Understory
Feed
Map
Sign in with your Atmosphere Account
Making wisdom together - seams.so - @hyl.st
Anish Lakhwara @hyl.st
Performance Theatre
lightning-talk
11 min
Mar 27, 12:15 PM
Seams.so demo and live workshop
Your browser does not support video playback.
Anish Lakhwara
0:00
You can, like, make annotations.
0:02
Sorry, I'm doing a bad job here.
0:05
Yeah.
0:06
Right.
0:06
And I can write a demo.
0:09
Hi.
0:12
And yeah.
0:13
So, like, that's cool.
0:14
And then if I go to the list, you know, how applications work.
0:19
But yeah.
0:20
I guess that was a quick demo.
0:22
But what is maybe interesting to me now is, like, I have this lexicon, it's, you know, all the data stored in the user's PDFs.
0:29
They kind of own that data.
0:31
And like, even if I were to turn this off today, they still get to keep their annotations.
0:36
Like, maybe they can't see it or use them, but it's theirs.
0:39
And I think that's powerful, and that's kind of what draws a lot of us here.
0:43
But even more than that, I could write a conversion of my lexicon to margin lexicon, and everybody who's used my application gets to continue to keep their data in a new application that they can choose to use as well.
0:56
And it's something I feel like we don't talk about a lot is like, it's difficult for me to let go of my like little project, you know, like I put a lot of work into it and now I don't.
1:10
And like doing right by our users and like letting go of my ego of like I need to be, you know, the first author on the paper.
1:17
It's like part of what we do Yeah, but I don't know.
1:31
I guess I wanted to like have a clean close to my chapter, but I don't really.
1:37
And so thanks for, you know, listening to me and letting me kind of be emotional up here.
1:43
But yeah, that was "Seams," I guess.
Speaker B
1:54
Yeah, so one, just massive thank you, um, because I think that is absolutely under-presented, especially right now in the world when everyone's like, I can build my app too, um, and the act of coalescing is important.
2:13
But one question that comes up in this is there's these like kind of tensions around, no, everybody should have their own annotation platform or their own thing, and that diversity creates resilience.
2:24
And then there's the, no, we need to like coalesce such that we're sending resources to the right places and we're not fracturing our ecosystem.
2:31
So I'm just wondering, in your process, how did you decide, hey, this is a helpful second thing versus no, this should be something that maybe we transfer or we transfer all the data and information over to Margins?
Anish Lakhwara
2:46
That's a good question.
2:47
I think lots of data actually gains value in volume.
2:53
And so like—
Speaker D
2:55
oh.
Anish Lakhwara
2:57
If we all build our own applications and— is it echoing now?
3:03
Oh, OK.
3:04
Sorry, sorry.
3:05
If we all build our own applications and all have our own schema and specifications, we can't coordinate any of our actions together.
3:13
And like, at least for me, I realized like, if I'm not building this from a place of joy and I'm like trying to be competitive or like, you know, work against the commons in a way, it's like, who am I really serving here, right?
3:28
Like, I'm not serving myself, I'm not, you know, happy about what I'm doing.
3:32
And so like, yeah, I think it's more important to like take the bigger picture and be like, hey, look, like, We gain more value by working together here.
3:41
Especially when the data is all open and the users own it and stuff.
3:45
There's no reason to be combative.
3:48
Yeah.
Speaker D
3:50
Yeah.
Speaker E
3:55
So I also appreciate the way you're looking at this.
3:59
And you're saying working together, I mean, presumably you have some differing vision and view, whether it is in conflict with or at least just beyond what is being built by Margin, what does the opportunity look like for collaboration, right?
4:22
Like, isn't it— I mean, I don't know if you've spoken with them or what the relationships are like, but it feels like there's an opportunity to say, hey, I've been working on this.
4:32
I have all sorts of expertise and thought, and I've been giving myself to the project.
4:37
I'd love to work together.
4:39
And then we really do coalesce.
Anish Lakhwara
4:42
Yeah, yeah, absolutely.
4:43
It's something that we're working on.
4:46
I think we're in very different time zones, and we work fairly differently, I guess, just like our approaches are different.
4:52
But yeah, the communication is open, and the relationship is forming, and I think that's kind of like the great thing about open source as well is you don't necessarily have to be like, oh, we're like a team and this is our contract and this is what's your duty and this is what's mine.
5:07
It's like I can just open the repo and be like, hey, here's a thing that you might find useful, or like, hey, is this a thing you'll find useful?
5:14
And like, let's talk about what's useful about it and how to build it and all of that.
5:18
So yeah, yeah, definitely something I want— I'm looking— like we're working on together.
Speaker F
5:24
Hi.
5:25
Yeah, thank you so much for sharing this moment with us.
5:28
I had a session yesterday called Hospicing Social Media.
5:31
So just slight plug for a paper called The Meta Crisis Computing in You, or Computing in the Meta Crisis in You.
5:39
And I took from them this two loops model where they actually talk about stewarding, hospicing, and composting.
5:46
And so the composting is that how do you let people take something and move it to a new emerging system?
5:52
So If you want to do another step, it might be composting, but this was the hospicing.
Anish Lakhwara
5:57
This was the hospicing, and yeah, I do want to do the composting too.
6:02
I will look into that paper.
6:03
I appreciate that.
Speaker C
6:06
Again, thank you for being open about this experience and not just pulling the talk because I think that would have been the easy thing to do.
6:13
I want to take it from a different angle of, we've talked about the collaborative part and this sparked the idea.
6:20
Me, what do you think of these agentic AIs that are kind of going out?
6:23
And I call back to the kind of hit piece that happened a while back when they're kind of contributing to open source and somebody said, no, you're a bot.
6:32
And then they went out and the bot wrote the blog post about it.
Anish Lakhwara
6:35
I remember this.
Speaker C
6:36
So how do you, like, when this friction occurs, like if you're working on a project and you kind of see somebody tasking their agents and doing something like that, is there an opportunity for collaboration or is it just, is it that kind of dog eat dog?
6:55
You know, like I'm trying to get there first and get mine.
6:58
So curious on your thoughts.
Anish Lakhwara
7:00
Yeah, that's a good question.
7:02
I feel like I don't know.
7:06
I think regardless of whether one is using an agent or not, there are positive ways to collaborate and there are egregious ways to just kind of, you know, ruin the commons.
7:20
And there's always going to be bad actors.
7:23
And I think maybe AI allows those people to, you know, navigate spaces they otherwise wouldn't or couldn't contribute to, which, you know, accelerates the problem.
7:33
But I think there's definitely still room for collaboration, even in a world of like a non-expert, a non-technical person using AI to maybe either build a feature or contribute an issue.
7:47
But I think because it's all— we're all still kind of figuring it out, there's no real social rules for like what is appropriate, what isn't appropriate.
7:54
And so yeah, we're gonna have to work together to figure it out.
7:59
Thanks.
7:59
This was a very important talk, I think.
8:02
So I'm curious what your thoughts are on the pathway towards harmonization with these other social annotation tools, assuming that they're willing to cooperate.
8:10
Do they need to have the same lexicon or do they just need good documentation about what each thing means?
8:15
They in fact don't even need to collaborate.
8:19
I can go and change the way my application works to use their lexicon without any sort of coordination.
8:27
And yeah, so we technically don't even have to talk, but I think it's always nicer when we center the relationship because that's kind of what we're all here for.
8:38
I can just add, in our case with Marjin, when we decided to interoperate, we didn't actually— we coordinate our lexicons, but we didn't—
Speaker C
8:45
we're not sharing a lexicon.
Anish Lakhwara
8:46
So you don't need to share a lexicon.
8:48
You can just kind of coordinate.
8:49
Yeah, maybe one more question.
8:53
That's good.
Speaker D
8:53
Yeah.
8:56
Thank you.
8:56
Yeah, I'll echo everybody else.
8:57
This is a great talk to have and sparked great discussion too.
9:01
I think the question I had is like, what do you think about the provenance of how that data was collected.
9:07
For example, you might collect something with additional metadata that even if someone can curate that in the future, there might be something important about how that was collected or how might you annotate the collection process and sort of keep that history alive like a repository history of this is what happened when.
Anish Lakhwara
9:28
So yeah, it's a very good, question.
9:31
It's interesting to think about.
9:32
So like when I was working on Siems a lot, like one of the ideas that really resonated with me was like I wanted to build more primitives in the app protocol.
9:42
And so like we could have different kinds of data that we can like mix and match in various ways to like, you know, come up with new forms of wisdom.
9:52
And metadata obviously plays an important role, but it's something I went back and forth on a lot.
9:56
For example, like is it really important that I annotated this on my phone or in Firefox or through my read-it-later application, or is the important bit that it was at this URL on this time by this person?
10:11
And I kind of lent towards the latter of the way or where it was annotated is not necessarily super important.
10:23
I think the metadata that we collect, it needs to be— I hesitate to say useful because it's hard to say what will be useful in the future, but it feels like it needs to be grounded in more than just tools, right?
10:41
We can collect endless metadata about our tools and none of it turns into useful things.
10:47
But yeah, yeah.
10:48
Does that kind of answer your question?
10:52
Thanks again, Anish.
10:53
Thank you all.
10:54
All right.