believe. And the exercise I'm going to give you 60 seconds, no equipment, no medication, no app comes directly out of understanding why that sentence is true.

[00:00:35]Not as metaphor, as physics, as neuroscience, as the actual measurable, experimentally confirmed structure of how biological information storage works in a system that evolved under physical constraints.

[00:00:50]You want to know why you forget things?

[00:00:52]You want to know how to stop? The answer is not what you think it is. And I've been waiting for a room full of people willing to hear the real version. So, let's go.

[00:01:03]Here is what most people believe about memory. They believe the brain is something like a very sophisticated recording device. You experience something. The experience gets encoded.

[00:01:14]It sits somewhere in neural tissue. And when you try to remember it, you play it back like a tape, like a hard drive. And when you forget something, the file got corrupted or the tape got erased or the hard drive developed a bad sector. The memory is gone, absent, no longer present in the system. That is the intuitive model. That is what almost every person in this room walked in here believing, whether they've articulated it that way or not. And that model is wrong in a way that matters enormously for what you do next. When you can't remember something, the brain does not record. It reconstructs.

[00:01:56]Every single time you remember something, you are not retrieving a stored file. You are running a reconstruction algorithm assembling a version of the past from fragments, cues, contextual signals, and something closer to active inference than passive playback. This is not a subtle distinction. This changes everything about how forgetting works and therefore everything about how recovery works. Let me give you the physics of this first because I think the physics is actually cleaner than the neuroscience and it sets up the right frame. In physics, we distinguish between information being destroyed and information being inaccessible. These are not the same thing. In classical mechanics, information is conserved. Absolutely.

[00:02:43]If you know the exact state of every particle in a system at one moment, you can in principle reconstruct the entire history of that system backwards and forwards through time. The information never disappears. It transforms. It disperses. It becomes extraordinarily difficult to recover, but it does not vanish. In quantum mechanics, we had a long vicious argument about whether this was still true. the black hole information paradox which uh some of you know I spent the better part of two decades fighting about with Hawking and which I will come back to because it turns out to be deeply relevant here and the answer which I am confident is correct is that information is conserved even there even in the most extreme physical process we know of information does not get destroyed it gets scrambled it gets hidden it gets encoded in correlations so complex They are practically unreachable but it does not cease to exist. Your brain is a physical system. It obeys the same rules.

[00:03:50]When you forget something, the information associated with that experience has not been deleted from physical reality. What has happened is that the pathway for reconstruction has become inaccessible, degraded or has lost the contextual scaffold it needs to run properly. That is a completely different problem and it has a completely different solution. I want to pause here because I know what some of you are thinking. You are thinking okay but the brain is not the same as a closed Hamiltonian system with perfectly conserved information. Neurons die.

[00:04:28]Synaptic connections degrade. The physical substrate changes.

[00:04:33]So isn't it possible that the information really is sometimes gone?

[00:04:36]Not just inaccessible. And the answer is yes. Sometimes it is. I am not going to stand here and tell you that uh every forgotten memory is perfectly recoverable. That would be false. And I don't say false things to make people feel better. What I am telling you is that for the overwhelming majority of what we call ordinary forgetting, the name you can't pull up, the word on the tip of your tongue, the memory of where you put your keys, the conversation you had last Tuesday that seems to have evaporated the physical substrate is almost certainly intact. The problem is access. The problem is retrieval architecture.

[00:05:15]And that problem is solvable. Here is where we need to get into some actual neuroscience. And I want to do this carefully because there is a lot of pop neuroscience out there that gets the details wrong in ways that lead people to do the wrong things. The brain stores memories in a fundamentally distributed fashion. There is no single location where a memory lives. When you encode an experience, what you are actually doing is creating a pattern of synaptic modifications spread across multiple cortical regions linked together by something called an engram. A physical trace of the experience distributed across a network of neurons.

[00:05:58]When you remember what you are doing is reactivating that pattern, you need enough of the right nodes in the network to fire together to trigger the reconstruction. And here is the crucial part. You do not need all of them. You need enough. This is why cues work. A smell, a sound, a physical sensation, a particular angle of afternoon light. Any of these can serve as a partial activation of the network enough to pull the rest of the reconstruction along with it. You have experienced this.

[00:06:33]Everyone in this room has experienced the thing where you walk back into a room and suddenly remember what you came in for. That is not magic. That is the spatial context of the room providing enough of the network activation to complete the retrieval that the different context of the hallway could not support. The room is a cue. The spatial and sensory environment you were in when the memory was encoded is part of the memories retrieval architecture.

[00:07:01]Now I have been thinking about the structure of this for a long time and I want to tell you about something that happened to me about 15 years ago that shifted how I thought about the whole problem. I was working on something related to the holographic principle.

[00:07:18]The idea that the information content of a region of space is encoded on its boundary like a uh hologram where the whole image is distributed across the surface rather than sitting at one point in the volume. And I was having a conversation with one of my graduate students, a very sharp young man about the implications of this for how information gets organized in complex systems. And he said something that stopped me. He said, "Is the brain holographic?"

[00:07:49]And I said, "Probably not in the strict sense that applies to quantum gravity."

[00:07:53]But then I kept thinking about it and I realized that in a functional sense in the sense of distributed storage where partial information can reconstruct the whole the answer is actually closer to yes than no. The memory network in the brain has something like holographic redundancy. Partial activation can drive full reconstruction. The information is not localized at a single address. It is distributed such that multiple partial inputs can converge on the same output.

[00:08:22]That is a holographic architecture, loosely speaking, and it has direct implications for how you recover what you think you've lost. Here is where it starts to get genuinely strange. And I mean that in the good way. The way where um the strangeness is pointing at something real. Most people when they try to remember something and fail do one of two things. They try harder. They strain at it. They push. They concentrate intensely on the target. Or they give up and assume it's gone. Both of these are wrong. The first is wrong because memory retrieval is not a motor task. You cannot muscle your way to a memory.

[00:09:03]straining increases arousal which can actually suppress the kind of uh relaxed associative pattern completion processing that successful retrieval requires. The second is wrong for the reasons I've already given you. The information is almost certainly still there. The right thing to do is something that most people find deeply counterintuitive which is exactly why I want to walk you through the physics and neuroscience of it before I give it to you. Because if you understand why it works, you'll actually do it. And uh if you just get a technique without the foundation, you'll try it twice, decide it's not working, and go back to straining uselessly. The 60-second exercise is this. When you cannot retrieve something, stop trying to retrieve it directly. Instead, spend 60 seconds deliberately reconstructing the context in which the memory was formed.

[00:09:58]Not the memory itself, the context, the physical environment, what room were you in, what time of day, what you were wearing, what you could hear, what the light looked like, what your body felt like, who else was there, and where they were standing, and what they were doing.

[00:10:19]You are not trying to remember the target. You are trying to reinflate the network of contextual cues that surrounded the moment of encoding. You are in physical terms trying to partially reactivate enough of the distributed engram to trigger the reconstruction cascade that delivers the memory you're after. Let me give you a concrete thought experiment. So this is not abstract. Imagine you met someone at a conference three years ago. You had a long conversation.

[00:10:51]You remember the conversation was important. You remember you like the person, but the name is gone. Completely gone. You think now most people at this point either try to dredge up the name directly. They run through the alphabet.

[00:11:04]They try rhyming words. They think it started with a J or they accept the loss. The 60-cond method says neither.

[00:11:12]Instead, where was the conference?

[00:11:15]Picture the room. Picture the particular corner of the room. What did the room smell like? The way conference rooms always have a particular smell of carpet and recycled air and coffee from 3 hours ago. Picture the time of day. Was it morning when you were sharp or afternoon when you were fading? What were you wearing? What were they wearing? What was the first thing they said to you before you knew their name? Get as many of those sensory and contextual nodes reactivated as you can. And then this part is important. Stop. Do not push toward the name. Let the network run.

[00:11:49]Let the associative activation propagate through the connections that were laid down at the time of encoding. The name, if it is recoverable at all, will often surface within seconds. This works because memory is not a record you access by finding the right address. It is a reconstruction that you trigger by providing enough of the right initial conditions for the pattern completion process to run.

[00:12:15]You are not looking for the file. You are rebooting the program. Now, I want to go deeper because there's a layer here that most people don't know about and that I find genuinely fascinating as a physicist. Uh the process I just described, contextual reinstatement, which is what this is called in the memory literature, works in part because of something called encoding specificity. The principle formalized by Endo Tolving in the 1970s is that a memory is not just an encoding of an experience. It is an encoding of an experience together with the context in which it was experienced. The context is not a label attached to the memory. The context is part of the memory. It is woven into the physical structure of the engram.

[00:13:03]And this means that the most efficient way to access a memory is to reinstate as much of the original context as possible because you are not just providing a cue, you are providing part of the memory itself. Tolving was one of those figures who was right for a long time before people fully believed him, which is a situation I find personally familiar. His encoding specificity principle was initially controversial because it conflicted with the prevailing view that memories were contextindependent representations that you encoded the thing not the thing plus context. He was right and they were wrong. And it took a decade of converging experimental evidence to settle it. I mention this not to relitigate cognitive psychology debates from the 1970s but because the story matters. Science moves by people being willing to say the consensus is wrong and being right about it and then enduring a long process of getting the field to accept it. Tolving did that.

[00:14:06]The encoding specificity principle is now foundational and it is the direct theoretical basis for the exercise I'm describing. But here is what I find even more interesting and this is where the physics really starts to bite. The encoding specificity principle implies something about the relationship between information and context that has echoes in some deep questions in physics that we do not yet have complete answers to.

[00:14:31]The question of whether information is always in principle recoverable, which I touched on earlier with the black hole problem, turns out to hinge on questions about context and correlation that are formally similar to the questions Tolving was asking about memory. In the black hole information paradox, the question is whether information that falls into a black hole is permanently destroyed when the black hole evaporates or whether it is encoded in the correlations between the outgoing Hawking radiation in a way that is in principle recoverable. Hawkings original position was that the information is destroyed. My position for which I argued for over 20 years before the field came around to it was that information is preserved but encoded in correlations so complex they are practically unreachable. The information is there. The context for reconstructing it. The detailed quantum correlations between every photon of Hawking radiation is almost impossibly difficult to assemble. But it exists. The memory so to speak has not been erased. The retrieval is just extraordinarily hard.

[00:15:48]When I think about what Toloving established about biological memory, I see the same deep structure. Information is preserved. Context determines accessibility. The failure mode is not eraser but loss of retrieval architecture and the recovery method is contextual reconstruction. It is the same principle operating at wildly different scales in wildly different physical systems. That is the kind of thing that uh when you notice it you either think is a profound unifying insight about the nature of information and physical systems or you think is an interesting coincidence.

[00:16:27]I think it is the former and I will tell you honestly that I have not worked out a rigorous mathematical version of what I mean by that which is exactly the kind of open problem that is worth losing sleep over. Let me now tell you something about uh the neuroscience of sleep because it connects directly to all of this and because most people have it exactly backwards. You have been told correctly that sleep is important for memory consolidation. What most people think this means is that sleep somehow saves memories, moves them from a temporary buffer into permanent storage like hitting control S on a document.

[00:17:11]And this is not entirely wrong, but it misses the most important thing about what consolidation actually does, which is that it reorganizes the retrieval architecture. During sleep, specifically during slowwave sleep and REM sleep, the brain replays uh recently encoded experiences. This replay is not random.

[00:17:31]It is a process of reactivating the engrams formed during the day and integrating them with previously existing memory networks. The result is not just a stronger memory trace. The result is a richer, more interconnected retrieval network where the new memory has more contextual connections, more associative pathways, more roots by which partial cues can trigger full reconstruction.

[00:17:57]Sleep does not save the file. Sleep builds more roads to the file. This is why the common advice to sleep on it when you can't remember something is actually grounded in real mechanism, not folk wisdom. A night of sleep adds connections to the retrieval network, which means that a cue that was insufficient to trigger reconstruction yesterday may be sufficient today. I have had this experience myself many times, going to bed unable to pull up a name or a reference, waking up with it, sitting in my head fully formed, having been assembled overnight while I was doing other things. The reconstruction ran successfully in the background and delivered its result on schedule. And here is the thing about the 60-second contextual reconstruction exercise that ties directly to this. When you deliberately reinflate the encoding context, when you spend 60 seconds rebuilding the sensory and situational surround of the moment you're trying to remember, you are essentially doing manually and consciously what the sleeping brain does automatically and non-conciously during consolidation replay. You are reactivating the network in which the memory is distributed. You are providing enough of the initial conditions for the pattern completion process to run. The difference is that you can do it in 60 seconds while awake because you can deliberately direct your attention to the contextual elements in a way the sleeping brain cannot. I want to say something here that uh might make some of you uncomfortable, which is that this exercise requires a skill that modern culture actively discourages, which is the ability to sit still and let your mind reconstruct something without immediately resorting to external lookup. When you can't remember something and you immediately reach for your phone, you are not exercising the retrieval architecture. You are bypassing it. And there is reasonable evidence, not conclusive, not settled, but reasonable and accumulating that this changes the architecture over time.

[00:20:05]Not in a catastrophic way, not in the way that the more alarmist commentators suggest, but in a real measurable way.

[00:20:14]The retrieval pathways that you do not use do not remain equally ready.

[00:20:19]Synaptic connections that are not activated are subject to pruning. The brain is not a static storage system. It is a dynamic system that reorganizes itself based on what it is asked to do.

[00:20:30]I want to be careful here because this is an area where the science is genuinely not settled and I do not want to overstate it. The claim that smartphones are destroying memory is too strong and too simple. The claim that there are no costs to routinely outsourcing retrieval to external systems is also too simple. The truth as usual is more complicated and more interesting than either of the easy narratives and it involves questions about uh what kind of memory capacity you want to develop and uh what you are willing to trade for the uh convenience of instant lookup. Those are not purely scientific questions, but the scientific component of them is real and it points toward the value of practicing the exercise I'm describing, not just as a trick for recovering individual forgotten items, but as a way of keeping the retrieval architecture exercised and robust.

[00:21:28]Let me take a step back and talk about something that I think gets missed in almost every popular discussion of memory and forgetting, which is the question of why we forget at all. From an engineering standpoint, forgetting looks like a bug. You are a biological computing system with an apparently vast storage capacity and yet you routinely fail to retrieve information that was clearly encoded and was clearly useful.

[00:21:54]Why would a system shaped by evolution have this apparently obvious deficiency?

[00:21:58]The answer, I think, is that forgetting is not a deficiency. It is a feature.

[00:22:03]And I mean this in a precise, not poetic sense. A system that retained everything with equal salience and equal accessibility would be computationally paralyzed. The famous literary example of this is Bourhees's F memorialius. A man who after an injury can forget nothing. He becomes unable to function, unable to generalize, unable to see similarity and pattern because his mind is overwhelmed by the infinite detail of every specific instance. Borges wrote this as fiction, but he was pointing at something real. A memory system that does not selectively forget, cannot efficiently extract patterns from experience, cannot form useful generalizations, cannot distinguish the signal from the noise of of its own stored data.

[00:22:54]Forgetting is the brain's compression algorithm. It is the process by which the system trades off perfect fidelity of specific instances for efficient extraction of general structure. And like any lossy compression, it involves real information loss about the specific details of individual experiences in exchange for better representation of the patterns that give those experiences meaning and utility. The system is not broken when it forgets your keys. The system is doing its job, which is to be useful over the long haul rather than perfect in any given instance. But here is a thing about lossy compression that is directly relevant to what we're doing today. The compressed representation is not the only thing stored. The information about the original before compression is encoded in the full system in the correlations between neurons in the contextual connections built up over time in the retrieval architecture that links the compressed representation back to the details that were nominally discarded. If you can access the right set of cues, if you can reconstruct enough of the original retrieval context, you can often unpack details from what seemed like a compressed and simplified residue. The 60-second exercise is in physical terms a targeted decompression attempt. You are trying to recover detail from what looks like a simplified residue by providing enough of the original context to run the decompression algorithm. Now I want to give you a second thought experiment that I find particularly illum illuminating because it makes the uh reconstruction nature of memory vivid in a way that I think most people have not thought through. Consider what it means to remember a conversation you had 5 years ago. Not a summary of the conversation. The actual conversation, the specific words, the specific back and forth, the specific order in which things were said. Most people, if pressed, would say they remember it reasonably well. And most people are substantially wrong about this in ways they cannot detect because the reconstruction feels like retrieval from a recording. Here is what actually happens when you remember that conversation. Your brain assembles a version of it that is consistent with your current knowledge about the people involved, your current understanding of the situation, your current narrative about what kind of person you are and what kind of person they are, and your current emotional veilance toward the memory. Details that fit this assembly are readily incorporated. Details that conflict with it are smoothed over, reinterpreted, or dropped. The memory you construct is not the conversation that happened. It is the conversation that is consistent with everything you currently know and believe with the specific moments of vivid encoding serving as anchor points around which the reconstruction is built. This is not a pathology. This is the system working as designed. The brain is not trying to be a recording device. It is trying to be useful. And a memory system that is fully integrated with current knowledge and current context is more useful for guiding future behavior than a perfectly accurate but isolated recording would be. But it means that what you experience as remembering is always partly reconstruction, always partly inference, always partly the present tense projecting itself backward into the apparent past. The implication for the 60-cond exercise is that when you reconstruct context, you are not trying to run backward to a perfect record. You are trying to reinstall enough of the original conditions that the reconstruction produces something closer to the original event rather than a version fully overwritten by everything that has happened since. Every detail you can genuinely reconstruct the room, the time of day, the physical sensations acts as a constraint on the reconstruction, pulling it toward the original and away from the current context distortions. The more constraints you provide, the closer the reconstruction comes to what actually happened. You will never get perfect fidelity, but you can get substantially better than you would with no contextual reinstatement.

[00:27:22]I want to talk now about something specific in the neuroscience that most people don't know but that has direct practical implications which is the role of the hippocampus in all of this. The hippocampus is a structure in the medial temporal lobe that is essential for the formation and retrieval of episodic memories. Memories of specific events, things that happened at specific times and places. people with hippocample damage. Most famously, the patient known as HM who had his hippocampi removed surgically in 1,953 in and attempt to treat severe epilepsy cannot form new episodic memories. They can remember how to do things. They can learn new skills, but they cannot remember that they learned them or when or where. Every day is in a terrible sense the first day. Hm is one of the most important cases in the history of neuroscience. And what he tells us about memory is profound. He tells us that episodic memory, the kind of memory we are talking about today, the kind where you remember specific events from your personal history, is not the same as procedural memory or semantic memory. It depends on a specific structure that can be selectively damaged. And it tells us something important about what the hippocampus does, which is not simply store memories, but actively participate in their retrieval. The hippocampus is the part of your brain that during the 60-second contextual reconstruction exercise is doing the heavy lifting.

[00:29:02]When you deliberately rebuild the environmental and situational context of a past event, you are activating hippocample circuits that were active during the original encoding of that event. The hippocampus acts as an index.

[00:29:16]It links together the distributed cortical representations of all the different elements of an experience so that activating any subset of them can through the hippocample index pull in the rest. When you reconstruct context, you are providing the hippocampus with enough of the index entries to locate and activate the full distributed engram. You are giving it enough to work with. Now, here is something genuinely strange and I mean strange in the sense that uh it is not fully understood and the implications have not been fully worked out. The hippocampus does not only activate during retrieval, it activates during imagination. When you imagine a future event in detail, a vacation you are planning, a conversation you are anticipating, the hippocample activation pattern is similar to the activation during episodic memory retrieval. The same system that reconstructs the past is building simulations of the future. This is not a coincidence. This is the same underlying mechanism. Memory and imagination are not opposites.

[00:30:27]They are the same process running in different directions. Memory is reconstruction of a constrained simulation constrained by the actual encoding from the original event.

[00:30:39]Imagination is reconstruction of an unconstrained simulation. The hardware is identical. This has an immediate implication for the 60-second exercise that I find striking. When you reconstruct context and you deliberately engage the vivid sensory and situational details, you are doing something more like imagination than most people realize. You are running the reconstruction engine not just as a lookup but as a simulation.

[00:31:08]And the simulation, if you provide enough of the right initial conditions, will tend to converge on the actual past event rather than a freely invented one because the actual encoding of the original event provides constraints, biases, and the network weights that pull the simulation toward fidelity. The line between remembering and imagining is thinner than most people assume, and the mechanism of recovery sits right on that line. There is something here that genuinely unsettles me and I want to be honest about that because I think the unsettling questions are more valuable than the comfortable answers. The fact that memory and imagination use the same underlying machinery means that the line between what you remember and what you have constructed is not always clear even to you. Even with full access to your own experience, the reconstruction feels like retrieval. The simulation feels like fact. You cannot always tell from the inside whether the memory you just recovered is accurate or is a plausible confabulation assembled from available fragments in a way that feels complete but is not. This is not a bug in the system you are using. It is a feature of the system itself and it means that the recovered memory however genuine the recovery process should be held with appropriate epistemic humility. You got closer to what happened. You did not necessarily arrive at an exact replica of what happened. I find this genuinely productively unsettling. The tool we are using to navigate our personal histories is not a tape recorder. It is a physicist building a model of the past from incomplete data and running that model forward, confident in its output, but not infallible. Sometimes the physicist gets it right. Sometimes the physicist fills in a gap with a plausible interpolation and doesn't know it. The 60-cond exercise gives the physicist better input data. It does not make the physicist omnisient. And uh now I want to bring this home because I think there is a larger point here that has to do with how you move through your life, not just how you recover a particular forgotten fact. The mistake most people make about their own memory is the same mistake that led to a century of wrong thinking about black hole information.

[00:33:32]They assume that what is inaccessible is gone. They treat the failure to retrieve as evidence of non-existence.

[00:33:39]They see the empty screen where the memory should be and they conclude the file has been deleted. But the physics of information does not work that way.

[00:33:47]Inaccessibility is not non-existence.

[00:33:50]Difficult retrieval is not evidence of eraser. The correct response to a blank screen is not to accept the loss. The correct response is to ask, "What retrieval architecture am I missing? And can I reconstruct it?" The 60-second exercise is the answer to that question put into practice. 60 seconds of deliberate contextual reconstruction done with the specificity and the patience that the reconstruction process actually requires. Not straining toward the memory, not giving up on it.

[00:34:22]rebuilding carefully and concretely the conditions under which it was formed so that you can give the pattern completion system in your hippocampus enough to work with.

[00:34:33]It takes 60 seconds because that is roughly the amount of time required to activate enough contextual nodes to drive a successful retrieval cascade if the memory is retrievable at all. Some memories will not come back even with perfect contextual reconstruction. Some have genuinely faded below the threshold of activation. And I am not going to pretend otherwise. But most of what you think you have forgotten, you have not forgotten. You have simply lost temporary access to it. And access can be restored. Here is the final thing I want to leave you with, and it is not comfortable, but I think it is true. The way you treat your own memory, whether you exercise it or outsource it, whether you practice retrieval or bypass it, whether you hold your memories as reconstructions to be examined or as recordings to be trusted, this matters more than most people appreciate.

[00:35:29]Not because memory is some mystical faculty that needs protection from the modern world, but because the reconstruction engine is plastic. It is shaped by how you use it. A retrieval architecture that is regularly exercised, that is regularly asked to do the work of contextual reconstruction rather than being bypassed by a search engine, is a better retrieval architecture. Not dramatically better in the short term, substantially better over years and decades. The physicists I admire most are the ones who built their intuition brick by brick through years of turning problems over in their minds, constructing and demolishing and reconstructing their understanding, not relying on a reference book to look up the answer, but actually building the answer from first principles each time until the first principles were internalized. That is not just a romantic notion about how science should be done. It is a description of how Understanding is actually built in a biological neural network. You do not build intuition by looking things up.

[00:36:37]You build it by constructing and reconstructing over and over until the reconstruction is so fast and so reliable it feels like knowing. That is what memory is for. Not storage, not recordkeeping. construction of a model of the world and of your own history that is rich enough, connected enough, and dynamically retrievable enough to be genuinely useful for navigating what comes next. The 60-second exercise is not just a trick for recovering uh a forgotten name. It is practice in using the system as it was designed to be used. And the more you practice it, the better it works and the less often you will need it.

[00:37:21]What I cannot tell you, and I want to be honest about this, because I think false certainty is more dangerous than admitted ignorance, is exactly where the boundary lies between recoverable and unreoverable, between the memory that is inaccessible and the memory that is genuinely gone.

[00:37:39]We do not have a clean answer to that.

[00:37:42]The neuroscience does not yet give us a test you can run to determine in advance whether a given retrieval attempt will succeed. We know the variables that affect it. Quality of encoding, number of retrieval pathways, interference from similar memories, elapse time, sleep quality, stress levels, but we cannot sum them into a single predictive equation. That problem is open. It is one of the problems that I think is genuinely hard in the way that the best problems are hard. Not because we lack the data, but because we may lack some conceptual framework that would let us put the data together correctly. And maybe that is the right place to end.

[00:38:24]Not with a neat answer, not with a technique delivered cleanly and wrapped up, but uh with the observation that uh your memory, the thing that makes you continuous across time that gives you a personal history that connects the person you were to the person you are, is not a vault you have access to or don't. It is a living physical system that you participate in, that you shape by how you engage with it, and that uh you can learn to use better. And that somewhere in the space between inaccessible and gone, between the blank screen and the recovered name, between the perfect recording we think we have and the probabilistic reconstruction we actually have, there are answers we still do not have. And those answers are worth going after. That is what physics taught me that uh the blank screen is not the end of the problem.