Nice work! I've shared a workshop and presentation called "Discovering Useful Ideas" or "Discovering Good Ideas" for years and this makes me think of it:
Using algorithms we create ourselves to help with the serendipitous (yet intentional) discovery of great content is not only fun, I think it's valuable from civics and pedagogical perspectives.
I look forward to learning more about how you created https://plot.fyi !
This is incredible. Great work. It gives me feelings of how the internet was when it was people with their own little corner on the web, making their own cool things just because they could. Thank you so much for sharing.
This is wild and is making my head spin, and cant say too much after clicking around. So all that tagging was you in an hour on a lunch break? Or are a lot of tags from the data.
I got confused where it told me to "click explain" as I searched just on "THe Road" and never saw buttons. I was curious about the connection with Life of Pi, but managed ti figure it out with the command line /explain The Road / Life of Pi
The total tagging is a multi month project, but you can see in some of the videos me running one or two of the tags in 10 minutes or so. So it’s an accretion of small realizations like “we should have a tag about creepy brother sister relationship as a plot point” and then having the LLM run that against my vault. The initial vault entries are wikidata that are then enriched by omdb then by my custom tag sets, where I add a bit to it each night.
Well that makes sense, because I think vibe building is bullshit! And I don't really have years of appreciation of film; until about a year ago I barely watched films at all. For at least 15 years I fell into the whole "Let's watch the new prestige TV" thing and it was only because of a series of life events I got interested in film. I always watched good stuff, and for two semesters of college I thought I might be a film major because it seemed pretty fun. But that was it really.
I think the problem with vibe building is that it assumes that the thing you're trying to do is build a solution in a day and say See! Look at all the work I saved! In the end, who cares. What I'm trying to demonstrate is something different that we can give our students. Take a problem, take the LLM, work it a bit each day. In each iteration learn something about the domain. In the end the most valuable thing you walk away with is not the code, but the understanding. If you grasp this you can get smarter about things you choose to get smart about at amazingly fast rates. But it's the understand, build, fail, rethink, rebuild cycle that is the heart of the humane use of these tools and people keep missing it.
I think in some of my later stuff I express this better. I'm trying to straddle the line between, for god's sake just start trying to do the thing instead of talking about it, and stressing that the real value will be engagement over time and I might be failing to do that.
As it has always been it takes me a bit longer to catch up to your thinking, Mike.
I see where you are aiming but also feel doubt how many are willing to take this on when all the pressure is do more in less time which we know is asinine.
It’s not your film experience I was thinking about it’s your deep understanding of knowledge organization, how to think about a problem or area of interest with they way you are doing by tagging.
I guess you are wanting students to have this experience through doing their own version of what you did.
That's fair. I've thought for ages about definitions, boundaries, themes, etc. And of course tagging. Definitions are the foundation of so much important analytical work but they don't come naturally to people.
I suppose though that that is also a piece of why I am passionate about this, even if we scale it down. To the extent people understand and surface what their definitions of a thing they ask for with these tools is (does vaccine injury mean reported complications in hospitals, things that happened about the same time as the vaccine, or just someone reporting something into an unmoderated form online?) they can get really good answers. To the extent they don't they will likely find a lot of answers that just reinforce their bias. So the challenge is high, but I think the target is really central to what it means to live in a society with this analytical power at your fingertips, and maybe(?) that means we can spend more than average effort in hammering out some activities.
Nice work! I've shared a workshop and presentation called "Discovering Useful Ideas" or "Discovering Good Ideas" for years and this makes me think of it:
https://wiki.wesfryer.com/Home/handouts/ideas
Using algorithms we create ourselves to help with the serendipitous (yet intentional) discovery of great content is not only fun, I think it's valuable from civics and pedagogical perspectives.
I look forward to learning more about how you created https://plot.fyi !
This is incredible. Great work. It gives me feelings of how the internet was when it was people with their own little corner on the web, making their own cool things just because they could. Thank you so much for sharing.
This is wild and is making my head spin, and cant say too much after clicking around. So all that tagging was you in an hour on a lunch break? Or are a lot of tags from the data.
I got confused where it told me to "click explain" as I searched just on "THe Road" and never saw buttons. I was curious about the connection with Life of Pi, but managed ti figure it out with the command line /explain The Road / Life of Pi
https://plot.fyi/#c/%3Dexplain%20The%20Road%20%2F%20Life%20of%20Pi&v=card
Thanks for today's distraction. Also, say hi to Todd for me, he likes chatting with you ;-)
The total tagging is a multi month project, but you can see in some of the videos me running one or two of the tags in 10 minutes or so. So it’s an accretion of small realizations like “we should have a tag about creepy brother sister relationship as a plot point” and then having the LLM run that against my vault. The initial vault entries are wikidata that are then enriched by omdb then by my custom tag sets, where I add a bit to it each night.
Ok. But riddle me this. You are able to so cleverly wrangle this app based on years of experience and appreciation of film as a foundation.
But how does a future generation Mike ever develop this working knowledge if the bulk of their experience is LM output consumption?
What you have been building these past years is impressive use cases, but I think you are the wrong case example for vibe building.
Well that makes sense, because I think vibe building is bullshit! And I don't really have years of appreciation of film; until about a year ago I barely watched films at all. For at least 15 years I fell into the whole "Let's watch the new prestige TV" thing and it was only because of a series of life events I got interested in film. I always watched good stuff, and for two semesters of college I thought I might be a film major because it seemed pretty fun. But that was it really.
I think the problem with vibe building is that it assumes that the thing you're trying to do is build a solution in a day and say See! Look at all the work I saved! In the end, who cares. What I'm trying to demonstrate is something different that we can give our students. Take a problem, take the LLM, work it a bit each day. In each iteration learn something about the domain. In the end the most valuable thing you walk away with is not the code, but the understanding. If you grasp this you can get smarter about things you choose to get smart about at amazingly fast rates. But it's the understand, build, fail, rethink, rebuild cycle that is the heart of the humane use of these tools and people keep missing it.
I think in some of my later stuff I express this better. I'm trying to straddle the line between, for god's sake just start trying to do the thing instead of talking about it, and stressing that the real value will be engagement over time and I might be failing to do that.
As it has always been it takes me a bit longer to catch up to your thinking, Mike.
I see where you are aiming but also feel doubt how many are willing to take this on when all the pressure is do more in less time which we know is asinine.
It’s not your film experience I was thinking about it’s your deep understanding of knowledge organization, how to think about a problem or area of interest with they way you are doing by tagging.
I guess you are wanting students to have this experience through doing their own version of what you did.
That's fair. I've thought for ages about definitions, boundaries, themes, etc. And of course tagging. Definitions are the foundation of so much important analytical work but they don't come naturally to people.
I suppose though that that is also a piece of why I am passionate about this, even if we scale it down. To the extent people understand and surface what their definitions of a thing they ask for with these tools is (does vaccine injury mean reported complications in hospitals, things that happened about the same time as the vaccine, or just someone reporting something into an unmoderated form online?) they can get really good answers. To the extent they don't they will likely find a lot of answers that just reinforce their bias. So the challenge is high, but I think the target is really central to what it means to live in a society with this analytical power at your fingertips, and maybe(?) that means we can spend more than average effort in hammering out some activities.
This project also reminds me of Freq, tho it's built as a small community platform (for music discovery)
https://publicinfrastructure.org/2023/05/24/introducing-freq/
Dang Freq has been decommissioned. I was excited to check it out!
Cool!