Summary
Marina Wyss (senior applied scientist, Amazon) walks through a cognitive-science-backed retention system built around five principles: active recall over rereading, pretesting before learning, spaced repetition at expanding intervals, interleaving over block studying, and the walk-and-talk method for multi-channel encoding. The throughline: most study methods feel productive but produce the fluency illusion — mistaking recognition for retrieval. The system is uncomfortable by design, and that difficulty is the mechanism.
Notes
- The fluency illusion — rereading makes material feel familiar; the brain mistakes that familiarity for knowledge. Recognition and retrieval are separate cognitive processes. The one that matters in practice (interviews, on the job) is retrieval. → Fluency Illusion
- Active recall — after studying, close everything and pull from memory. Students who test themselves retain ~80% after a week vs ~34% for re-readers. Retrieval strengthens neural pathways; rereading does not. → Active Recall
- Pretesting — quiz yourself before you learn the material. You’ll get it wrong; that’s the point. A 2025 study confirmed that failed attempts prime the brain to encode the correct answer more deeply when it arrives. Bookend every session: test before AND after.
- The forgetting curve — without review, ~80% of new information is lost within 48 hours. Ebbinghaus mapped this in the 1800s. One strong session is not enough. → Ebbinghaus Forgetting Curve · Hermann Ebbinghaus
- Spaced repetition — review at expanding intervals (1 day → 3 days → 1 week → 2 weeks…). Each retrieval pushes the curve further out. Meta-analysis: combining SR with active recall improves outcomes ~25% over either alone. → Spaced Repetition
- Interleaving vs block studying — interleaved group scored 63% on delayed test vs 20% for blocked. Paradox: blocked group felt they were learning better during practice (fluency illusion again). Interleaving forces the brain to identify which concept applies — real-world thinking. → Interleaving
- Walk and talk / dual coding — walking (motor cortex active) + explaining out loud (language production) + hearing yourself (auditory feedback) = three encoding channels simultaneously. Can’t explain it = don’t understand it (Feynman technique). → Dual Coding · Richard Feynman
Transcript
how to actually remember what you study
0:00 · I’ve spent hours studying, then gone to use it 2 weeks later and drawn a complete blank. For a long time, I thought that was just how my brain worked. Like, maybe I’m just not the kind of person who retains things well. 0:11 · Turns out, it wasn’t me.
0:12 · [music] It was the method I was using to study. 0:14 · I’m a senior applied scientist at Amazon. I came from a non-tech background and taught myself most of the skills that got me where I am today. And I’ve mentored almost 200 people into AI and machine learning roles. So, I’ve [music] watched this play out hundreds of times. They’re putting in the hours, doing the courses, and when it actually matters, in an interview, on the job, or when someone asks them to explain a concept they studied last month, it’s gone. The problem isn’t effort. It’s that most of us were never taught how learning actually works. So, I’m going to show you the system I use to retain what [music] I study. It’s backed by cognitive science, and most of it goes against what feels productive, which is exactly why it works. The first thing I changed was how I reviewed material. I used to reread. That was my whole strategy. Go through the notes again, go through the documentation again, maybe pilot some stuff. And it felt productive. After a couple of passes, the material starts to feel familiar.
the illusion of learning
1:03 · But, here’s the [music] thing. There’s a well-documented phenomenon in cognitive science called the fluency illusion. 1:09 · Your brain recognizes familiar material and mistakes that recognition for actual knowledge. You’re not learning, you’re pattern matching. You see something you’ve seen before and your brain goes, “Oh, yeah, I’ve got this.” But, recognition and retrieval are completely different cognitive processes. And the one that matters, on the job, for an exam, in an interview, when you actually try to apply something, that’s retrieval. So, here’s the fix. After you study something, close everything. Close the docs, tutorial, your notes, and try to remember what you just learned from scratch. [music] Write it down, say it out loud, whatever works. The point is that you’re pulling it from memory, not looking at it again. This is called active recall, and it’s one of the most impactful changes you can make. This isn’t just my opinion. Research shows that students who test themselves retain roughly 80% of the material after a week, compared to about 34% for those who just reread. The reason it works is that retrieving information strengthens the neural pathways associated with it. 2:04 · Rereading strengthens recognition. 2:06 · Retrieval strengthens recall. And those are not the same thing. I know it feels harder, and that’s actually the point. The difficulty is what makes it work. 2:13 · Now, here’s the part that really surprised me when I dug into the research. You don’t even have to know anything yet for this to help. There’s a technique called pretesting. Basically, quizzing yourself on material before you even learned it. Of course, you’ll get the answers wrong. That’s fine. That’s kind of the point. A 2025 study in memory and cognition confirmed what earlier research suggested. Attempting to answer a question before you know the answer primes your brain to encode the correct answer more deeply when you encounter it later. Think about what that means for how you approach a new course or a new topic. Before you watch the lecture on, say, gradient descent, try to explain to yourself what you think it is. You’ll probably be wrong, but when the actual explanation hits, your brain is already primed to absorb it because you’ve created a gap it wants to fill. So, the system isn’t as simple as just test yourself after. It’s test yourself before and after. Bookend every learning session with retrieval, even when you don’t know anything yet. But, there’s another step that’s really important to make this work. Let’s say you start doing active recall. You can actually pull the material from memory without looking at your notes. You feel good. But, 2 weeks later, it’s gone. And this is where a lot of people get frustrated, because they did the work.
the spacing problem
3:19 · They studied correctly. They tested themselves. And they still forgot most of it. Hermman Ebbinghaus mapped this out in the 1800s. It’s called the forgetting curve. 3:27 · And what it shows is very sad. Without review, you lose roughly 80% of new information within 48 hours. It doesn’t matter how well you learned it in the moment. One strong study session is not enough. This is why people finish a course and 3 months later, it’s like they never took it. One exposure, even a really good one, doesn’t create durable memory. The solution here is spaced repetition, or reviewing at expanding intervals. You review the material 1 day after learning it, then 3 days after learning it, then a week, 2 weeks, and so on. Each time you retrieve, you push the forgetting curve further out. The intervals get longer because the memory gets stronger. A meta-analysis found that combining spaced repetition with active recall improves outcomes by about 25% more than using either strategy alone. So, these first two methods aren’t just additive, they compound. The key thing is that spaced repetition doesn’t actually require more time, just different timing. Instead of one 4-hour study block, try doing four shorter sessions spread out over 2 weeks, for example. Now, spacing out these review sessions means you’re studying a lot. 4:26 · And if you’re like me, you’re not always doing it at home. I do a lot of my deep work at coffee shops, on flights, or killing time before appointments. And every one of those places has the same problem. You’re on public Wi-Fi. Here’s the thing many people don’t think about. 4:39 · When you’re connected to a cafe’s network or an airport hotspot, your traffic is exposed. Anyone on that same network can potentially see what you’re doing, the sites you’re visiting, the credentials you’re entering, and the data you’re sending. And if you’re logging into course platforms, your email, GitHub, whatever, that’s information that you don’t want intercepted. Surfshark is the VPN I use. 5:00 · It encrypts everything between your device and the internet, so even on a completely open network, your data stays private. I just connect before I open anything and forget about it. They also just released a web content blocker that strips out ads and trackers while you browse, which I love for study sessions. 5:14 · Fewer distractions, cleaner pages, and one less thing pulling my focus when I’m trying to stay in a review session. Go to surfshark.com/marinavpn, or use the code marinavpn at checkout to get four extra months. There’s a 30-day money-back guarantee, so no risk to try it. The link’s in the description.
interleaving vs. block studying
5:30 · All right, back to the system. Before I get to the third method, which is the one that makes the biggest difference for me personally, there’s one more trap I want to flag, because it’s related to everything we just talked about. Most people study topics in blocks, like a week or two on a given topic before they move on to a completely new topic. But, the research says this is another fluency illusion. When researchers compared block study against music interleaved study, where you shuffle topics instead of doing them in neat little blocks, the interleaved group scored 63% on a delayed test versus 20% for the blocked group. But, weirdly, during practice, the blocked study group felt like they were learning better. 6:07 · Your subjective sense of, “I’m getting this,” is lying to you. Again, interleaving works because it forces your brain to constantly identify which concept applies and how it differs from the last one. That’s the kind of thinking you actually need in an interview or on the job, not, “I studied neural networks for 3 hours straight,” but, music “Given this problem, which approach is the right one?” I’m not saying you need to overhaul your whole study plan around this. But, when you’re doing those spaced review sessions, mix the topics up. And while you’re at it, there’s one more tip I have to make everything stick. This is the one that feels the strangest, but it’s the one that consistently produces the best retention for me. I walk around and talk to myself out loud about the topic.
the walk and talk method
6:45 · Like, actually out loud, like I’m explaining it to someone who isn’t there. I know that sounds ridiculous, but hear me out, because this is the part of my system that I think most people have never tried, and it changes everything. Most studying is passive and single-channel. You’re sitting, reading, maybe typing. There’s only one input mode. But, the brain encodes information more deeply when multiple systems are engaged simultaneously. This is called dual coding, and it’s backed by a solid body of research on embodied cognition. 7:12 · Here’s why the walk and talk works. When you’re walking, your motor cortex is active. Movement creates a different physiological state than sitting, and that state affects how information is encoded. When you’re speaking, you’re forcing yourself to organize information into language. And this is where it gets fun and slightly painful, because you can’t explain something out loud without actually understanding it. When you try, the gaps become obvious immediately. 7:34 · You’ll be going along fine and then suddenly stumble on a part you thought you understood, but can’t actually articulate. That’s the Feyman technique in action. If you can’t explain it simply, you don’t understand it yet. And when you hear yourself, you get an auditory feedback loop, too. That’s another input channel for the same information. So, you’ve got movement, speech production, and auditory processing all happening at the same time, and all encoding the same material. Three channels instead of one. 7:59 · I do this most mornings. Some concepts I’ve talked through 10 or more times. 8:02 · Those are the ones I don’t hesitate on in a technical discussion. They’re not stored as something I read once. They’re stored as something I’ve explained, heard myself explain, and physically, like, moved through. That’s a completely different kind of memory. Now, if you’re thinking, “This all makes sense, but I know myself. I’ll do it for a week and then fall off.” That’s exactly why I built the AI and ML Career Launchpad. 8:23 · It’s a learning community for aspiring AI and ML professionals, where you’ve got structure, accountability, and other people going through the same thing. 8:31 · Because the system only works if you actually use it consistently. I’ll drop the link in the description. So, here’s what this comes down to. Everything most people do to study, rereading, highlighting, cramming, it feels productive, and that feeling is the trap. The system that works is uncomfortable by design. That discomfort is the learning. But, even if your approach is perfect, it doesn’t matter if you don’t do it consistently. I made a video on how I get myself to study every day without using much willpower at all. Check that out next.