super helpful to see how Dwarkesh preps for his interviews, imbibes knowledge and uses Claude and other tools to make the process much more efficient. Now I want to go back to my book shelf, create a bunch of Claude projects and then re-grok them. I already have started doing something similar with Iain M Banks Culture novels and also built a Startup Mentor using books of Ben Horowitz, Thiel etc. I think there is a lot of value in mixing up books from different genres, finding threads etc. For example I have tried to find similarities between Balaji S's Network State and David Graeber's Dawn of Everything..as they both talk about decentralisation, one in the future and the other in the past.
Extremely insightful conversation and so many impactful ways to use AI models to enhance our thinking and planning. Especially loved showcasing Mochi and stressing the importance of spaced repetition!
I saw you mention on Twitter, Dan, that you were writing a blog post about how you're using Claude when reading books. Is that article published yet, or are you still working on it?
It's so interesting how spaced repitition is similar to a fine tune dataset for our brains lol. But makes me wonder if we could teach llms with spaced repitition also?
Very interesting to see how Dwarkesh uses AI to help prepare for interviews. I'm not sure if the on-the-fly prompting parts are the best suited for a podcast though. Some of that can be interesting but can also start to feel more like a livestream.
Great conversation! If anyone has the full prompt Dwarkesh uses to create the spaced repetitions, i would love to see it. We only got to see part of it
Generate spaced repetition prompts from the text at the very end using the following guidance: Spaced repetition prompts should be focused. A question or answer involving too much detail will dull your concentration and stimulate incomplete retrievals, leaving some bulbs unlit. Unfocused questions also make it harder to check whether you remembered all parts of the answer and to note places where you differed. It’s usually best to focus on one detail at a time. Spaced repetition prompts should be precise about what they’re asking for. Vague questions will elicit vague answers, which won’t reliably light the bulbs you’re targeting. Spaced repetition prompts should produce consistent answers, lighting the same bulbs each time you perform the task. Otherwise, you may run afoul of an interference phenomenon called "retrieval-induced forgetting": what you remember during practice is reinforced, but other related knowledge which you didn't recall is actually inhibited. Now, there is a useful type of prompt which involves generating new answers with each repetition, but such prompts leverage a different theory of change. We'll discuss them briefly later in this guide. Spaced repetition prompts should be tractable. To avoid interference-driven churn and recurring annoyance in your review sessions, you should strive to write prompts which you can almost always answer correctly. This often means breaking the task down, or adding cues. Spaced repetition prompts should be effortful. It's important that the prompt actually involves retrieving the answer from memory. You shouldn't be able to trivially infer the answer. Cues are helpful, as we'll discuss later-just don't "give the answer away." In fact, effort appears to be an important factor in the effects of retrieval practice. That's one motivation for spacing reviews out over time: if it's too easy to recall the answer, retrieval practice has little effect. Output question answer pairs like so: Why did blah blah blah? --- Because blah blah blah. How big was xyz? --- abc.
Problem with key points extraction is like, why even bother with writing full articles in the first place? And you can always extract key points from "key points" text indefinetly with ai tools. Who's even going to bother to write stuff themselves?
I want you to read this section of my book, and break it down into interesting and logically progressing concepts. For each concept, given all the quotes you know, create a unique, original, powerful, inspiring and insightful quote which encapsulates the concept.
You can also modify with : For each concept, given all the quotes you know by Mother Teresa, create... ... which encapsulates the concept in the same voice.
![The moment we stopped understanding AI [AlexNet]](/img/1.gif)
Dwarkesh is like an insightful version of Lex Fridman who doesn't have to try to make his observations and questions sound deep.
Dwarkesh also actually understands the subject matter and respects his audience. Dwarkesh and Lex are not even in the same league
Dwarkesh is the man!
super helpful to see how Dwarkesh preps for his interviews, imbibes knowledge and uses Claude and other tools to make the process much more efficient. Now I want to go back to my book shelf, create a bunch of Claude projects and then re-grok them. I already have started doing something similar with Iain M Banks Culture novels and also built a Startup Mentor using books of Ben Horowitz, Thiel etc. I think there is a lot of value in mixing up books from different genres, finding threads etc. For example I have tried to find similarities between Balaji S's Network State and David Graeber's Dawn of Everything..as they both talk about decentralisation, one in the future and the other in the past.
Great convo between you both, got a lot of value out of it!
Extremely insightful conversation and so many impactful ways to use AI models to enhance our thinking and planning. Especially loved showcasing Mochi and stressing the importance of spaced repetition!
Another awesome episode! AI&I has rapidly become my favorite podcast/RUclips video series - keep it up
Two absolute GIGACHADS! Thanks for this!
Stay curious Dwarkesh, love your energy 🙏👍
I saw you mention on Twitter, Dan, that you were writing a blog post about how you're using Claude when reading books. Is that article published yet, or are you still working on it?
Comment for the algorithm,. Great content. The quest to learn everything could be a good phrase that can describe the modern human
being
It's so interesting how spaced repitition is similar to a fine tune dataset for our brains lol. But makes me wonder if we could teach llms with spaced repitition also?
This was really cool and helpful, thanks👍
Dwarkesh Superchad Patel
I highly recommend reading "Mental Models I Find Repeatedly Useful" by Gabriel Weinberg
Can Dwarkesh make his huggingface space public?
Very interesting to see how Dwarkesh uses AI to help prepare for interviews. I'm not sure if the on-the-fly prompting parts are the best suited for a podcast though. Some of that can be interesting but can also start to feel more like a livestream.
Great video, thanks. What is the online service Dwarkesh uses for converting an epub file to text ?
What is the Card Dashboard Patel using?
this must have been pre the latest version of NotebookLM
Never start with an ad!!
Interestingly, I tried to do this with the book Atomic Habits (Spanish edition) that I’m currently reading and it filled my project completely
Any one has Dwarkesh spaced repetition prompt as huggingface link doesn't work anymore?
Great conversation! If anyone has the full prompt Dwarkesh uses to create the spaced repetitions, i would love to see it. We only got to see part of it
Generate spaced repetition prompts from the text at the very end using the following guidance:
Spaced repetition prompts should be focused. A question or answer involving too much detail will dull your concentration and stimulate incomplete retrievals, leaving some bulbs unlit. Unfocused questions also make it harder to check whether you remembered all parts of the answer and to note places where you differed. It’s usually best to focus on one detail at a time.
Spaced repetition prompts should be precise about what they’re asking for. Vague questions will elicit vague answers, which won’t reliably light the bulbs you’re targeting.
Spaced repetition prompts should produce consistent answers, lighting the same bulbs each time you perform the task. Otherwise, you may run afoul of an interference phenomenon called "retrieval-induced forgetting": what you remember during practice is reinforced, but other related knowledge which you didn't recall is actually inhibited. Now, there is a useful type of prompt which involves generating new answers with each repetition, but such prompts leverage a different theory of change. We'll discuss them briefly later in this guide.
Spaced repetition prompts should be tractable. To avoid interference-driven churn and recurring annoyance in your review sessions, you should strive to write prompts which you can almost always answer correctly. This often means breaking the task down, or adding cues.
Spaced repetition prompts should be effortful. It's important that the prompt actually involves retrieving the answer from memory. You shouldn't be able to trivially infer the answer. Cues are helpful, as we'll discuss later-just don't "give the answer away." In fact, effort appears to be an important factor in the effects of retrieval practice. That's one motivation for spacing reviews out over time: if it's too easy to recall the answer, retrieval practice has little effect.
Output question answer pairs like so:
Why did blah blah blah?
---
Because blah blah blah.
How big was xyz?
---
abc.
@@manuelmao4700 nice one. I fed Andy matyushaks article about spaced repetition into Claude and had it generate a system prompt for me.
Doesn't Descript do a lot of the podcast operations functionality? Or are there still quite a few things missing?
The author mentioned is Peter Watson, not Jackson fyi
Oh this is about to be the one
great ep!! this is annoying feedback but can you get the microwave to stop blinking?
Interesting
I did now know I could upload EPUBs to Claude/CGPT ... Damn
Problem with key points extraction is like, why even bother with writing full articles in the first place? And you can always extract key points from "key points" text indefinetly with ai tools.
Who's even going to bother to write stuff themselves?
I want you to read this section of my book, and break it down into interesting and logically progressing concepts. For each concept, given all the quotes you know, create a unique, original, powerful, inspiring and insightful quote which encapsulates the concept.
Cool prompt.
You can also modify with : For each concept, given all the quotes you know by Mother Teresa, create...
... which encapsulates the concept in the same voice.
Wait you have a white background and approximately 9 books in the background too?! let's do a podcast.