You finish a good nonfiction book and feel like you’ve learned a lot. Two weeks later you can name the title, one idea, and the color of the cover. If you want to know how to remember what you read, the short answer is this: stop rereading and highlighting, close the book and write down what you remember, repeat that a few days later, and turn each chapter into one thing you’ll do. The research behind each of those steps is unusually solid.

I co-founded Podex because of the same problem with podcasts: people keep far less of what they hear in a long episode than they think. Our rule when we turn an episode into a summary is that every tactic has to come with one step you can run this week. Books have the same problem and, it turns out, the same fix. Here’s the evidence, then a routine you can use on the book you’re reading now.

Why you forget most of what you read

Forgetting isn’t a sign that you read badly. It’s the default. In the 1880s Hermann Ebbinghaus memorized lists of nonsense syllables and measured how fast he lost them: steeply in the first hours and days, then more slowly. In 2015 two researchers in Amsterdam, Jaap Murre and Joeri Dros, repeated his experiment with one volunteer who spent about 70 hours learning and relearning lists. The curve they got was very close to the original.

Books aren’t nonsense syllables, and you remember meaningful ideas better. But the shape holds. Unless something brings an idea back, it fades fast, and the first few days matter most.

The good news is that the fix is not more reading. It’s a different kind of effort after you read.

Why rereading and highlighting feel useful but don’t help much

The two most common ways people try to remember a book are going back over it and marking it up. Both feel productive. Neither does much.

In 2009, Aimee Callender and Mark McDaniel at Washington University ran a set of experiments with textbook chapters and a Scientific American article. Reading a text a second time, straight after the first, gave no reliable benefit on the kind of tests students actually take. That held across prior knowledge, reading ability and test delay.

A few years later, a team led by John Dunlosky reviewed the evidence on ten popular study techniques for the journal Psychological Science in the Public Interest. Highlighting and rereading landed in the “low utility” group. Two techniques earned the top rating: practice testing and spreading your study over time.

Part of the reason rereading misleads you is that a familiar page feels easy. Easy is easy to confuse with “I know this.” You only find out what you know when the book is closed.

Close the book and recall: the strongest single step

The most cited study on this is from 2006. Henry Roediger and Jeffrey Karpicke gave students short prose passages to learn. One group read the passage in four study periods. Another read it once and then spent three periods writing down everything they could remember, with no feedback.

On a test five minutes later, the rereaders did better. A week later the picture flipped. The recall group remembered 61% of the passage. The rereading group remembered 40%. The rereaders had gone through the text about 14 times; the recall group about 3.

In a classic memory study, students who closed the text and wrote down what they remembered recalled 61% a week later. Students who kept rereading recalled 40%, despite reading it four times as often.

A common objection is that recall only helps with facts, not understanding. A 2011 study in Science by Karpicke and Janell Blunt tested that. Students who practiced recall did better a week later than students who built detailed concept maps with the text open, including on questions that asked them to draw inferences the text didn’t state directly.

What this means for a reader: the moment that does the work is the one where you look away and try to say what the chapter said. It feels harder than rereading. That difficulty is the point.

Space it out: when to come back

Recall once and you’ll keep more. Recall a few times, with gaps, and you’ll keep it much longer.

How long should the gaps be? A 2008 study led by Nicholas Cepeda had 1,354 people learn obscure facts in two sessions, with gaps from 20 minutes to 105 days and final tests up to 350 days later. The best gap depended on how long you want to remember. To remember something for a week, the best gap between sessions was one to a few days. To remember something nearly a year later, it was about three to four weeks.

To remember facts for almost a year, the best gap between study sessions was about three to four weeks. For one week, it was one to a few days.

You don’t need flashcard software to use this, though it helps if you already like it. A note in your calendar saying “recall chapter 4 of the pricing book” works.

Turn each chapter into a decision

For most nonfiction readers, the goal isn’t to pass a test. It’s to change something you do. That needs one more step, and it’s the one almost every “remember what you read” guide skips.

Write each idea you want to keep as an if-then plan: “When I’m writing a pricing page, I will show the annual price first.” Psychologists call these implementation intentions. A 2006 meta-analysis by Peter Gollwitzer and Paschal Sheeran, covering 94 tests and more than 8,400 people, found a medium-to-large effect (d = 0.65) on reaching goals for people who made them, compared with people who only set the goal.

An idea you can’t turn into “When X happens, I will do Y” is an idea you probably won’t use.

This is also a good filter. If a chapter doesn’t give you a single if-then, you may not need to remember it at all.

The close-the-book routine

Here’s the whole thing as a routine. It adds roughly 15 minutes per chapter, spread over a month, most of it in two-minute pieces.

1. Before a chapter, read its headings and turn the main one into a question. “Why do anchors work in pricing?”
2. At the end of the chapter, close the book. Write three bullet points from memory. Don’t peek.
3. Open the book and check. Fix what you got wrong in a different color, or mark it. The misses are what you study next time.
4. Write one if-then plan from the chapter, or decide there isn’t one.
5. One or two days later, recall the three bullets and the plan again without looking. Two minutes.
6. About a week later, do it again. Then once more around a month later, for the ideas you still care about.
7. When you finish the book, write a short summary from memory: what it argued, and the two or three plans you’re actually using. James Clear describes doing a three-sentence version of this for every book he finishes.

A worked example. Say you’re reading a chapter on pricing and the main idea is that people judge a price against whatever number they saw first. Your three bullets might be: the first number sets the reference; showing an expensive option makes the middle one look reasonable; this works even when people know about it. You check and find the book actually said the third point was weaker than you remembered. Your if-then: “When I put prices on a page, I’ll list the most complete plan first.” On Thursday you recall it in the queue for coffee. A week later you check whether you actually did it.

The same method works for podcasts and audiobooks

None of this depends on the page. If you listen to a lot of audiobooks or long interviews, the problem is the same and so is the fix: pause at the end of a segment, say or jot down what you remember, and pick one thing to do. I wrote a podcast-specific version in How to Retain Information From Podcasts (and Actually Use It).

If you’ve wondered whether listening counts as reading in the first place, the short answer from the research is that comprehension is close, with a few traps around attention. I went through 46 studies on that in Is Listening as Good as Reading?

Where this advice is weaker

Most of these studies used students, short passages and tests a week or so later. A 300-page business book read over a month is a different thing, and nobody has run the trial for it that I could find. The direction of the evidence is very consistent, though, across materials and ages, which is why Dunlosky’s review rated recall and spacing so highly.

The Ebbinghaus replication used one person and meaningless syllables. It tells you the general shape of forgetting, not how fast you’ll lose a good argument.

Recall is also less fun than rereading. If the routine makes you stop reading, it has failed. Use it on the books where you want to change something, and read the rest for pleasure.

Questions people ask about remembering books

Is highlighting useless? Not useless, but weak on its own. Use it to mark what you’ll test yourself on later, then close the book.

Should I take notes while I read? Notes written from memory at the end of a section do more than notes copied while the page is open. The UNC Learning Center gives the same advice: read a section, pause, then write from memory.

How many times should I review? The studies above suggest two or three spaced recalls for things you want to keep for months: a day or two after, about a week after, and about a month after.

Why do I forget books so fast even when I enjoyed them? Enjoying a book and retrieving its ideas are different activities. Enjoyment happens while reading. Memory gets built when you try to bring the idea back later.

One thing to try today

Pick the chapter you read most recently. Without opening the book, write three things it said and one if-then plan. Then check. Put a reminder in your calendar for two days from now that just says the chapter name.

Disclosure: I’m a co-founder of Podex. We turn long podcast episodes into short playbooks where every idea comes with a step to try this week (200+ summaries so far), and you can paste a link to an episode you don’t have time for. If that would help, you can try Podex here.

Sources: Roediger & Karpicke (2006), Psychological Science, doi:10.1111/j.1467–9280.2006.01693.x. Karpicke & Blunt (2011), Science, doi:10.1126/science.1199327. Callender & McDaniel (2009), Contemporary Educational Psychology, doi:10.1016/j.cedpsych.2008.07.001. Dunlosky et al. (2013), Psychological Science in the Public Interest, doi:10.1177/1529100612453266. Cepeda et al. (2008), Psychological Science, doi:10.1111/j.1467–9280.2008.02209.x. Murre & Dros (2015), PLoS ONE, doi:10.1371/journal.pone.0120644. Gollwitzer & Sheeran (2006), doi:10.1016/S0065–2601(06)38002–1. UNC Learning Center, “Taking Notes While Reading”.

Does AI make you dumber? The honest answer from the research so far is no, not in general. But one specific way of using it does make you worse at the task once the AI is gone: asking it for the finished answer before you’ve done any thinking of your own.

Several randomized experiments now point the same way, and the gap between people who use AI well and badly is often bigger than the gap between AI users and non-users.

I’ve long suspected that people keep much less of what they hear in long podcasts than they think they do. AI summaries promise a shortcut around that problem. So I wanted to know what the evidence says about learning from AI, and where exactly the cost comes from.

The study everyone quotes

The one that made the headlines came from the MIT Media Lab in 2025. Nataliya Kosmyna and colleagues had 54 people write essays with ChatGPT, with a search engine, or with no tools, while EEG recorded their brain activity.

The ChatGPT group showed the weakest neural connectivity. And when asked to quote a sentence from an essay they had written minutes earlier, 15 of the 18 ChatGPT users couldn’t do it accurately. In each of the other groups, only 2 of 18 had that trouble.

That last detail is striking, but the study is small, it’s a preprint, and it tested one task: writing an essay you didn’t really have to write. It’s a warning sign, not a verdict. The more convincing evidence came later, from larger trials.

What the bigger experiments found

In a field experiment in a Turkish high school, Hamsa Bastani and colleagues gave nearly 1,000 students access to GPT-4 during math practice. One group got a plain ChatGPT-style interface. Another got a version prompted to act as a tutor and give hints rather than answers.

During practice, the plain version improved scores by 48% and the tutor by 127%. Then came an exam without AI. Students who had practised with the plain version scored 17% lower than students who had never had AI. The tutor group showed no such drop.

A 2026 paper presented at the Conference on Language Modeling took the same question into a lab, across 1,222 people. Grace Liu, Brian Christian, and colleagues gave participants fraction problems or SAT reading questions, with GPT-5 in a sidebar for some of them, then took the AI away without warning.

In the first experiment, people who’d had AI solved 57% of the final problems, compared with 74% for people who never had it, and they skipped problems almost twice as often. The AI session lasted only 10 to 15 minutes.

The detail that matters most is in the second experiment. When the researchers split AI users by how they had used it, the drop was concentrated among the 61% who used it mainly to get answers. People who used it for hints or clarifications did about as well as people with no AI at all.

Coding shows the same split. In a randomized study by Anthropic researchers Judy Hanwen Shen and Alex Tamkin, 52 developers learned a Python library they hadn’t used before, with or without an AI assistant.

The AI group scored 17% lower on a quiz about the library and wasn’t significantly faster. But the AI users weren’t all alike. Those who asked conceptual questions, or asked for explanations alongside generated code, scored far higher than those who handed the task over.

What about summaries?

Most of those studies are about solving problems. If you use AI the way I’m interested in, to digest long content, the most relevant paper is by Shiri Melumad and Jin Ho Yun at the University of Pennsylvania.

Across seven experiments with more than 10,000 participants, people learned about a topic, such as how to grow a vegetable garden, either from an LLM or from ordinary Google results, and then wrote advice for a friend.

The LLM learners felt they’d learned less, put in less effort, and wrote advice that was shorter, less factual, and more generic. Independent readers, who didn’t know which tool had been used, found it less helpful.

This held even when both groups saw the same set of facts. And when the researchers gave people a version that showed source links next to its summary, it didn’t help, because once people had the summary, they weren’t motivated to dig into the sources. The summary felt like the end of the job.

Retention follows the same line. In a 2025 trial with 120 business students, André Barcaui had half learn a topic with ChatGPT and half with traditional methods, then gave everyone a surprise test 45 days later. The ChatGPT group scored 57.5% against 68.5%. They had also studied for less time, but the gap remained after accounting for that.

The evidence on the other side

It would be easy to stop there and conclude that AI is bad for learning. The research doesn’t support that either.

At Harvard, Gregory Kestin and colleagues ran a randomized trial with 194 physics students. An AI tutor, built to guide students through problems step by step, produced learning gains more than double those of an in-class active-learning lesson, in less time.

And a 2026 experiment in Frontiers in Psychology, with 659 participants in its main analysis, tested the order of use directly. One group wrote down their own position on an essay topic first and only then questioned the AI. Another group adopted what the AI gave them. Then everyone wrote a second essay with no AI.

The direct adopters reasoned worse than people who had never used AI. The think-first group did not fall below that baseline. The effect was small, but in the authors’ words, “What separated the arms was not access to the model but the order in which the interaction required the user to produce something.”

The pattern underneath

Put these side by side and a fairly consistent picture appears. The cost shows up when AI does the part of the work that produces learning: retrieving, struggling, connecting, and putting things in your own words. It largely disappears when the AI is set up, by its designers or by you, to make you do that part.

That fits older findings that have nothing to do with AI. In a classic 2006 study, Henry Roediger and Jeffrey Karpicke showed that trying to recall material beat rereading it on tests days later, even though rereading felt more productive. A polished AI summary is the ultimate rereading. It feels like understanding and asks nothing of you.

A Microsoft Research survey of 319 knowledge workers adds a self-reported angle. The more confidence people had in generative AI, the less critical thinking they reported putting into tasks. The more confidence they had in themselves, the more they reported.

How to use AI summaries without paying for it

None of this means giving up summaries. It means changing what you ask for and when.

Decide whether you’re looking something up or learning it. For a quick fact, a summary is the right tool. Melumad’s own advice is that the cost appears when the goal is deeper knowledge you’ll use later.

Think first, then ask. Before you open a summary of an episode or article you’ve consumed, write two sentences on what you think the main point was. Then compare. This is the think-first order from the Frontiers study, and it takes a minute.

Ask for questions, not answers. The tutor-style setups in the Bastani and Liu studies worked because they held the answer back. You can do that yourself:

I just listened to a podcast episode on [topic]. Don't summarize it.
Ask me five questions, one at a time, about its main ideas.
After each of my answers, tell me what I got right, what I missed,
and point me to the part of the transcript I should reread.

Use the summary as a map, not the destination. Let it tell you which 10 minutes of a two-hour episode are worth your attention, then go and listen to or read those 10 minutes yourself.

Close it and recall. After reading a summary, close the tab and write down the three points you’d act on. If you can’t, you’ve read it, but you haven’t kept it.

Where the evidence runs out

Most of these studies are short. Sessions lasted minutes or a few hours, and most tests came right afterwards. Only Barcaui’s waited weeks. Nobody yet knows what years of daily use do, in either direction.

The tasks are also narrow: fractions, SAT passages, essays, one coding library, gardening advice. The MIT study is small and not yet peer-reviewed. The Microsoft survey is self-reported and correlational. And participants in lab studies use AI the way an experiment allows, not the way they’d choose to.

What does hold across very different studies is the narrow claim. AI doesn’t make you dumber by being there. Letting it do your thinking before you’ve tried makes you worse at thinking without it, and that’s something you control every time you open a chat.

One thing to try

Next time you’re about to ask for a summary of something you want to remember, stop and write two sentences first. Then ask the AI what you missed.

This story is published under the Generative AI publication. Connect with us on LinkedIn and follow Zeniteq to stay in the loop with the latest AI stories. Let’s shape the future of AI together!

Does AI Make You Dumber? What the Studies Actually Measured was originally published in Generative AI on Medium, where people are continuing the conversation by highlighting and responding to this story.

The fastest way to summarize a podcast transcript with ChatGPT, Claude or Gemini is to paste the whole thing in and type “summarize this.” For a 20-minute episode, that’s usually fine.

For a two-hour interview, it tends to go wrong in two quiet ways: the middle of the conversation thins out, and some of the quotes come back smoother than anything the guest actually said.

Neither problem is a bug in one model. Both show up in published research on how language models handle long inputs, and both are easy to work around once you know where they come from.

Below is the workflow I’d use: split the transcript, extract before you summarize, and check the output against the source before you rely on it.

How long a two-hour transcript actually is

It helps to see the size of the job. The published transcript of Lex Fridman’s March 2024 conversation with Sam Altman runs just under two hours and comes to about 20,000 words, with roughly 500 timestamps. That fits comfortably inside the context window of every major chatbot today.

Fitting isn’t the problem. Using all of it evenly is.

Failure one: the middle goes missing

In 2023, Nelson Liu and colleagues at Stanford tested how well language models find information placed at different points in a long input. Performance was highest when the relevant passage sat near the beginning or end, and dropped significantly when it sat in the middle, even for models built for long contexts. They called it “lost in the middle.”

That study used question answering. A year later, Mathieu Ravaut and colleagues checked whether summaries have the same weakness. Across 6 models and 10 datasets, they found that the same U-shape holds for summarization: models lean on the start and end of the input and largely ignore what’s in between. When they built a test set where the important content sat in the middle, the models did noticeably worse.

A related study of book-length summaries, called FABLES, found that models systematically over-emphasized events near the end of the book and omitted important earlier material.

For a podcast, that’s a real loss. Hosts open with small talk and close with “where can people find you.” The part where the guest explains how they actually did the thing is usually somewhere around minute 40 to 80.

Ravaut’s team tested two fixes. One was summarizing the input in blocks and then summarizing the summaries. The other was building the summary up block by block. Both helped, especially for smaller models. They also tried simply asking the model to pay attention to the middle.

That worked reasonably well for GPT-3.5 but did little to change how the open models leaned on the start of the input. So asking nicely isn’t a strategy. Changing the shape of the input is.

Failure two: quotes that were never said

The second problem is faithfulness. The FABLES researchers had people who had read each book check 3,158 claims made in AI-written summaries. Most of the unfaithful claims concerned events and the state of characters, and they usually took some reasoning over the story to spot. A sentence could sound right and still be wrong.

The finding that matters most for this workflow: when the researchers asked language models to judge faithfulness, none of the AI raters agreed strongly with the human annotators, especially at catching the unfaithful claims. Asking the same chatbot “is this accurate?” is a weak check.

Outside the lab, the picture is similar. In a 2025 study coordinated by the EBU and led by the BBC, journalists reviewed more than 3,000 answers from ChatGPT, Copilot, Gemini and Perplexity about news stories.

31% had serious sourcing problems, meaning missing, misleading or wrong attributions. That’s news, not transcripts, but attribution is exactly what goes wrong when a summary puts a sentence in a guest’s mouth.

The workflow

Step 0: Get a transcript with timestamps and names

On YouTube, open the video description and click “Show transcript.” Many shows also publish their own transcripts, which are usually cleaner than auto-captions. Keep the timestamps and speaker names. They’re what make the later checks possible.

Step 1: Split it into chunks of 20 to 30 minutes

Use the timestamps to cut the transcript into four to six pieces, ideally at natural topic breaks. This is the “summarize in blocks” fix from the Ravaut paper, done by hand. It also means each chunk gets the model’s full attention, rather than competing with an hour of text on either side.

Step 2: Extract from each chunk, don’t summarize yet

Paste one chunk at a time with this prompt:

Below is one section of a podcast transcript (timestamps [START] to [END]).
Work only from this section.

List every distinct claim, recommendation, story or number the guest gives.
For each one:
- one plain sentence describing it
- the timestamp where it starts
- a short verbatim quote (8 to 20 words) copied exactly from the text

Rules:
- Copy quotes character for character. Do not tidy grammar or fill gaps.
- If you can't find an exact quote for a point, write NO QUOTE.
- If it's unclear who is speaking, write SPEAKER UNCLEAR.
- Leave out ads, intros and small talk.

[PASTE CHUNK]

Extraction is a narrower job than summarizing, so there’s less room to drift. The NO QUOTE and SPEAKER UNCLEAR options matter. They give the model a permitted way to say “I don’t know” instead of inventing something to fill the slot.

Step 3: Summarize the notes, not the transcript

Once every chunk is done, start a fresh chat and paste only the extracted notes:

These are notes extracted from a podcast, in order, with timestamps.
Write a summary of the main ideas.

- Keep the timestamp next to every point.
- Use only what is in the notes. Add nothing from outside knowledge.
- Give each section of the episode fair weight. Don't let the
  final section dominate.
- End with a short list of anything that looks contradictory
  or that the guest said with low confidence.

Because the model is now working from a few hundred lines rather than 20,000 words, the middle of the episode is no longer buried in the middle of a giant input.

Step 4: Turn it into actions, and label who said what

Summaries are good for remembering what was said. If you listened to learn something, you want the next step:

From these notes, list up to five concrete actions a listener could
take this week. Mark each one:
- GUEST: the guest recommended this directly (give the timestamp)
- INFERRED: you derived it from what they said

Don't present an INFERRED action as the guest's advice.

The GUEST/INFERRED label is the single most useful line in the whole workflow. Language models are good at turning a vague remark into a crisp instruction, and the crisp version is easy to misremember later as something the guest told you to do.

The five-minute check

Don’t skip this, because the research above says the model can’t reliably do it for you.

1. Search the quotes. Pick every quote you plan to reuse, copy four or five distinctive words, and Ctrl+F the transcript. If the phrase isn’t there, the quote isn’t real, however plausible it sounds.
2. Count by chunk. Look at how many points came from each section. If the second hour produced two points and the first 20 minutes produced eight, something was skimmed.
3. Spot-check one timestamp at random. Jump to it in the transcript or the video and read 30 seconds either side. Does the point match what’s there, including who said it?
4. Read the INFERRED list with suspicion. These are the model’s ideas, which can be useful, but they belong to the model, not the guest.

Which tool to use

The workflow works the same in ChatGPT, Claude, and Gemini, because it doesn’t depend on any one model’s features. NotebookLM takes a different route: you upload the transcript as a source, and its answers come with citations you can click back to the passage, which makes step 3 of the check faster. It still runs on a language model, so the same rules about quotes apply.

What this doesn’t fix

If the transcript is wrong, the summary will be too. Auto-captions mishear names, numbers, and jargon, so treat any figure in a summary as a pointer to go and listen to that moment, not as a fact.

It’s also slower. One prompt takes a minute; this takes ten to fifteen. For an episode you’ll never think about again, the one-prompt version is fine. The longer workflow is for the episodes where you want to act on what you heard or repeat it to someone else.

Models will keep improving, and some of these problems may shrink. But the habit of keeping timestamps, separating what was said from what was inferred, and checking quotes against the source will stay useful whatever model you use next year.

One thing to try

Take the last long episode you meant to learn from. Run step 2 on just the middle chunk and compare it with what you remember. If the notes surprise you, the middle was exactly the part you, and a one-shot summary, would have lost.

This story is published under the Generative AI publication. Connect with us on LinkedIn and follow Zeniteq to stay in the loop with the latest AI stories. Let’s shape the future of AI together!

How to Summarize a Podcast Transcript With ChatGPT Without Losing the Middle was originally published in Generative AI on Medium, where people are continuing the conversation by highlighting and responding to this story.

You don’t put off the hard email because you can’t find 20 minutes. You put it off because opening it feels bad. So if you want to know how to stop procrastinating, start with that feeling, not your calendar. Psychologist Fuschia Sirois, who has studied procrastination for more than 20 years, put it plainly on the American Psychological Association’s podcast: “we’re not avoiding the task per se, what we’re avoiding is the negative emotions associated with that task.”

That one sentence changes what you should do next.

I co-founded Podex, where we turn long podcast episodes into short playbooks with steps you can run this week. That’s the lens I used on Sirois’s 36-minute interview: what does it tell you to do on Monday morning? Quite a lot, once you check her claims against the studies behind them.

The episode is Speaking of Psychology, episode 210, “Why we procrastinate and what to do about it” (October 2022). APA publishes a full transcript, and the timestamps below are from the episode audio in APA’s feed.

Why time-management tips keep missing

Most advice on procrastination is about structure. Make a to-do list, use a Pomodoro timer, break the task down, block your calendar. Some of that helps, and we’ll get to which parts. But Sirois thinks it aims at the wrong target.

At 6:17 she describes the common wisdom: “teach somebody to manage their time better and they’ll stop procrastinating, or give them better self-control or tell them they’re lazy and crack the whip.” Her verdict is that this looks at “the symptoms of procrastination rather than the causes.”

The biggest review of the field agrees with her more than it disagrees. In 2007, Piers Steel pooled 691 correlations from procrastination research. Among the strongest and most consistent predictors was task aversiveness, meaning how unpleasant the task feels. Others were how far away the reward is, low confidence that you’ll do it well, and impulsiveness. Being disorganized is on the list too. But notice how many of those are about how the task feels to you, not how much time you have.

Sirois tells a story at 18:13 that makes the same point. A journalist told her he hated cleaning, “but whenever I had a big paper or exam coming, boy, was my house ever clean.” As she puts it, procrastinators aren’t sitting around being lazy. “They’re doing other things. It’s just not that thing.”

Procrastination isn’t a shortage of time or effort. It’s effort spent somewhere that feels safer.

That’s also why being busy is such a good disguise: it feels productive. I wrote about my own version of being busy without building anything in I Traded Crypto for Two Years. I Was Busy, but I Wasn’t Building Anything.

The relief is the trap

If procrastination is about feelings, why is it so hard to stop? Because it works, briefly.

“What procrastination does is it gives you immediate relief,” Sirois says at 7:58. “That’s why we say it’s sort of a short-term mood repair.” You push the report aside, you feel better, and at 8:31 she names the problem: “it’s immediately reinforcing. You get rewarded for doing that.” Do that enough times and putting things off becomes the habit your brain reaches for whenever a task feels heavy. Sirois and Tim Pychyl made the same argument in a 2013 review: procrastination puts how you feel now ahead of the goal, and the bill goes to your future self.

The bill is real. In a classic 1997 study, Dianne Tice and Roy Baumeister followed students across a semester. Procrastinators reported less stress and less illness early in the term, then more stress and more illness late in the term. Overall they were sicker, and they got lower grades on every assignment.

Procrastination works, just not for long. That’s exactly why it’s so hard to stop.

For founders and anyone else who runs their own work, the task that gets avoided often isn’t the one that takes the longest. It’s the one where someone might judge the result: the pricing page, the cold email, the investor update, the post you’re not sure is good enough. Her own example, at 8:05, is a work report where “you’re worried about what your manager’s going to think because maybe a promotion rides on it.”

So the fixes worth trying go after the feeling. Here are three, with the evidence behind each and how strong it is.

1. Shrink the forecast before you start

Near the end of the interview, at 31:48, Sirois describes what she’s studying next: the idea that we “make emotional mountains out of molehills.” We imagine how hard, complicated and painful the task will be, and that’s enough to keep us from starting.

At 32:10 she links this to affective forecasting, the research on how badly people predict their own future feelings. “We predict that we’re going to feel a lot worse when we’re doing the task than when we actually get on and just do it,” she says. “And we also think that it’s going to last longer.”

She’s careful to say this part is still being tested in procrastinators specifically. But you can test it on yourself in about ten minutes. Before you start the avoided task, write down how bad you expect it to feel, from 1 to 10. Work on it for ten minutes. Then write down how bad it actually felt. If your forecast keeps coming in higher than reality, you have your own evidence that the dread is lying to you, and that evidence is easier to trust next time than any pep talk.

2. Forgive yesterday’s delay instead of punishing it

This is the one that feels wrong, and it’s the one with the most interesting data.

At 10:18 Sirois says it “sounds counterintuitive” because we like to “crack the whip and get going.” Then she cites the study: students “who forgave themselves for procrastinating on academic work were found two weeks later to actually procrastinate less.”

That’s a 2010 study by Michael Wohl, Tim Pychyl and Shannon Bennett, titled “I forgive myself, now I can study”. They followed 119 first-year students across two midterm exams. Students who forgave themselves for procrastinating before the first exam procrastinated less before the second, and the effect ran through lower negative feelings about it. Sirois’s own work points the same way. Across four samples of students and adults, she found that procrastinators had lower self-compassion and more stress, and self-compassion explained part of that link.

Self-compassion here isn’t a free pass. At 11:08 Sirois gives the script: “I’m struggling with this task. I’m having difficulty, but you know what, it’s okay. Other people have struggled with this task too. There’s nothing special about my struggle.” At 34:00 she shortens it: “You’re not the first person to procrastinate, nor will you be the last.” Then get on with it.

Beating yourself up about yesterday’s delay adds a second bad feeling to the task you’re already avoiding.

3. Decide the start, and decide what you’ll do with the dread

Planning still matters. It just works better when the plan covers the feeling too.

The best-supported planning tool is the implementation intention, an if-then plan for when and where you’ll act. Peter Gollwitzer and Paschal Sheeran’s meta-analysis of 94 independent tests found a medium-to-large effect on reaching goals (d = 0.65), and they help with the hardest part, getting started. It’s the same cue-then-action idea I wrote about in Huberman’s 21-day habit plan.

There’s also direct evidence that training the emotional side reduces procrastination. In a 2016 paper, Marcus Eckert and colleagues ran a randomized controlled trial in which training people to tolerate and modify unpleasant emotions reduced procrastination. In the first of the paper’s three studies, the ability to tolerate those emotions was what linked emotion skills to less procrastination.

Put the two together and you get a plan with two lines instead of one:

If it’s 9:00 and I’m at my desk, then I open the pricing doc before I open email.

If I open it and feel the dread, then I write one bad sentence and keep going for ten minutes.

The first line removes the decision. The second line tells you what to do when the feeling shows up, because it will.

Where this advice runs out

Sirois is open about the limits of her field. At 29:20 she says procrastination is close to impossible to study in a lab, because it has to be a task the person intended to do. So most of the evidence comes from students, self-report questionnaires and diaries. The forgiveness study followed students from one psychology course across two exams.

Not every delay is procrastination, either. At 3:12 she quotes Pychyl: “all procrastination is delay, but not all delay is procrastination.” Waiting for information you genuinely need before starting is what she calls sagacious delay, and it’s just good judgment.

And if procrastination is wrecking your work or your health, or comes with ADHD, anxiety or depression, these three moves are a starting point, not a treatment. Talk to a professional.

One thing to try today

Pick the task you’ve moved to tomorrow at least twice. Write your dread score from 1 to 10 on a sticky note. Say Sirois’s line to yourself: other people have struggled with this too. Then work on it for ten minutes and write down how it actually felt. If the second number is lower, you’ve just caught your own forecast overshooting, and tomorrow’s start gets a little easier to argue for.

Disclosure: I’m a co-founder of Podex. We turn long podcast episodes into short playbooks with the steps worth keeping (200+ summaries so far). If you’d rather get the steps from an interview like this one without listening to all of it, you can paste an episode link into Podex.

The format barely matters for simple facts. What you do with your attention, and with the pause button, matters a lot.

If you’ve ever felt slightly guilty that your “reading” this year was mostly podcasts and audiobooks, the best evidence we have should cheer you up. Is listening as good as reading? For understanding the same text, mostly yes. In 2022, Virginia Clinton-Lisell pooled 46 studies with 4,687 participants that compared reading a text with listening to it, and across all of them the difference in comprehension was close to zero.

That’s where most articles stop. It’s also the least useful part of the paper.

I co-founded a company on the bet that people keep much less of what they hear in long podcasts than they think they do. So I wanted to know where that loss actually happens. Is it the ear, or is it something we do while we listen? Once you read past the average, the answer is mostly the second one. And the details tell you how to listen differently.

The average is a tie

Start with the good news, because it holds up.

Clinton-Lisell’s meta-analysis reported an overall effect size of g = 0.07, which wasn’t statistically different from zero. In plain terms: line up people who read a passage and people who heard it, test them, and the two groups land in almost the same place.

A smaller experiment makes this concrete. In 2016, Beth Rogowsky, Barbara Calhoun and Paula Tallal randomly assigned 91 adults to get the preface and a chapter of a nonfiction book as a digital audiobook, as e-text on a tablet, or as both at once. Everyone took the same written comprehension test right away and again two weeks later. There were no significant differences between the groups at either point.

So the ear isn’t a broken channel. For an adult who already reads well, sound carries meaning about as well as print.

One detail from that study is worth keeping. The third group, who read the text while hearing it, didn’t do better than either of the others. Doubling up the input added nothing. Plenty of people who struggle to keep what they hear reach for a transcript to read along with, and this result suggests that on its own, it won’t change much. The fix has to be something other than more input.

But averages hide conditions, and the meta-analysis found three that matter.

Where reading still wins

The first is control over pace. When readers could move at their own speed, reading came out ahead of listening (g = 0.13). When the experimenter set the pace, the reading advantage disappeared. Reading lets you slow down at a hard sentence, skip back two lines, or stop and think. Listening, by default, keeps going whether you followed or not.

The second is the kind of understanding being tested. For literal comprehension, meaning facts stated directly in the text, reading and listening were the same (g = -0.01). For inferential comprehension, meaning conclusions you have to work out yourself, reading had a clearer edge (g = 0.36). That’s still a modest effect. But it sits in exactly the place that matters most when you learn from a long interview, where the useful part is rarely a sentence the guest says outright. It’s the connection you draw between what they said at minute 12 and what they said at minute 58.

Rogowsky’s team made a related point. They noted that earlier studies on adults conflict, and suggested one reason: whether participants could go back and study the material before the test. When people can return to the text, the format starts to matter more, because text is easy to revisit and audio isn’t.

The third condition is attention. It’s the most important one, and the meta-analysis can’t see it directly, because lab participants are usually sitting at a desk with nothing else to do.

The attention tax

In 2010, David Daniel and William Douglas Woody gave college students a primary source either as text or as a podcast. The podcast group did worse on the quiz. The students had preferred the podcast before the quiz, and their preference flipped right after it. As the University of Delaware’s Stephanie Del Tufo describes the study, students used the material on their own time, and many of the listeners reported multitasking, such as clicking around on their computers while the podcast played.

A 2013 study from the University of Waterloo measured the same problem in real time. Trish Varao Sousa, Jonathan Carriere and Daniel Smilek had 235 students read one excerpt from Bill Bryson’s A Short History of Nearly Everything silently, read another aloud, and listen to a third read by Bryson himself. At random moments a box asked whether they’d been mind wandering. Listeners said yes about half the time (a proportion of 0.506). Silent readers said yes about a third of the time (0.356). Listening also produced the worst memory scores of the three conditions.

The authors suggest a reason I find convincing. When you read, your eyes have to keep moving, and when they stop, you notice. When you listen, nothing about the task tells you that you’ve drifted. The voice carries on. You can lose 40 seconds and never know it happened.

Listening doesn’t lose to reading because the ear is weaker. It loses because drifting off while listening doesn’t feel like anything.

Reading aloud produced the least mind wandering of all, yet its memory scores were about the same as silent reading. Producing words kept people on task, even if it didn’t, by itself, make them remember more. That fits a pattern you’ll see again below: what helps is anything that makes you do something with the material instead of letting it pass through.

Not all multitasking is the same

Here the evidence gets more interesting, because it doesn’t simply say “never listen while doing something else.”

In a 2022 study in Cureus, Wolpaw and colleagues had 61 anesthesiology residents and medical students learn three topics: one from a podcast while seated, one from a podcast while walking on a treadmill, and one from a textbook. EEG measures of attention didn’t differ between the conditions. Podcast learning, seated or walking, produced a better immediate learning gain than the textbook, and equal retention for two of the three topics at four weeks.

In 2025, a McMaster University team put 19 resident physicians and 22 undergraduates in a high-fidelity driving simulator. They listened to medical education podcasts while driving through a busy city, while driving on empty country roads, or while sitting at a desk. Across the three conditions, recall wasn’t significantly different. Busy city driving lowered immediate recall slightly compared with country driving, but that gap was gone at the one-month test.

Put next to Daniel and Woody’s students clicking around their screens, these results suggest a useful line. The McMaster authors lean on a theory called threaded cognition: tasks that use different mental resources, such as steering with your eyes and hands while processing speech with your ears, can run side by side with little interference. Tasks that compete for the same resource collide. Browsing, reading messages and answering email all use language. Walking and driving a familiar road mostly don’t.

Listening while your hands are busy costs little. Listening while your words are busy, in email, messages or a feed, costs a lot.

Two cautions. Both trials used medical content with medical learners, small samples, and one institution each. And recall fell for everyone between the immediate test and the one-month test, whatever the condition. Listening while walking may be fine for taking ideas in. It does nothing on its own to keep them, which is the problem I wrote about in how to retain information from podcasts.

Speed is less of a problem than you’d think

If attention is the real constraint, faster playback seems like it should make things worse. The data say otherwise, up to a point.

Dillon Murphy and colleagues at UCLA showed students lecture videos at different speeds and tested them right away and a week later. Going from normal speed to 1.5x or 2x cost very little, and comprehension dropped beyond 2x. Watching a video twice at 2x wasn’t better than watching it once at normal speed. But using the saved time to rewatch at 2x shortly before the test did beat a single normal-speed viewing a week earlier.

That’s a lecture with slides, not a two-hour conversation, so treat it as a rough guide. The pattern still helps: moderate speed costs little, and what you do with the time it saves is the part that counts.

The fix that works in a lab

The most practical result in this whole literature comes from a 2013 Harvard study that wasn’t about podcasts at all.

Karl Szpunar, Novall Khan and Daniel Schacter had students learn from a 21-minute statistics lecture, split into four segments of about five and a half minutes. Some students got a short quiz after each segment. The quizzed students reported less mind wandering, took more notes, learned more, and felt less anxious about the final test.

Notice what the quizzes did. They didn’t make the lecture more interesting. They gave students a reason to stay with it, because a test was always a few minutes away. In effect, they put back the signal that listening lacks: a moment where drifting off becomes visible.

A podcast has no quizzes. But you can add the pause.

How to listen so you keep more

None of these studies tested a founder interview or a three-hour science podcast, so this is an inference, not a finding. With that said, here’s how I’d listen if I wanted to keep what I heard.

Match the format to the job. For a story, a news update or plain facts, listening is as good as reading, so listen anywhere. For an argument you want to use, where the value is in the reasoning, you’ll do better with something you can slow down and reread. That might mean pulling up the transcript for the 10 minutes that matter. A guest walking through how they priced their first product is a good example: the story is easy to follow by ear, but the reasoning behind each decision is the part you’d want written down.

Listen while your eyes and hands are busy, not while your words are. Walking, driving a familiar route, washing up: the evidence above says these are fine. Email, messages and scrolling are not. They use the same channel the podcast does.

Add your own quiz. Every 15 or 20 minutes, or at a natural break, pause and say out loud what the last segment argued. If you can’t, you drifted, and now you know. It’s the closest a listener can get to Szpunar’s quizzes. I used the same idea of building checkpoints into a plan in Huberman’s 21-day habit system, and it works for the same reason: it makes slipping visible.

Use speed to buy review time. Listening at 1.5x on a long episode saves real time. Spend part of it going back over the one section you most want to remember, ideally shortly before you plan to use it.

Where the evidence runs out

Most of this research uses short passages and students in quiet rooms. Long-form podcasts are conversational, repetitive and emotionally engaging. Those features might help memory, and they might also make drifting easier. Nobody has tested that properly.

The effect sizes are small to moderate. A g of 0.36 for inference is meaningful, but it doesn’t mean readers understand and listeners don’t. It means readers, on average, do somewhat better on questions that need working out.

The medical podcast trials, the ones most favorable to listening, are small. They’re encouraging, not the last word.

And the two studies most often quoted against listening, Daniel and Woody’s and the Waterloo one, are more than a decade old. Both tested students on material someone else picked. People choose their own podcasts, usually because they care about the topic, and in the Waterloo data interest went with less mind wandering. A listener who picked the episode may drift less than a student who was assigned it.

The narrow claim survives all of that. Your ears aren’t the problem. Unnoticed drifting is, and so is treating listening as the only step. Both can be fixed with a pause button and an honest attempt to say what you just heard.

One thing to try today

On your next episode, pause at the first natural break and say, in two sentences, what the guest just argued. If it comes easily, keep going. If it doesn’t, rewind five minutes and listen again with your phone face down.

Disclosure: I’m a co-founder of Podex. We turn long podcast episodes into short, action-first lessons (200+ summaries so far), built around the one step worth keeping. If you’d like the takeaways in a form you can reread, you can try Podex here.

The short answer to how to build a habit: pick a small behavior, tie it to a cue you already hit every day (“after I pour my coffee, I…”), put hard habits early in your day, and expect it to take around two months, not three weeks. Missing one day doesn’t undo it. Then check after 21 days which habits run on their own, and fix the rest.

That’s the whole method. The rest of this piece shows where each part comes from, because the moment most plans fail is predictable. It’s the second week. You skip one tired evening, and by Friday the habit feels like something you used to do. Most advice answers that moment with a number: 21 days, 30, 66.

Andrew Huberman spends a section of his January 2022 episode, “The Science of Making & Breaking Habits”, taking that idea apart. Then, more than an hour later, he offers his own 21-day system. Those two things sound like they contradict each other. They don’t, and the reason why is the most useful thing in the episode.

I co-founded Podex on the belief that people keep far less of what they hear in long episodes than they think, even from the good ones. This one runs close to two hours. So here is what it actually tells you to do, where its main number comes from, what newer research adds, and a plan that fits on one page. The timestamps are the episode’s own section markers, so you can check any of it yourself.

How long does it take to build a habit?

In the section that starts at 11:40, Huberman goes after the 21-day rule directly. He cites a 2010 study by Phillippa Lally and colleagues at University College London and says it found that “it can take anywhere from 18 days to as many as 254 days for different individuals to form that habit.”

That range is right. But the way the episode describes the study is a little off, and it matters.

In the episode, Huberman says the study looked at one specific behavior, taking a walk after dinner, and that the same habit took some people 18 days and others 254. The paper itself describes something different. Ninety-six volunteers each chose their own new eating, drinking or exercise behavior and did it daily in the same context for 84 days. The examples in the paper are things like “eating a piece of fruit with lunch”, “drinking a bottle of water with lunch” and “running for 15 minutes before dinner”.

So the 18 to 254 day spread comes from many different habits, chosen by the people doing them, rather than one habit across many people. Drinking water with lunch and running before dinner are not equally hard, and it shouldn’t surprise anyone that they take different amounts of time to become automatic.

Two more details from the paper are worth knowing. The median time to reach close to full automaticity was 66 days, among the 39 people whose data fit the researchers’ model well. And missing a single day “did not materially affect the habit formation process.”

Newer evidence points the same way. A 2024 systematic review from the University of South Australia pooled 20 studies with 2,601 participants. Across the four studies that reported a time to habit, the median was 59 to 66 days and the average was 106 to 154 days, with individual times anywhere from 4 to 335 days. The same review found that habits done in the morning and habits people chose for themselves tended to be stronger.

Habits take about two months on average, not 21 days. The range runs from under a week to most of a year, and it depends on how hard the habit is for you.

The authors are careful to say that most of the included studies carried a high risk of bias, so treat two months as a rough middle, not a deadline. The useful part is less about 66 versus 21 and more about this: the number depends on how hard the habit is for you. Which is exactly where the episode goes next.

Why some habits feel harder than others

Huberman’s central concept is something he calls limbic friction. He’s upfront that this is his own label:

“Limbic friction is not a term that you’re going to find in the formal biological literature or even psychological literature. It’s frankly a term that I coined.” (Huberman Lab, “The Science of Making & Breaking Habits”, section starting at 11:40)

He uses it to describe the effort it takes to get yourself to do something. In his framing, that resistance comes from one of two states. Either you’re too anxious and wound up to settle into the task, or you’re too tired and flat to start it. A morning run is hard when you’re exhausted. Meditating is hard when you’ve just walked in from a stressful day.

You don’t have to accept his neuroscience to find the label useful. It gives you a question to ask about each habit you want: how much resistance do I feel when I imagine starting this, today, as I am? That one question does most of the sorting work in the plan below.

It also explains why the most common advice in habit guides, start tiny, works. Stanford’s BJ Fogg, who built the Tiny Habits method, told NPR’s Life Kit to scale a new habit back until it’s “super-tiny”: read one paragraph, or take three calming breaths. A tiny version has almost no friction, so it survives the tired days that kill the full version.

Tie the habit to a cue, not to willpower

One thing Huberman’s episode doesn’t stress, and most top-ranking habit guides do, is the cue.

Habits are behavior you repeat in the same situation until the situation starts it for you. In one diary study, Wendy Wood and colleagues found that about 43% of the everyday behaviors people logged were done almost daily and usually in the same place. A lot of your day already runs on cues. The trick is to borrow one.

The best-tested way to do that is an if-then plan, which researchers call an implementation intention: “When situation X happens, I will do Y.” A 2006 meta-analysis by Peter Gollwitzer and Paschal Sheeran pooled 94 tests with more than 8,400 people and found a medium-to-large effect on reaching goals (d = 0.65) for people who made these plans compared with people who only set the goal.

Habit stacking is the same idea with an existing habit as the “if”. Fogg’s own wording on NPR is a good template: “After I sit down on the subway, I’ll open my book and read a paragraph.” Write yours in that exact form.

A plain if-then plan (“When X happens, I will do Y”) was one of the strongest effects in the goal research: d = 0.65 across 94 tests.

There is one tension to know about. Huberman argues against tying a habit to an exact clock time (more on that below), while the if-then research says be specific about the situation. These fit together if your “if” is an event, like finishing breakfast or closing your laptop, rather than a minute on the clock.

Four tools from the episode worth keeping

The episode has six numbered tools. Four of them fit together into a single plan.

Walk through the steps once (24:03). Before starting a new habit, Huberman suggests thinking through the exact sequence of steps from start to finish. He doesn’t mean a guided meditation. His example is making espresso every morning: walk into the kitchen, turn on the machine, pull the shot. He says this one mental run-through makes the habit more likely to stick and that “the literature is really robust” on it. I haven’t checked that claim here, so treat it as his. It costs about a minute, which makes it an easy bet.

Anchor to a state, not a clock (37:08). This is the episode’s most contrarian point. Huberman says being very specific about exactly when you’ll do a habit helps in the short term but not the long term, because behavior follows your state (how alert, focused or tired you are) more than the time on the clock. So instead of “7:15 a.m., run”, the plan uses broader windows of the day, with an event as the trigger.

Match the habit to the phase of the day (38:16). Huberman splits the day into three phases, counted from when you wake up. Phase one is roughly the first eight hours. In his account, this is when you’re most alert and action-oriented, so it’s where the hardest habits go: “take the habits for which you know there’s the highest degree of limbic friction… and put those in this zero to eight hours after waking.” Phase two, about nine to fourteen or fifteen hours after waking, is for what he calls “mellower activities”, and his examples are journaling, practicing music, or learning a language. Phase three is the rest of the cycle and is mostly about sleep. If you wake at 7 a.m., phase one ends around 3 p.m. and phase two runs until about 9 or 10 p.m. The 2024 review’s finding that morning habits tend to be stronger fits this part.

Use the 21-day install-and-test system (1:18:22). This is the part people remember, and it’s more careful than the “21 days to a habit” slogan it sounds like. You write down six new habits. You try to do them every day for 21 days, but “the expectation is that you’ll only complete four to five of those each day.” If you miss, you don’t make it up the next day by doing extra. He treats the 21 days as a run of two-day blocks: get four or five done for two days, then reset. And then comes the important bit. For the next 21 days you add nothing new. You just watch which of the habits keep happening without effort. Those are the ones that stuck.

Read that way, the 21 days isn’t a claim about how long a habit takes to form. It’s a test window. That’s why it doesn’t contradict the research. The first 21 days install the habits, the second 21 days tell you which ones you’ve really got, and the ones that didn’t survive go back on the list.

A one-page plan to build a habit

Here’s the episode and the research turned into something you can write on an index card.

1. Pick three to six habits you want. Six is Huberman’s number. If you’ve failed at habit plans before, start with three. You’ll learn more from three that survive than from six that collapse in week two.
2. Shrink each one. Write the smallest version you’d still count: ten pages, not a chapter; a 10-minute walk, not a 5K.
3. Rate the resistance. For each one, ask how much limbic friction you feel imagining doing it today: high, medium or low. Be honest. A 20-minute workout might be high for you and low for someone else.
4. Place each habit in a phase and give it a cue. High resistance goes in your first eight hours awake. Low and medium go in the late afternoon and evening. Write it as an if-then: “After I close my laptop for lunch, I run.” A worked example: say your three habits are a 20-minute run, reading 10 pages of a nonfiction book, and writing down one thing you learned that day. For most people the run is the high-resistance one, so it goes in the morning or at lunch. Reading and the one-line note are low effort, so they sit in the evening, where Huberman would put journaling. Your ratings may differ. The point is that you place each habit by how it feels to you, not by what a morning-routine video says.
5. Walk through each one once. For every habit, think through the steps from the first physical action to the last. Where are your shoes? Which app opens? What do you do the moment it’s done? This takes a minute per habit.
6. Run 21 days with a pass mark. A good day is most of your list, not all of it. With six habits, that’s four or five. With three, it’s two. If you miss a day, the next day is a normal day. No doubling up. Lally’s data back this part: a single missed day didn’t derail habit formation.
7. Run 21 days with nothing new. Stop tracking on purpose. Note which habits you’re still doing without having to push yourself. Those are installed. The rest either move to a different phase of the day, get smaller, get a better cue, or go back on next round’s list.

That’s 42 days, which lands inside the range both studies found. For simple habits, it may be more than enough. For hard ones, the second 21 days tells you whether you need more time, which is more useful than a promise that you’re done.

If the habit you keep failing at is a work task you dread rather than a health routine, the problem is often the feeling, not the plan. I wrote about that in How to Stop Procrastinating When You Already Know What to Do.

Where the episode and the research are weaker

A few honest limits, some of which Huberman states himself.

He says plainly that the 21-day system has nothing special about it scientifically: “there’s nothing neuroscientifically unique about it, except for the fact that it’s not just 21 days broken up into two-day chunks.” It’s a structure he likes, and he says it comes from someone he knows who has used it for a long time. That’s a good reason to try it, not proof that it works better than other plans.

Limbic friction is his coined term, as he says. The neurochemistry he attaches to each phase of the day (more dopamine and norepinephrine early, more serotonin later) is his summary, and this piece doesn’t check it. The phases also assume a normal daytime schedule. He notes that shift workers and parents of newborns will need to adjust.

The transcript on the Huberman Lab site carries a note that it is still under human review and may contain errors. The quotes above match it, but listen to the sections at the timestamps if you plan to rely on the exact wording.

And the research base is thinner than the confident numbers suggest. Lally’s study is small: only 39 people’s data fit the model well, and automaticity was self-reported. The 2024 review found just four studies that reported a time to habit, and most of its studies had a high risk of bias. Almost all of them are about health behaviors like flossing, water and exercise, not work habits.

Habit questions people ask

Is the 21-day rule a myth? As a claim about how long habits take, yes. The best data put the middle at around two months, with a huge spread. As a test window, like Huberman uses it, 21 days is fine.

What if I miss a day? Carry on the next day as normal. In Lally’s study, one missed day didn’t materially change the habit curve. What hurts is treating one miss as proof the plan failed.

How many habits should I build at once? Huberman suggests six and expects you to hit four or five a day. If you’ve struggled before, start with three. Fewer, smaller habits survive more often.

What’s the best time of day for a new habit? Hard habits early, when you’re most alert, and easy ones later, per the episode. The 2024 review found morning habits tended to be stronger, but the evidence is limited.

Does habit stacking work? It’s an if-then plan with an existing habit as the trigger, and if-then plans have strong evidence (d = 0.65 across 94 tests). The trigger needs to be something you really do every day, in the same place.

One thing to try today

Take one habit you’ve already started and dropped at least once. Shrink it, rate how much resistance you feel about it right now, then move it into the phase of the day that matches and write its if-then cue. Hard things early, gentle things late. Walk through its steps once, start tomorrow, and set a reminder for 21 days from now that asks one question: am I still doing this without thinking about it?

If you’re building a learning habit, like taking one idea from every podcast you hear, the hard part is the “keep it” step, not the listening. That’s what How to Retain Information From Podcasts (and Actually Use It) is about.

Disclosure: I’m a co-founder of Podex. We turn podcast episodes like this one into short, action-first playbooks (200+ summaries so far), and you can paste a link to a long episode to get its steps without the two-hour listen. If that’s useful, you can try Podex here.

Sources: Lally et al. (2010), European Journal of Social Psychology, doi:10.1002/ejsp.674. Singh et al. (2024), Healthcare 12(23):2488, https://pubmed.ncbi.nlm.nih.gov/39685110/. Gollwitzer & Sheeran (2006), Advances in Experimental Social Psychology 38, doi:10.1016/S0065–2601(06)38002–1. Wood, Quinn & Kashy (2002), Journal of Personality and Social Psychology, https://pubmed.ncbi.nlm.nih.gov/12500811/. BJ Fogg on NPR Life Kit (Feb 25, 2020). Huberman Lab, “The Science of Making & Breaking Habits” (Jan 2022), https://www.hubermanlab.com/episode/the-science-of-making-and-breaking-habits.

For a little over two years, I traded crypto. Mostly memecoins, plus some utility tokens. Some months went well and some didn’t, and the results never settled into anything consistent. It kept me occupied. It didn’t leave me fulfilled.

I’m writing this because I think a lot of people are in the same spot with something other than crypto, and because the lesson I took from trading turned out to be the exact problem I now work on. After those two years I started building a company that tries to save people time, and the thing it fights is the same trap, just with headphones on.

Motion that leaves nothing behind

Trading is very good at making you feel busy. There’s always another chart, another token, another thread to read. Any hour of the day can be filled with it.

What it didn’t give me was something that added up. In trading, a good week doesn’t make the next week easier, and a bad week doesn’t leave you holding anything you can point to. That’s a big part of why it felt unfulfilling, and why I wanted to build something real instead.

Robert Greene put words to this on The Diary of a CEO, in an episode about confidence:

“The finest sense of confidence is actually creating things and having success and meeting goals and achieving things and having a record of that.”
Robert Greene, The Diary of a CEO, E232, at 1:02:03

The phrase that matters is “a record of that.” Trading gave me plenty of activity and very little record. That’s the clearest way I can describe being busy but not productive: you can be fully occupied and still have nothing that stacks.

The same trap, with headphones on

The company I co-founded is built on one belief: people think they keep what they hear in a full podcast episode, and mostly they don’t.

That is the trading problem in a different outfit. A two-hour episode feels like learning the way a full day of watching charts feels like work. You were engaged the whole time and you agreed with most of it. Ask yourself a week later what you’re doing differently because of it, and the answer is often thin.

This isn’t only a founder’s hunch. In a 2010 study in Teaching of Psychology, David Daniel and William Douglas Woody had one group of students listen to a primary source as a podcast while another group read the same material as text. The podcast group scored worse on the quiz. The detail I like most: the students started out preferring the podcast, and changed their minds right after the quiz. Listening felt easier, and easier turned out to be part of the problem.

A second study explains what to do about it. In “Test-Enhanced Learning”, Henry Roediger and Jeffrey Karpicke had students either reread a prose passage again and again, or read it once and then practice recalling it from memory. Five minutes later, the rereaders did better. A week later it flipped. The group that practiced recalling remembered 61% of the ideas and the rereaders remembered 40%, even though the rereaders had gone through the passage about 14 times and the recall group about 3.4 times.

Put the two together and you get an uncomfortable picture. Hearing something once and moving on is close to the setup that loses the most. Relistening feels like the responsible fix, but the research suggests it mostly makes the next five minutes feel better.

What actually sticks

The question worth asking about any habit that eats your time is the one trading taught me: what does this leave behind?

For a podcast, the answer can be small. When an episode ends, close the app and write down, from memory, the one idea you’d act on. Then write the step: what you’ll do this week, and when. If you can’t pull anything out without scrolling back, that tells you something too. Maybe the episode was entertainment, which is fine, as long as you call it that.

Here’s what that looks like with the Greene clip above. The idea: confidence comes from a record of things you’ve finished, not from how you feel about yourself. The step: today, finish one thing you said you’d do, and start a list of what you finish. Two lines, written from memory, and the episode has left something behind.

That small act lines up with both studies. Pulling the idea out of your own head is the kind of practice that held up after a week in Roediger and Karpicke’s experiment. And turning it into a dated step gives you, in Greene’s words, a record.

Where this falls short

Not every episode should become homework. Some podcasts are company on a long drive, and squeezing a takeaway out of them misses the point. Both studies also used students and course-style material, not people listening to interviews for fun, so they’re a strong hint about podcasts rather than a verdict. And writing a step down isn’t doing it. The record only counts once you run the step.

I’m also not saying trading is worthless. Some people do it well and build real skill from it. My problem was that, for me, it produced motion without anything that compounded.

If you want one thing to try today: pick the last episode you finished that you’d call useful. Without opening it, write one sentence about what it argued and one action you’ll take by Sunday, and put that action in your calendar. If nothing comes to mind, you’ve learned something about how much of it stuck.

Disclosure: I’m a co-founder of Podex, which turns podcast episodes into short tactic summaries (200+ so far), each with a step you can run this week. If you’d rather start from a summary than a blank page, you can try Podex here.

I used AI tools to help draft and edit this piece; the story and opinions are my own.

Deepstash got one big thing right. It took the urge to scroll and pointed it at something useful. Instead of a social feed, you swipe through short idea cards pulled from books, articles, podcasts and videos, and you save the good ones to your own library. Its site promises access to 200,000+ ideas, a personalised growth plan, reading and listening synced across devices, and a scanner that turns pages from physical books into searchable notes. There's a free plan, and the Pro plan lists unlimited saving, listening to ideas, no ads and offline access.

That's a genuinely good product. So why do so many people search for Deepstash alternatives? The reasons usually fall into five complaints, and they point in different directions. The cards start to feel thin. The stash grows to hundreds of saved ideas you never look at again. The feed has no end, so it slowly turns into the scrolling you were trying to replace. The free version feels limited. Or you realise the ideas you actually use come from long podcast conversations, not from three-sentence cards.

Each problem has a different best answer, so that's how this list is built. All product facts come from each company's own site as of October 2026. Deepstash doesn't publish its Pro price on its website, and app-store prices vary by country, so I don't quote it. I only quote other prices where the official page shows them.

Disclosure: I'm a co-founder of Podex. It shows up in one section below, for one specific kind of reader. Everything else here is someone else's product.

A two-minute stash audit before you switch

Open your Deepstash library and look at the last ten ideas you saved. For each one, ask two questions. Can I explain this idea to someone without looking? Did I ever do anything with it?

If you can explain most of them but used none, your problem is action, and the "nothing changes" fix near the end matters more than which app you pick. If you can't explain them, your problem is depth or review. If you didn't even recognise half of them, the feed is doing the choosing for you. Keep your answer in mind as you read.

If the cards feel too thin: Blinkist, Shortform or Headway

An idea card gives you the conclusion. A good summary gives you the argument behind it, which is what you need when you want to judge whether an idea applies to your situation.

Blinkist is the natural step up. Its plans page lists 9,000+ book and podcast summaries you can read or listen to in about 15 minutes, with personal daily recommendations. As of October 2026 on blinkist.com/en/plans (US), Premium is $79.99 for the first year, then $99.99 a year, and each Premium member gets a second Premium account to share. Pro adds Blinkist AI, which summarises videos, podcasts, documents and articles, at $139.99 for the first year, then $174.99 a year. If you cancel a trial, Blinkist still gives you one pre-selected book a day.

Shortform is for people who want the whole reasoning, not a quicker version of it. Its guides go deep, with commentary, highlights and notes, audio narration, PDF downloads and exercises, and the pricing page says it covers 12,000+ books. As of October 2026, it's $16.42 a month billed annually. Pick it for the three or four books a year you actually want to change how you work. It's a reading session, not a coffee-break habit.

Headway sits between the two. Its site lists 1,700+ summaries of about 15 minutes each, plus streaks, spaced repetition and short themed plans such as "Stop delaying, start doing". If you liked Deepstash's daily rhythm but want complete books instead of loose cards, Headway is the closest swap in feel. Its price varies too much across signup paths to quote a single number, so check it in your app store.

If you save ideas but never see them again: Readwise or Anki

This is the quiet failure of every "save" button. Saving feels like learning, but a library you never reopen is a bookmark folder with nicer fonts.

Readwise fixes the review side. It pulls in your highlights from places like Kindle, Instapaper and Apple Books, then sends you a daily review that resurfaces a few of them, so the best parts of what you read come back before you forget them. You can tag and annotate highlights and export them to Notion or Obsidian. It's free for the first 30 days. After that, as of October 2026 on readwise.io/pricing, the full plan is $9.99 a month billed annually or $12.99 month to month, there's a Lite plan at $5.59 a month billed annually, and students can ask for 50% off. It suits people who already read a lot and want that reading to stay with them. It doesn't give you content, so pair it with books, articles or a summary app.

Anki is the heavy-duty option. It's a flashcard program that asks you to rate how well you recalled each card, then schedules the next review for when you're most likely to forget it. The computer version is free on Windows, Mac and Linux, and AnkiWeb sync between devices is free too. The catch is that you write the cards yourself. That's also why it works: turning a Deepstash idea into one good question forces you to understand it.

A simple bridge: once a week, pick the three Deepstash ideas you most want to keep and turn each into one Anki card or one tagged Readwise note. Three a week adds up to more than 150 ideas a year that you can actually recall.

If the feed never ends: Chunks, Imprint or Nibble

Deepstash is built to replace doomscrolling, but an infinite feed of good content is still an infinite feed. If you open it for five minutes and look up after forty, you want something with a finish line.

Chunks is built around exactly that. Its stories run four to eight minutes, are narrated with the text highlighting as it plays, and then they end. The library covers history, philosophy, literature, science, art, music, nature and health. It's free to start, and unlimited access is $9.99 a month or $49.99 a year as of October 2026. Choose it if what you loved about Deepstash was curiosity rather than self-improvement.

Imprint teaches big ideas through visual, tap-through lessons, mainly in psychology and related topics, and makes it easy to save and review the ideas you want to keep. If text cards slide off your brain but diagrams stick, try it first.

Nibble offers 670+ pieces across 35+ topics, with lessons of 3 to 10 minutes that mix quizzes, games, audio, video and chats with historical figures. It runs on iOS, Android and the web, and the first lesson is free. It's broader and more playful than Deepstash, and the short lessons give you a natural place to stop.

If you want it free: Khan Academy, Kinnu, Anki and Blinkist's daily pick

"Deepstash alternative free" is one of the most common versions of this search, so let's be precise. A 7-day trial isn't free. These keep working without paying:

• Khan Academy is a nonprofit, and its lessons, exercises and videos are free. It's structured courses rather than a feed, strongest in maths, science, economics and the humanities.
• Kinnu offers free pathways for general knowledge, built around learning a subject step by step rather than grazing.
• Anki's computer version is free, as covered above.
• Blinkist gives you one pre-selected book a day after a trial ends.

A free stack that genuinely replaces Deepstash: the Blinkist daily pick for book ideas, one Kinnu or Khan Academy lesson for depth, and Anki for the handful of ideas you want to keep for good.

If your best ideas come from podcasts: Snipd or Podex

Here's the reason that gets the least attention. Deepstash's cards come from books, articles, podcasts and videos, but many of the most practical ideas right now come from long podcast interviews. A founder explaining exactly how they priced their first product, or a researcher walking through a specific sleep routine, rarely fits on a card, and it's often more useful than a book summary written years ago.

Snipd is the best pick if you love listening and just want to keep what you hear. It's a podcast player where you tap your headphones whenever something is worth remembering, and its AI saves a snip with the audio, the transcript and a summary. It syncs those snips to Notion, Readwise, Obsidian and other note apps, and it can import YouTube videos too.

Podex is the option for people who'd rather skip the three-hour listen and go straight to what to do. Remember, I co-founded it. We built it on a simple belief: most people keep far less of a long episode than they think they do. You pick your goals, such as "Build something people want", "Make money make sense" or "Get sharper every day", and you get short, action-first summaries pulled from shows like The Diary of a CEO, Huberman Lab, Acquired and My First Million. Every tactic comes with a concrete step to run this week, and there are 200+ summaries in the library. You can also paste a link to any YouTube or podcast episode and get the tactics pulled out. Where Deepstash hands you an idea, Podex hands you the next step. If that's the gap you've been feeling, you can pick your goals and try Podex with no card needed.

If nothing changes: keep five active ideas, not five hundred

Whichever app you choose, this rule fixes more than any switch will. Keep a list of no more than five active ideas. Each one gets a single line in the form "Next time this happens, I will do that." When you've used an idea twice, it graduates and leaves the list. When it has sat there two weeks untried, delete it with no guilt. You only add a new idea when a slot opens.

It sounds restrictive. That's the point. A stash of five ideas you're running beats a library of five hundred you're admiring.

Deepstash vs Blinkist vs Headway

This comparison gets searched a lot, so here's the short version. Deepstash gives you single ideas from many kinds of sources, which makes it great for range and quick sparks. Blinkist gives you whole books and some podcasts in 15 minutes, with the biggest catalog of the three. Headway gives you whole books plus the strongest habit structure, with streaks, spaced repetition and short plans. Want range? Deepstash. A book's full argument? Blinkist. A guided daily routine? Headway.

When to stay with Deepstash

If you use Deepstash to replace social media and it's working, stay. If you love its book scanner and your own library, nothing else on this list does both in quite the same way. The fix may be smaller than a switch: run the stash audit monthly and add the five-active-ideas rule.

The quick version, by reason

• Cards too thin: Blinkist, Shortform or Headway.

Saved ideas never come back: Readwise or Anki.

• The feed never ends: Chunks, Imprint or Nibble.
• You want free: Khan Academy, Kinnu, Anki and Blinkist's daily pick.
• Your best ideas come from podcasts: Snipd to keep what you hear, Podex to turn episodes into steps.

Common questions

What is the best Deepstash alternative?

It depends on why you're leaving. For deeper content, Blinkist or Shortform. For remembering what you save, Readwise or Anki. For sessions that end, Chunks. For learning from podcasts, Snipd or Podex.

Is there a free Deepstash alternative?

Yes. Khan Academy and Kinnu's pathways are free, Anki's computer version is free, and Blinkist offers one pre-selected book a day after a trial ends.

Is Deepstash better than Blinkist?

Neither is better overall. Deepstash is better for variety and quick ideas from many sources. Blinkist is better when you want a book's full argument in 15 minutes.

Is Deepstash free?

Yes, there's a free plan. The Pro plan lists unlimited saving, listening to ideas, no ads and offline access. Its price isn't listed on Deepstash's website, so check the App Store or Google Play for your country.

Pick by the problem, not by the biggest library. That's the whole trick.

PS: What’s one idea you saved in Deepstash and actually used? I’m curious which ones survive the trip from the stash to real life.

Type “apps like Duolingo for” into Google and the suggestions tell you what people want: other subjects, coding, science, finance, sign language, anatomy. They like the five-minute lesson, the streak they don’t want to break, and they want that feeling for the thing they actually need to learn.

There’s no single app that does Duolingo for everything well. There are good apps for specific subjects, and a simple way to rebuild the Duolingo loop around anything else. This piece covers both, sorted by subject, so skip to yours.

Product details come from each company’s own site as of October 19, 2026. I’ve left prices out where they vary by country or promotion, or where an app is free to start.

Disclosure: I’m a co-founder of Podex. It appears once, under self-improvement, where it fits. Everything else on this list is someone else’s product.

What Duolingo actually gets right

Before swapping apps, it helps to name which part of Duolingo you’re missing. Three mechanics do most of the work:

1. The lesson is small enough to start. Five minutes doesn’t trigger the “I’ll do it later” reflex that an hour-long course does.
2. You produce something every few seconds. You type, tap, speak or choose. You’re never just reading.
3. It asks for you tomorrow. The streak, the reminder and the review of old material make today’s lesson part of a chain.

Any app that does all three will feel like Duolingo, whatever the subject. Apps that only copy the cartoon and the points, without that second mechanic, tend to feel like Duolingo for about a week.

Duolingo itself, beyond languages

It’s worth checking the obvious option first. Duolingo now offers Math, Music and Chess alongside its languages. You can add them from the course picker, and any lesson you finish, in any subject, counts toward the same streak.

Math covers school-level content for elementary, middle and high school students, from arithmetic through algebra, in bite-sized lessons with a built-in tutor. Chess has guided lessons, puzzles and matches against Oscar or other learners. Music teaches reading music and playing songs on an on-screen keyboard. If one of those is your subject, you get the real Duolingo loop without installing anything else.

For coding: Sololearn

Sololearn is the closest thing to Duolingo for programming. Its site describes free, bite-sized interactive courses with in-lesson quizzes, and you write and run real code with instant feedback. There are more than 20 courses covering Python, JavaScript, SQL, web development and a growing set of AI topics, plus certificates when you finish. It works on the web and in the app, and keeps your place across both.

Its limit is the same one Duolingo has with languages: it gets you started, but small exercises won’t make you a working developer on their own. Once the basics click, start building a tiny project of your own alongside it.

For maths, science and logic: Brilliant, or Khan Academy for free

Brilliant is the standout for maths, coding, data and science. Every lesson is an interactive problem rather than a lecture, which covers Duolingo’s second mechanic better than almost anything else. Its homepage now leads with an AI tutor, Koji, that asks guiding questions instead of handing you answers, on a shared canvas for maths and in a shared coding environment for computer science. Expect more effort per session than Duolingo; it rewards focus.

Khan Academy is free, run by a nonprofit, and covers maths, science, economics and more with structured practice and mastery tracking. Sessions are longer and feel more like school, but nothing beats it on price.

For money and personal finance: Zogo

“Duolingo for finance” is a common search, and Zogo is the closest match. It’s a gamified financial-literacy platform with short lessons and quizzes on budgeting, saving, debt, investing and similar basics. Its site says there are 1,200+ lessons across nine categories, and you earn points as you learn that you can redeem for rewards in an in-app Marketplace. It’s often offered through banks and credit unions as part of their financial education, so you may find a sponsored version through an institution in your area.

It’s a solid foundation. For anything beyond the basics, like investing strategy or taxes in your country, you’ll need sources with more depth.

For sign language: Lingvano

Duolingo doesn’t teach American Sign Language, which is why “apps like Duolingo for ASL” shows up so often. Lingvano is purpose-built for it. Its lessons take about 10 minutes a day, and its site describes 600+ interactive lessons taught by native Deaf teachers, with quizzes, sign recognition feedback for fingerspelling, and real-life dialogues. Learning from real signers matters a lot for a visual language, and it’s the main reason to choose a dedicated app here.

For music

Start with Duolingo Music if you want to read music and learn simple songs in the same five-minute format. For a real instrument, look for an app that listens to you play through the microphone and responds. That’s the music version of “you produce something every few seconds”, and it’s what separates a learning app from a library of videos.

For anatomy and medical terms

“Duolingo for anatomy” is a surprisingly common search, mostly from students. There isn’t a polished Duolingo-style anatomy app that clearly leads, but this is where Anki shines. Medical students have used it for years, and there are large shared decks for anatomy and other medical subjects that you can download instead of building cards from scratch. Pair a deck with a good anatomy atlas or a 3D anatomy app for the visual side, and do your Anki reviews daily. That gives you small sessions, constant recall and a reason to come back tomorrow, which is everything that makes Duolingo work.

Anki itself is free on desktop and Android. The official iOS app, AnkiMobile, is paid, and those purchases help fund Anki’s development.

For history, psychology and general knowledge

This is the most crowded category, and the right pick depends on how you like to take things in.

Kinnu offers free general-knowledge pathways built around memory science, with regular review of what you’ve covered. It’s the strongest pure “Duolingo for knowledge” option on the free side.

Imprint teaches big ideas in psychology, philosophy and science through illustrated tap-through lessons. It’s for people who learn from visuals.

Nibble covers 35+ topics through 670+ short interactive pieces of 3 to 10 minutes, with quizzes, games, video, audio and chats with historical figures and AI tutors. It’s broad and playful, available on iOS, Android and the web, and the first lesson is free.

NerdSip uses AI to generate five-minute courses with quizzes on almost any topic you type in, with heavy gamification. As with any AI-generated content, check facts you plan to rely on.

For self-improvement and career skills

Here the Duolingo model needs one change. In a language, practice is the goal. In self-improvement, practice is a means; the goal is doing something differently at work or at home. An app that quizzes you on habit research is still teaching you about habits, not building one.

The Headway app turns nonfiction books into 15-minute summaries with streaks, spaced repetition and short themed plans. It’s the most Duolingo-like of the book-summary apps.

Podex is built around that change from knowing to doing, and since my brother and I made it, take this with a pinch of salt. It’s a micro-learning app built on the world’s best podcasts. You pick a goal, such as building something people want, getting your money in order or getting sharper every day, and you get one short lesson a day with the exact step to run in real life, then a check-in the next day to see whether it stuck. The library has 200+ summaries from shows like Diary of a CEO, My First Million, Acquired and Huberman Lab. Instead of a quiz, the step you take tomorrow is the exercise. You can also paste any YouTube or podcast link to turn an episode into tactics. The free tier lets you browse everything and start the first chapter of every daily track, no card required: getpodex.com/onboarding

Nibble also has a self-development angle, and Imprint covers communication and mindset topics visually.

For anything else: build the Duolingo loop yourself

Wine regions, a professional exam, a new job’s product catalogue, the names of everyone on your team: no app covers all of it. But the three mechanics are easy to rebuild with free tools.

1. Make it small. Break the subject into pieces you can learn in five minutes: ten terms, one diagram, one process.
2. Make yourself produce. Put each piece into Anki as questions rather than notes. “What does the deltoid do?” beats “Deltoid: shoulder abduction”. Anki handles the spaced review automatically.
3. Make tomorrow ask for you. Pick a fixed trigger, such as your first coffee. Keep a streak on paper or in a habit tracker. The streak isn’t the point; it’s a reminder that tomorrow is part of the chain.

For skills rather than facts, change step 2: instead of a flashcard, write one thing you’ll practise tomorrow and check it off the next evening.

The trap to avoid

Duolingo’s own success comes with a well-known side effect: the streak can become the goal. Plenty of people keep a 500-day streak doing one easy lesson a night and plateau for months.

The same thing happens in every Duolingo-style app. So every month, ask one question: can I do something now that I couldn’t do a month ago? Explain a concept, write a working function, sign a full sentence, handle a money decision with more confidence. If the answer is no, make the lessons harder or switch to a different mechanic. Don’t just protect the streak.

A quick way to choose

Coding: Sololearn.

Maths, science and logic: Brilliant, or Khan Academy for free. Duolingo Math for school-level maths.

Money basics: Zogo.

Anatomy and medical terms: Anki with a shared deck.

ASL: Lingvano.

Music, chess, or school maths in the Duolingo format: Duolingo’s own courses.

History, psychology, general knowledge: Kinnu, Imprint, Nibble or NerdSip.

Books and self-improvement ideas: Headway.

Acting on advice from podcasts: Podex.

Anything else: Anki plus a daily trigger.

Common questions

Is there a Duolingo for everything?

Not one app that does everything well. Duolingo itself has expanded to maths, music and chess. For other subjects, the best results come from a dedicated app for that subject, or from rebuilding Duolingo’s loop with Anki and a daily trigger.

What app is like Duolingo but for coding?

Sololearn is the closest: free, bite-sized interactive lessons with quizzes, where you write and run real code. Brilliant also teaches coding through interactive problems.

Is there a Duolingo for finance?

Zogo is the closest match, with gamified lessons and quizzes on budgeting, saving, debt and investing, and points you can redeem for rewards.

Is there a Duolingo for ASL?

Duolingo doesn’t offer ASL. Lingvano is purpose-built for it, with lessons of about 10 minutes a day and dialogues with Deaf teachers.

Are there free apps like Duolingo for other subjects?

Yes. Khan Academy is free, Kinnu has free knowledge pathways, Sololearn’s lessons are free, and Duolingo’s Math, Music and Chess courses are free to start. Anki is free on desktop and Android.

The short version

The thing you like about Duolingo is three mechanics: small lessons, constant doing and a reason to come back tomorrow. Find an app that does all three for your subject, or build them yourself with a flashcard app and a daily trigger. Then check once a month that you’re getting better at the subject, not just at keeping the streak.

The Headway app is well made. It turns nonfiction bestsellers into 15-minute summaries you can read or listen to, adds daily insights and streaks, and wraps it in short themed plans with names like "Speak like a CEO". Its own site says the library holds 1,700+ summaries, and it uses spaced repetition to bring key ideas back. For a lot of people that's exactly the right amount of structure.

So when someone searches for Headway app alternatives, it's rarely because Headway is bad. It's because one specific thing stopped working. Maybe the summaries feel thin for the books you care about. Maybe the library doesn't have the titles you want. Maybe the renewal price showed up and felt like a lot for something you open twice a week. Or maybe you've finished forty summaries and can't name three ideas you actually use.

Each of those has a different best answer, so that's how this list is organised. Product facts come from each company's own site as of October 2026. I've left out prices I couldn't confirm on an official page, because summary-app pricing changes by country, promotion and signup path.

Full disclosure: I co-founded Podex with my brother. It appears in one section below, for one specific reason, and I've tried to keep it in proportion.

First, a 10-minute test before you switch

Most people switch apps and repeat the same problem in a new interface. Before you install anything, answer three questions honestly:

1. What did you open Headway for in the first place? Book discovery, a daily habit, deep understanding of a few books, or changing how you work and live?
2. Which of those is it failing at? Pick one.
3. In the last month, how many ideas from the app did you actually try?

If the answer to question 3 is zero, read the last section first. Changing apps won't fix that by itself.

If the summaries feel too shallow: Shortform

Shortform sits at the opposite end from 15-minute summaries. Its guides go chapter by chapter, add commentary and context from other books, and include exercises. The pricing page lists highlights and notes, audio narrations, PDF downloads and community discussions, with iOS and Android apps.

Price on shortform.com as of October 2026: $16.42 a month billed annually.

Choose it if you read a handful of important books a year and want to understand them properly, or if you often finish a Headway summary thinking "but why?" Skip it if you want something you can finish while waiting for coffee. A Shortform guide is a reading session, not a snack.

A good way to use it: read the Shortform guide for the one book you're most likely to apply this quarter, do the exercises, and skip everything else. Depth on one book beats breadth on twenty.

If the library feels too small: Blinkist

Blinkist is the biggest name in book summaries and has the largest catalog of the mainstream apps. Its plans page lists 9,000+ book and podcast summaries, readable or listenable in about 15 minutes, with personal daily recommendations. Premium also includes a second Premium account to share with someone.

Price on blinkist.com/en/plans as of October 2026 (US): Premium is $79.99 for the first year, then $99.99 a year. A Pro plan adds Blinkist AI for summarising videos, podcasts, documents and articles, at $139.99 for the first year, then $174.99 a year.

Free: Blinkist's site says that if you cancel a trial, you can still read or listen to one pre-selected book a day.

Choose it if Headway kept failing to have the book you wanted, or if you share a subscription with a partner. It feels less like a guided habit app than Headway, which some people love and others find makes it easier to drift.

If you're tired of books entirely: Deepstash, Imprint, Nibble or Kinnu

Sometimes the problem isn't the app. It's the format. A book summary is still a book, compressed. If you want ideas without the book wrapper, these are the strongest options.

Deepstash collects bite-sized ideas from books, articles, podcasts and videos and lets you save them into your own library. Its site claims access to 200,000+ ideas, a personalised growth plan, listening as well as reading, and a feature that scans pages from physical books. The paid tier removes ads and adds offline access. It's the closest to scrolling, in a good way, which makes it a natural swap for a social feed.

Imprint teaches through illustrated, tap-through visual lessons on psychology and other big topics. Its site emphasises visuals that clarify key ideas, and saving and reviewing the ideas you want to keep. If text summaries slide off your brain but diagrams stick, try it.

Nibble covers 35+ topics through 670+ interactive pieces, with lessons that take 3 to 10 minutes and include quizzes, games, video, audio and chats with historical figures and AI tutors. It's on iOS, Android and the web, and the first lesson is free. It's broader and more playful than Headway, less about self-improvement and more about curiosity.

Kinnu offers free general-knowledge "pathways" built around memory and review. If cost was your reason for leaving and you're happy to learn subjects rather than books, it's the obvious first install.

If the price is the problem: the genuinely free routes

"Apps like Headway but free" is one of the most common versions of this search, so it's worth being precise about what free means. Many apps call a 7-day trial free. These keep working without paying:

Blinkist's free daily pick, one pre-selected title a day.

StoryShots, which offers free book summaries in text, audio and animated formats, with paid extras on top.

Kinnu's free pathways.

Your local library, through Libby. With a library card you can borrow whole ebooks and audiobooks for free. If you liked Headway mostly for audio on the commute, this is the most underrated option on the list. You get the full book, and what's available depends on your library's catalog, so check it before you assume a title is missing.

A sensible free stack: use a free daily summary to decide which books deserve your time, then borrow the winners through Libby. You'll read fewer books and get more out of each one.

If nothing is sticking: change what you do after the summary

This is the reason people rarely admit to, and it's the most common one. You read a summary of Atomic Habits on Monday, nod along, and by Friday it has merged into a general sense that habits matter.

That's not a Headway problem. It's how memory works. Reading a summary gives you recognition ("oh right, I've heard that"), not the ability to use the idea. Spaced repetition helps you recall the idea. It doesn't make you apply it.

A method that works with any app, including Headway:

1. After each summary, write one sentence that starts with "Tomorrow I will". Make it something you can do in under ten minutes.
2. Attach it to a moment you can't miss: after your first coffee, or when you open your laptop.
3. The next evening, answer one question in one line: did I do it, and what happened?
4. Only move to a new summary once you've run the last idea at least once.

That last rule will cut how many summaries you consume by a lot. It will also make the ones you do read count.

Where podcasts fit: Podex

There's one more reason people leave book-summary apps that doesn't get discussed much. A lot of the most practical advice now comes from long-form podcast interviews rather than books. A founder explaining exactly how they priced their first product on My First Million, or a researcher on Huberman Lab walking through a specific morning routine, is often more actionable than a summary of a book written five years earlier.

Podex is built on that idea, and we built it, so read this with that in mind. It's a micro-learning app built on the world's best podcasts. You pick a goal, such as building something people want, getting your money in order, or getting sharper every day, and you get one short lesson a day with the exact step to run, then a check-in the next day to make it stick. The library has 200+ summaries from shows like Diary of a CEO, Acquired, The Tim Ferriss Show and Huberman Lab, and every tactic traces back to something a guest actually said.

It's a different source rather than a Headway clone: podcasts instead of books, and built around doing the step rather than recalling the idea. The free tier lets you browse the catalog, see the top tactic of every summary, open three full summaries you choose, and start the first chapter of every daily track, no card required. If the "Tomorrow I will" method above sounds useful but you'd like it done for you, you can pick a goal and get your first lesson here: https://www.getpodex.com/onboarding?utm_source=medium&utm_medium=article&utm_campaign=headway-app-alternatives

A quick comparison by reason

Too shallow: Shortform. Longer guides with commentary and exercises.

Too small a library: Blinkist. 9,000+ summaries, plus a free daily title.

Tired of books: Deepstash for idea cards, Imprint for visual lessons, Nibble for broad curiosity, Kinnu for free structured pathways.

Too expensive: Blinkist's free daily title, StoryShots' free summaries, Kinnu, and full books through Libby.

Nothing sticks: keep any app and add the one-action rule. Or try a format built around a daily step and a check-in.

Want ideas from podcasts rather than books: Podex.

When to stay with Headway

If what you value is a guided daily habit with streaks, short plans and a polished listening experience, Headway does that well, and switching to a bigger library like Blinkist may actually make you open the app less. The structure is the product. Plenty of people would do better adding the one-action rule to Headway than starting over somewhere else.

Common questions

What's the best free alternative to the Headway app?

For book summaries, Blinkist's free daily title and StoryShots' free summaries. For learning without books, Kinnu's free pathways. For full books, Libby with a library card. Which is best depends on whether you want summaries, subjects or the real book.

What is the closest app to Headway?

Blinkist is the closest in format: 15-minute book summaries you can read or listen to, with a much larger library. Headway leans harder into habit features like streaks and plans.

Is Shortform better than Headway?

For depth, yes. Shortform guides are far more detailed, with commentary and exercises. For a quick daily habit, Headway is easier to stick with. They suit different people more than one is simply better.

Are there apps like Headway for podcasts?

Yes. Blinkist includes some podcast summaries in its library, and Podex is built entirely on long-form podcast episodes, turning them into daily lessons with a step to try. Snipd is a podcast player that helps you save and summarise moments while you listen.

The short version

Figure out which part of Headway stopped working, then pick for that. Shallow summaries point to Shortform. A small library points to Blinkist. If you're tired of books, try Deepstash, Imprint, Nibble or Kinnu. If cost is the issue, the free daily summaries plus a library card go a long way. And if the real problem is that nothing sticks, change what you do in the 24 hours after you read, not which app you read in.