The superpower to me is the interactivity. You hint at this a bit with the generate tests and quizzes, but just the ability to ask questions and clarify in real time is amazing and much more like having a tutor than not.
Agreed! I actually showcase some of my favorite use cases of NotebookLM's "Interactive Mode" feature where you can interrupt the podcast hosts in my companion video guide.
Very impressive and useful collection of resources in a logical and thoughtful manner of presentation. Cheers! Need to try out the tools at the start of the post that I have never used but seem promising!
checking out rabbit hole this morning and it is pretty cool! is the whole idea that you can go done the various potential follow-ups and trace back up when you need to?
do you have any notable uses where rabbit hole has really helped you explore something better? i already see value in this over these deep research tools which really does not provide human user input in the middle of a research run, which this does by mapping the potential possibilities to take the discussion.
is the paid plan worth it or are you just using the free version?
Yeah pretty much: You get to go on an exploration (wherever it takes you) and in the process you end up building a sort of visual mind map of the topic, since each branch is a new path. So you can then zoom out to get a holistic picture of the exploration.
Very different from Deep Research, which is an autonomous agent that's meant to go beyond what you can do in a short span of time - this is more of a "learn at your own pace and build your own path" approach.
I wouldn't say the paid plan is worth it at all at this stage, unless you're a REALLY heavy user. You already get 3 separate rabbit holes per day as a free user (with unlimited branches/follow-ups per rabbit hole). Looks like they'll add new features like better LLMs and the ability to chat to your rabbit holes later, at which point it might make sense to upgrade. But for now, the free plan is plenty!
If you don't like any of the three suggested follow-up questions, you simply hover over the main box (from which they branch out) and scroll to the bottom where it says "Ask Follow Up" - then you can type your own question to be added as the fourth branch.
That's yet another thing I showcase in the video guide ;)
OK, I feel like this is where I can actually weigh in a little bit and add some value, maybe. I do this every day, and I love it so much!
I've definitely used Pareto-style prompts (#1) and #2 resonates - I'll find a good book if I'm interested in a deep dive, or maybe even a good TV show to watch.
I need to save this one:
“Summarize key themes and insights in the attached sources. Identify any inconsistencies or knowledge areas that aren’t sufficiently covered. Provide potential research questions that might help address these knowledge gaps.”
Lately I've been trying to learn the nations of the world, just where everything is on a map. I can do most of the big nations now, but the smaller ones are tough. I play a little quiz game where I try to identify all nations with a UN subregion (there are 22 of them), and the idea is for a voice prompt to give me the first two letters of the nation, see if I can name it. I get the idea of the group that way, even without a map.
Also on voice: I am going to be practicing more Brazilian Portuguese soon too. So far, it's not awful, but voice can't go slower, so it's a slog right now.
I've found myself using voice mainly in the evening, and desktop GPT or Gemini mostly during the day.
I like that they talk slower. It also seems rock solid so far. It's def not ideal, though. Pretty glitchy at my end - like the voice keeps talking and doesn't recognize interruptions as well as other models I've tried, but I can very much see what you mean about the natural feel. This has tons of potential.
That's odd, because I was specifically impressed at just how instant their interruption-detection and general latency was. Maybe some temporary issue and/or Internet bandwidth issue. Try again later and see if they do better.
I also noticed that they seem to pick up on nonverbal cues like my tone of voice and tailor their responses accordingly. Or maybe I'm projecting that. In any case, we're getting really close to near-human-level chats.
Well also: i use headphones and my phone to talk to ChatGPT and Gemini, whereas this was on my laptop just now. I wonder if that matters - I don't take anything for granted any more, and figure it all might matter.
If you know of a way for me to like go for a walk and play with this thing, I'll spend a couple hours on it later today.
I guess you can always run it in your phone browser, but it's essentially a preview demo so it's not like they have a polished app yet. I do know that sometimes Advanced Voice had issues when I e.g. used Bluetooth headphones or had my VPN switched on, so lots of things can interfere.
I was thinking the same thing: my experience may not be similar to yours at all, and vice versa w/ChatGPT's voice. I swear, it's like magic most of the time, and it's far more reliable than any other service I've tried so far! But I'm here in the US, I'm using a particular cell network and a particular device and so on, and I think there are other notable differences we may not be picking up on.
Also, our objectives when hopping on could be a little different, but I think it's mostly the former in this case. I may well try them out for my voice convos this evening. I think it has become a part of my routine to play with these things for at least a few minutes at the end of the day.
The superpower to me is the interactivity. You hint at this a bit with the generate tests and quizzes, but just the ability to ask questions and clarify in real time is amazing and much more like having a tutor than not.
Agreed! I actually showcase some of my favorite use cases of NotebookLM's "Interactive Mode" feature where you can interrupt the podcast hosts in my companion video guide.
Very impressive and useful collection of resources in a logical and thoughtful manner of presentation. Cheers! Need to try out the tools at the start of the post that I have never used but seem promising!
Happy you found the list useful - would love to hear your thoughts on the tools when you end up trying them out.
checking out rabbit hole this morning and it is pretty cool! is the whole idea that you can go done the various potential follow-ups and trace back up when you need to?
https://www.rabbithole.chat/rabbithole/im-researching-for-a-jq4pvg4p
do you have any notable uses where rabbit hole has really helped you explore something better? i already see value in this over these deep research tools which really does not provide human user input in the middle of a research run, which this does by mapping the potential possibilities to take the discussion.
is the paid plan worth it or are you just using the free version?
Yeah pretty much: You get to go on an exploration (wherever it takes you) and in the process you end up building a sort of visual mind map of the topic, since each branch is a new path. So you can then zoom out to get a holistic picture of the exploration.
Very different from Deep Research, which is an autonomous agent that's meant to go beyond what you can do in a short span of time - this is more of a "learn at your own pace and build your own path" approach.
I wouldn't say the paid plan is worth it at all at this stage, unless you're a REALLY heavy user. You already get 3 separate rabbit holes per day as a free user (with unlimited branches/follow-ups per rabbit hole). Looks like they'll add new features like better LLMs and the ability to chat to your rabbit holes later, at which point it might make sense to upgrade. But for now, the free plan is plenty!
PS: If you enjoy rabbithole, you might like Heuristica or Albus from this list: https://www.whytryai.com/p/free-ai-learning-tools
Same vibes!
cheers!
i agree the 3 searches per day is more than enough and am looking forward to how they evolve!
it seems like i can't ask a question within the tree which is a limitation but not a big deal given the use case is to explore a new topic.
If you don't like any of the three suggested follow-up questions, you simply hover over the main box (from which they branch out) and scroll to the bottom where it says "Ask Follow Up" - then you can type your own question to be added as the fourth branch.
That's yet another thing I showcase in the video guide ;)
OK, I feel like this is where I can actually weigh in a little bit and add some value, maybe. I do this every day, and I love it so much!
I've definitely used Pareto-style prompts (#1) and #2 resonates - I'll find a good book if I'm interested in a deep dive, or maybe even a good TV show to watch.
I need to save this one:
“Summarize key themes and insights in the attached sources. Identify any inconsistencies or knowledge areas that aren’t sufficiently covered. Provide potential research questions that might help address these knowledge gaps.”
Lately I've been trying to learn the nations of the world, just where everything is on a map. I can do most of the big nations now, but the smaller ones are tough. I play a little quiz game where I try to identify all nations with a UN subregion (there are 22 of them), and the idea is for a voice prompt to give me the first two letters of the nation, see if I can name it. I get the idea of the group that way, even without a map.
Also on voice: I am going to be practicing more Brazilian Portuguese soon too. So far, it's not awful, but voice can't go slower, so it's a slog right now.
I've found myself using voice mainly in the evening, and desktop GPT or Gemini mostly during the day.
Thanks for your insights - looks like we're aligned on the approaches here!
As for voice, if you haven't already, check out Sesame: https://www.sesame.com/research/crossing_the_uncanny_valley_of_voice#demo
Incredibly realistic, real-time voice chat. More impressive than the current iteration of the Advanced Voice chat in ChatGPT I think.
I like that they talk slower. It also seems rock solid so far. It's def not ideal, though. Pretty glitchy at my end - like the voice keeps talking and doesn't recognize interruptions as well as other models I've tried, but I can very much see what you mean about the natural feel. This has tons of potential.
That's odd, because I was specifically impressed at just how instant their interruption-detection and general latency was. Maybe some temporary issue and/or Internet bandwidth issue. Try again later and see if they do better.
I also noticed that they seem to pick up on nonverbal cues like my tone of voice and tailor their responses accordingly. Or maybe I'm projecting that. In any case, we're getting really close to near-human-level chats.
Well also: i use headphones and my phone to talk to ChatGPT and Gemini, whereas this was on my laptop just now. I wonder if that matters - I don't take anything for granted any more, and figure it all might matter.
If you know of a way for me to like go for a walk and play with this thing, I'll spend a couple hours on it later today.
I guess you can always run it in your phone browser, but it's essentially a preview demo so it's not like they have a polished app yet. I do know that sometimes Advanced Voice had issues when I e.g. used Bluetooth headphones or had my VPN switched on, so lots of things can interfere.
I was thinking the same thing: my experience may not be similar to yours at all, and vice versa w/ChatGPT's voice. I swear, it's like magic most of the time, and it's far more reliable than any other service I've tried so far! But I'm here in the US, I'm using a particular cell network and a particular device and so on, and I think there are other notable differences we may not be picking up on.
Also, our objectives when hopping on could be a little different, but I think it's mostly the former in this case. I may well try them out for my voice convos this evening. I think it has become a part of my routine to play with these things for at least a few minutes at the end of the day.