Allison Parrish is a poet and programmer who researches and makes art about language, often in the context of computation and the Internet. She is the author most recently of Articulations, a book of generative poems from an algorithm which extracts linguistic features from over two million lines of public domain poetry, then traces fluid paths between the lines based on their similarities.She co-created Rewordable, a fantastic word-building card game and favorite Scrabble-alternative for my family. She is also the creator of the legendary everyword Twitterbot, where thousands followed the progression of English words from 2007 to 2014. Her github is here.

You may notice this interview has a different flow than the others, as it’s a transcribed conversation rather than the usual e-mail-based interview.

» Can you tell me a little about your background; how you learned to code, how you started experimenting with code and language?

Oh sure. When I was 5, we got a Tandy computer, a color TRS-80 computer 2 for Christmas and it came with a programming manual. It had just like the regular Microsoft BASIC that came on every 8-bit computer at the time, so that’s how I learned how to program. Just being 5 years old and sitting down in front of the TRS-80 and typing in the examples from the manual that came with it and doing fun stuff that way and that turned into what people expected me to do.

» So your first computer use was all command line stuff

Yeah, it was the Color Computer 2, and then DOS programming and then I took some UNIX classes as an undergrad, so it’s all been command line, textmode stuff, that’s always been the main interest of mine.

» So did that grow naturally into working with language?

It’s interesting, I hadn’t made the connection, but there was another game creation system – we were talking about MegaZeux earlier – but what came before that was ZZT. Do you know about ZZT?

» No, I don’t know about any of these things [laughs]

[laughs] So ZZT is this DOS game creation software that’s all in the text mode, it’s all in the extended IBM ASCII characters, 16 colors, a character can only be one color in the foreground and one color in the background and you’d use it to make adventure and action games and there’s a simple scripting language that went along with it, and for that I created a lot of games, I was part of this scene of people who made games for this and I still have a ton of friends who are part of that scene, the ZZT scene. And it is weird that it is all textmode, so you’re always fashioning these games out of letters and symbols, because there’s no way to actually draw pixel graphics, you had to pick a character. Like, if you wanted to place a plant in the world, it would have to be like the extended ASCII spade character that you made green, or something.

» So it’s still working across a 2D plane, it’s still visual, but you’re using characters –

Yeah, it’s all text based.

I was interested in language as a kid. I read Tolkien’s Lord of the Rings and stuff like that, got really interested in the made-up languages and did that for a really long time. I think that’s where my interest in language originally came from, from being fascinated with what Tolkien did in those books with the languages.

» Did you ever design your own languages that way?

Yes, I did. I was an active member of the conlang list for many years.

» Oh cool!

I’ve also met a number of friends I still keep in touch with, a number of close friends I still keep in touch with, from that

» Is that something you still follow now, in terms of people doing new work in that area?

A little bit. It’s not like something I do actively any more, and I think I would approach it a lot differently now if it were something I were still doing as an area of interest. Actually, David Peterson, who made Dothraki, was in my same class in Linguistics at UC Berkeley, so we knew each other back then. And he’s really amazing, really awesome, really smart – but yeah, now we’re totally going off in the weeds.

As a poet, I’m more in touch with the ways that language can be used to change minds, to do interesting things with language, things that affect the brain that I think was my original goal in getting into constructed languages was like: take this Sapir-Whorf Hypothesis at face value and then see where you can go with it. How can you make a language different so that in the process of learning it, you have to look at the world in a different way. And that’s also sort of what you do with poetry, right? So if I made a constructed language today it would be very focused on those issues of like when you learn this language you have to break down the world in different ways. And languages like Toki Pona and Láadan and languages like that, those are the languages I really look to in that area of creative practice as examples of poetic uses of constructed languages.

» Yes, we’ve talked about Toki Pona a little bit. You know, I wrote about that Oou language that Sonja Lang created – it’s all vowels, and it has a small vocabulary and is meant to create confusion, every word means so many different things, and you can have a conversation with someone and both of you can have completely different understandings of what’s being said.

Oh yes, right.

» So your background is that you studied Linguistics and poetry. How did you get into working in generative poetry? Was that the first computational work you were doing, in that area?

I wasn’t really interested in poetry until my creative writing teacher had us read Gertrude Stein. And then I was just amazed – it was just really awesome that language can do this. Because you know if poetry if it’s not Gertrude Stein, non-Gertrude Stein poetry, has this very like expressive, that narrative thing that happens in poetry sometimes, that’s sort of the poetry that you get exposed to when you’re growing up and getting a primary education, in the United States, at least. And I think, having at that point, already in high school, having that background in computation, just immediately made me think this looks like something that could have been generated by a computer. And then, from there, it’s a really easy step to go “Oh, I made a Perl program that randomizes the lines in a text program” and that’s where the energy kicks in and you can see now the text is doing something completely unexpected, that it wouldn’t have done before.

So that interest arose naturally from that combination of factors: constructed languages, Gertrude Stein, computer programming; all of those things together, it’s just sort of inevitable that it would turn into an interest in generative methods… maybe? I guess I don’t remember that moment when I was like “Oh, I’m interested in this thing.” I think it also was like it actually took me a while to understand that people didn’t recognize that as a methodology, that for most things, it’s a separate, weird thing that you do. Because since I was a kid, all of my creativity has been channeled through procedure in some way, right? So it’s been kind of a surprise, even now in my career, when I’ll have a conversation with someone who’s doing stuff that I like and i find out that they think computers are bad, or something.

Like, have a conversation with poets who are making interesting poetry that I tend to think of as procedural, and then realize that still using computers to do that kind of procedural composition is something that people don’t think of. That’s kind of abstract, I don’t have a specific example for that…

» It’s interesting how different it is between like you think of PhD music programs and I’m not sure what it’s like now, but like 8 years ago, everyone was doing computational music stuff. Algorithmic theater is still something that feels very new, even though the ideas are similar to what people have been doing in other forms for quite a while. Especially with poetry, where so much of it is procedural for the writer, but I guess the process of working through that procedure is important for them?

Yeah, and I totally understand that. But it’s always a surprise to me.

» You have a number of projects, more of less in order that dealing with travel and exploring the unknown, especially water-based travel, or maybe just – I tried the Our Arrival, the piece that takes your location and generates a

Um, what’s that called…

» Travel Guide

Oh right, A Travel Guide.

» But A Travel Guide, for me, for New York, it seemed to combine the idea of the subway into the idea of the river, and it felt like the sort of fantastical version of New York, kind of like Invisible Cities kind of thing, where you’re describing a place from a very specific point of view that gives it this very strange and beautiful idea of what the place is. And then The Ephemerides, all of them seem to have this epic travel kind of thing, and then your more recent work is more about sound and structure of poetry, and it’s interesting, obviously your recent work with the phonetic vector and work you’ve done around it, it seems like you’re honing in on something more structural about poetry itself, where the earlier ones have a more what you think of as poetic content. Does that make sense?

Yeah, that makes sense. What is your question?

[both laugh]

» Um… um, what is my question. For the phonetic vector stuff, a lot of it seemed to come together with like, a sort of toolkit you were building, and your own work. And I’m wondering how these come together. Did you start with: I want to see how this stuff works, and I’m going to build a toolkit, and then I build my own stuff around it, or like what um

Yeah, I mean, for me, my process almost always begins with technical experimentation. Like, “Oh I found out about this new math thing,” or – it’s going to look really bad in the transcript, “this math thing…” There’ll be a chance for us to edit this?

» Yes, I’ll put it in a Google Doc [Ed Note: As you can see, we decided to keep it all in]

I’m trying my best to make it as transcribable as possible.

Um, I’ll find out about a new library, or a new method that’s being used in machine learning, or in the case of the phonetic vector stuff, I was at the Alt-AI conference and it was Mario Klingemann’s talk about how he had taken all of these art assets essentially and then made them into a matrix by calculating certain properties of that matrix and then doing simple Euclidean distance on them to find similarities between them and then making these amazing automated collages of these art resources.

And from that, it was like, how do I do this with poetry? Because that’s what I’m doing. And I got on my laptop and made a directory and said this is my place for experimenting with that, and it just sort of came out of that, that idea of how do I use this math thing on poems. And with the Ephemerides, or Our Arrival, that was similarly based on I know how to parse a text into its grammatical constituents, how can that be used as a resource for poetic composition? Like being able to have access to that. Because like when Tristan Tzara is making a Dadaist poem, the only unit of language he had access to in an automated fashion was the word, his way of cutting up the newspaper was into words. The interesting thing about natural language processing tools is that it gives you access to all these other levels of language, like parts of speech, and grammatical constituents, and named entities, and stuff like that. So what would that process of randomization look like, with these other resources that the computer gives to you? So it was just totally based on now I know how to do this technical thing, how do I turn that into an actual manifestation that people can engage with. So that’s where it starts for me. It’s almost always bottom-up in that sense, from the technical procedure to the output. And then there are all these choices of what corpus am I going to use and all that

» For the phonetic vector stuff, do you see it as something that is a tool for computational poets to work with, or is it an interactive piece, to see the pieces progress, or what do you see as the engagement with it?

I wrote a paper about it and presented it at a conference last year, at the Experimental AI in Games conference accepted the paper. And the second half of the research for that piece was looking into what the actual research into phonetic similarity is, in the field of natural language, and in computational creativity and it turned out really nobody had been thinking about this issue of how do poems have phonetic coherence, which they obviously do, that’s one of the tools that poets have at their fingertips for making a poem cohere is things like alliteration, and assonance and stuff like that. And there’d been some studies of that, which were very limited and just looking at very specific, like does each word within a five word window start with the same sound, and counting that as the only kind of alliteration in poetry. But in fact in poems, a poem will have a very sophisticated phonetic coherence that affects the poem as a whole, affect it in non-local ways, and recursive ways, and stuff like that.

So one of the things I argue in the paper at least is that this procedure could be a resource for people who are doing computational poetry at least as a way of saying when you’re in the process of building a poem computationally, when you’re selecting what words to include, you could use this measurement of phonetic similarity and basically like scrunch those vectors of those words together so that you get some phonetic coherence out of it, right? And I haven’t used that in any other procedures of mine so far, but I do hope that others who are making computer-generated sonnets or whatever will be able to use this resource to do things other than just rhyme when they’re  talking about how a poem sounds. So I do see it as not just my own thing that I did, I see it more as some general research I’ve done in general methods of computer-generated poetry

» You have some pieces, obviously the game, that are interactive. But then you also have a lot of work that are more like classic computer art, the generative art. I’m wondering if you much at the work of computer artists, that you find inspiring. And also, you’re teaching at ITP, I guess I’m interested in that way that the computational literature people and the digital artists are meeting in that middle ground and how they’re both working with language, what that kind of conversation is like, especially at ITP.

Um ok, what part of that should I answer first

[both laugh]

» Yeah, whatever comes to mind first

Who are you putting in the category of computer artists

» That’s the thing, like I’m thinking of people like Alison Knowles, because I believe you mentioned her, I’m thinking of Fluxus, I’m thinking of people who work with language but approach it more from the contemporary art, and I’m wondering how the people approaching it from the literary side, how their concerns are different, what happens in that conversation between them–

Yeah, that’s a really interesting and complicated question. Because this is where I got stuck in the email* because it’s really difficult to explain. And it’s difficult to talk about without coming across like I’ve been wronged. Because the kind of, I mean I do very much look specifically at the work of Fluxus all the time because there’s so many artists who refer to that scene whose work resonates with the work I’m trying to do like Alison Knowles, Jackson Mac Low especially is my continual source for creative inspiration. The interesting thing about, say, Allison Knowles for example, is that House of Dust is really her only computer poem, right? I mean, she didn’t like work on House of Dust and then say well, this is an interesting field of practice, I’m going to stick with this for like five or ten years and really explore the space of it, and that’s really similar to a lot of other people in Fluxus who were working computation or with language, actually, like of the Fluxus people, is there anybody other than Jackson Mac Low that you could call a poet?

*[Ed Note: originally we started an interview via email]

» I don’t really think of them that way.

Well, right, exactly, for most people, that’s not even how you’d think of Jackson Mac Low, so – whereas that’s my primary interest in that scene is what were people doing with language in there. And that trend of like, oh, I’m an artist and I’m working with procedure or rules or computation, and I made this one little piece about language and then I’ll go back and do all this other stuff. That repeats itself to this day. Like, I know a lot of my contemporaries in New Media Art and stuff like that, will do their one computer language project and make the letters appear on the screen and make the letters move around with your Kinect sensor or whatever

» Right, right. Well, that’s more graphical then -

Well, that, or other related projects. Like Kyle McDonald will make one project about text and then just sort of like dust his hands off and walk away. So it’s weird because when I’m talking to people in that scene, I’m unusual because I’m focused on language, like that’s my gimmick at ITP is that I’m the person who’s interested in poetry, right – versus when I talk to people in literature – and I don’t have a background in literature, I don’t have an English PhD, I don’t have a poetry MFA or anything like that, so I’m very much an outsider in that scene. And there are many people who have very generously welcomed me. But there I’m an outsider because I’m interested in computation, right? So it’s like, on either side, nobody is really understanding what I’m saying. And I’m notable for opposite things in those two areas. Which is kind of frustrating. So when it comes to the intersection between those two things, there’s just me and Nick Montfort, like standing around awkwardly. And that can be frustrating and a little bit alienating. And that’s actually one of the things I’m hopeful about with this book, about this series of books that Nick is editing about computer generated works is having it on this established press, Counterpath, that’s known for publishing experimental poetic work, lends it this kind of legitimacy that I feel like means that it will feel more worthwhile for people in the literature scene to pay attention to it. I don’t know if it will help me with getting a promotion at ITP because I don’t know if they’ll understand like publishing a book of poetry as being something that is worthwhile, so it’s a trade off, I guess. Does that kind of get at that question?

»Yeah, yeah, definitely.

Is there something more specific you’re-

» What’s nice about that is that in contemporary art, you want to be defining your own scene, and defining who’s around you, rather than falling into this category, it’s the tension there is much more interesting -

I know that I’m not the only artist who has this problem. I mean, I imagine most artists defining their own genre-

» But when I think about Fluxus and writing, I think of the event score, which is sort of a way of bringing writing into performance or maybe performance into writing, but I think of Yoko Ono working within the poetic space

Yeah, sure

» But like what was I going to say… take the work of someone like Pall Thayer, whose code & language work is based very much in the contemporary art space, it’s not meant to function in the literary space, and then you have people like Mez Breeze who are much more in the literary space, is also merging code and language in a way, I’m wondering if there’s more understanding across these groups now, as opposed to oh, that code doesn’t run, it’s not code art but code poetry, like the language isn’t being used in a very expressive way here but I feel like functioning in that space of language and code, I think it’s richer drawing from both sides, and I’m wondering what the conversation across them is like, it sounds like there’s still somewhat of a divide. There’s like WordHack, which does bring those together.

Well, WordHack is a really interesting example because it is like - the practices of the people who speak at WordHack are just incredibly diverse, in a way that it’s sometimes hard to even understand what the similarities are, right? Like I absolutely love the curation of that series. Sometimes there are people there that I wouldn’t recognize as being practitioners in the area that it’s supposedly about. Which is ultimately a good thing. And I don’t know – maybe there’s a conversation going on and I’m just not a part of it.

But the conversation for me is mostly well, “why are you doing that, Allison?” Or: “That’s very cool.” I will say that I’ve had a more positive reception to my work in literary circles from people who aren’t explicitly working with digital stuff. I feel like traditional electronic literature people are not as interested or excited by procedural work, for better or worse. Which is super confusing to me because in my brain, when I hear “electronic literature,” when I hear “e-lit,” to me that means computational literature, that’s the only way for me to understand it.

» Yeah, I would think something generative or something interactive. Interactive fiction, those two things.

But it’s much a broader category that includes other things, that includes hypertext, and stuff like that.

And so I find that finding the place where people are doing the same kind of work is difficult to do. And I spend a lot of time trying to figure out how to talk about my work to particular audiences. To make it legible, and that is vastly different depending on whether I’m talking to a literary audience or a technical audience or  an arts audience or a science audience.

» It’s great that your work, especially the recent research has something to say to each of these groups.

Yeah, well, thank you for saying that. That’s what I was going for.

» So let’s talk about the state of Twitterbots and Mastodon-bots. So at the time you created everyword, this was very much a new idea, of this kind of procedural action being carried out on Twitter, but since then, your project has become very well known, and there’s been a lot of projects that sort of play off of that, or create similar kind of procedural type feeds. THen there was the whole horse_ebooks, where it felt like it was this magical bot that seemed to have all this wisdom, and everyone was so curious what kind of algorithm could do this and of course, in the end, it was a fraud and everyone lost interest, it wasn’t a bot at all. And now it seems like there are a lot of bots that are more expressive texts that are more edited, where many texts that are generated and only some are sent through, which is how I think that sex bot one works, you know what I’m talking about?


» But almost, it’s generative, but then there’s also this editing moment. And trying to maybe work back to what was promised by horse_ebooks but no one’s ever really accomplished. That’s what occurs to me as an outside person to that space, but I’m wondering what is the state of that kind of bot now, are there recent Twitterbots that you find interesting or-

Hmm, I’m trying to think. Well, my interest has sort of veered away from that recently. I think partially in response to just Twitter as a platform not being a satisfying place to do that kind of work anymore because a) it’s actively propping up an authoritarian regime here in the United States and elsewhere in the world and I don’t feel like making wallpaper for that house anymore I guess.

And also I think that the way they curate the feeds now makes it hard for artbot content to surface. Yeah, so I haven’t been doing a lot with that lately and I guess I haven’t really been paying attention to other makers in that space lately either, which isn’t a great thing.

» How about in Mastodon? You’ve been active there or -

A: So yeah, I’ve pretty much switched over to Mastodon pretty much 100% for my daily social media stuff, which feels really good, I mean it feels bad to not be able to talk to my friends on Twitter, but also Twitter was just like literally driving me mad. Like it was extremely toxic for me in a way that I couldn’t control. So resolving to not use Twitter except to check my notifications every day has been very positive for my mental health. And I love the community on Mastodon, it’s just a really good place for doing the kind of work that I do. I share a lot of work on Mastodon, but I haven’t really made any bots for Mastodon, for whatever reason. I think it’s partially because I’ve been focused more on these, I’ve had to focus more on larger scale ideas.

I mean, the great thing about a bot is that it can be just sort of a one-off, right? Like you have an idea, and then you turn it into a program and then putting it on as a bot is an easy way to essentially publish it in a way that people are going to be able to engage with easily. But like for this book, for Articulations, which just came out, that was the end of a process that was a year long research thing and there was no stopping place in the middle where it felt like I’ve done the appropriate work and I’m ready to publish it, right? And it’s harder now to feel okay with having those one-off ideas and putting them up as bots because it feels like I should be working toward these larger-scale things

» So like for the larger scale things, so you’re saying that this is it is a generative work over all but it’s also edited, so you also decided this was what you wanted to include in it

No, not really, So I did twenty different runs of the algorithms. And then just selected the one that I liked the best, but there are no - there are two lines I ended up excising from the manuscript after the fact because they contained weird slurs, but other than that, there was no copyediting of the text, it just came straight from the algorithm

» I guess I’m talking about an open system vs a closed system. With a bot, it can run forever, and continually writing new things, and with a book it’s like here’s the final version, it’s published and here it is.

Well, what I’m saying is that I did several runs of the program and selected the ones that I liked the best. But from one word to the next, in the text, the only decisions I made about them were algorithmic. So it’s not edited in that traditional sense of I picked which lines to include, the end.

» But would you imagine having a book that would run forever…

I guess I could. I don’t know if with that particularly algorithm that would be that satisfying.

» You’ve done it a few different ways. You have the bots that - well, with everyword obviously it had to come to an end, but with most bots it sort of follows a certain pattern and can generate new situations and then you have like the zines, where each one can be different, have different combinations within the same series, and that still feels open, but is closed, and then the book. I guess for each one it seems like it’s important that it’s not manipulated, that it all comes from the algorithm itself.

Yeah, no, it is. And that’s where my interest is; that’s what’s interesting to me is how to make those decisions algorithmically. It’s not as interesting to me personally to make those individual word by word choices in that intentional poetry way, so it’s not like a purism in that this is how I think it should be done, it’s just me trying to be true to what I’m actually interested in knowing. Is how to make those decisions by laying out the rules specifically.

» So maybe you can walk me through your process. Obviously with the phonetic thing, you had this bigger question you were trying to answer and from the research you had specific pieces in applying it to specific poems to see how they’d be transformed. But when you’re working on the algorithm for this specific project, for Articulations, how does that come together, does it start with an idea that you’re exploring or does it start from start, like how does it come together for you?

Yeah, I mean like with any project, it comes together from having knowledge about particular things. So the algorithm is based on the CMU Pronouncing Dictionary. That’s where I got the phonetic information about the words. And then the question was just like Okay I see all this interesting work being done with clustering and similarity-based work that was happening and is still happening in machine working crowds and you can see the big t-SNE like there’s this big Google project of t-SNE like artworks like the showing of clustering artworks

» Oh like the hallucinating one, like the -

No, not the deep dream stuff, I’m not even sure I’m referring to real a thing here, it was also a thing that Mario Klingemann made that – for me, recently a lot of the work has come from this machine learning thing is happening with images, how would you do that poetry instead? And obviously a lot of machine learning stuff is happening on text but it’s almost never just the playful kind of machine learning work that’s happening with images right now – so in that case, it was like what would it mean to do similarity-based clustering with poetry. And then I have Project Gutenberg, which is another resource available to me. And I have a dump of everything that was in Project Gutenberg, which was just sitting on a CMU server, and so I thought like well, I want to get poetry, where do I get poetry, I get it from Project Gutenberg, because there’s a lot of poetry in Project Gutenberg. I find all the lines of poetry in Project Gutenberg, put them in a big text file, and then how do I know if these lines are similar. So I did a whole bunch of different experiments. Like semantic similarities is one way of approaching that. And with semantic similarity, you can use word2vec word vectors to find semantic similarity. And if you look back to my Twitter account two years or whatever ago, you can see I did experiments in that area. THe semantic similarity was sort of interesting but it wasn’t like Earth-shaking and I didn’t quite know how to make it not just like lines that have the same words in them. I feel like I could do a better job of that now if I - and then I thought what’s another way poetry can be similar – it can be similar in how it sounds. And so then I worked on several different ways of trying to extract some way of telling whether two lines of poetry sounded the same. And there’s been a ton of research on whether things sound similar, like scoring two things whether or not they sound similar. But there hadn’t been a way, as far as I know, of positioning some stretch of text in a vector space according to how similar it was with neighboring things. So that was like the primary like thing about this research that was different was to think about phonetic similarity the same way that word2vec thinks about semantic similarity. And I worked on a few different ways of trying to do that and drew on my Linguistics background to figure out, like I want words that have phonemes that are similar to - I want to break down the phonemes that are similar, so that "b” and “p”, even they are distinct phonemes, are closer in sound than “b” and “sh” or something like that. So I broke down those into their features and then did this analysis of them, and that’s where it came from.

And then the idea for a random walk, which is sort of how this is composed, moving from one line of poetry, to the most similar line of poetry, to the most similar line of poetry to that, while avoiding anything that’s been already picked was like I need some way of unlike Mario Klingemann’s thing, where he shows the images in two dimensions, in an arrangement that way, I need to do it in one dimension, because poetry is one dimensional, it moves from beginning to end. So I needed some way of linearizing that space. And a random walk is just a handy, easy way, of linearizing a space, right?

»Okay, that makes sense. Have you looked at rhythm as well?

Yes, so this algorithm doesn’t actually take into account the stress of the lines. I think probably some information about stress does leak through, just because I think stress tends to be associated with particular phonemes, but maybe not. I think I would want to include that stress information in the future, but that’s also, a lot of poetry generation algorithms will use the CMU Pronouncing Dictionary or other methods of determining the stress patterns of the lines to match them up in iambic pentameter or like Ranjit’s Pentametron is basically doing that work. Meter feels different because you need to have like an exact match for it to work, and finding two lines that are similar metrically isn’t, feels more like a precise rule-based task than a statistical task. But maybe I’m wrong about that.

» That could be an interesting problem

No, it’s a super interesting problem. Like doing automatic scansion of a line of poetry is a an interesting problem that wasn’t aesthetically interesting to me in that moment, so yeah. So that’s like a very technical description of where I was coming from.

» So after Articulations, do you think you’re going to be moving on to another subject of inquiry, or do you still have some ways to go in exploring phonetic space, the CMU Pronouncing Dictionary, and all that?

I don’t know. I’ve been having real trouble coming up with a next project. I think that stems from the fact that I just sort of – this is my second year on the full time faculty at ITP. And before that, I was working as a writer in residence at Fordham. So it’s really just these past three years that I’ve been in academia. And before that, all of my ideas like basically my only choice was to do small scale projects. Because then it was like what do I have time to do in my spare time when I’m not working. And so that’s where projects like the Twitterbots came from, was just like - I can put these things together, that’s all I have time for, it’s over. Where as now, I feel pressure to make projects that I can write grants against, or hire an intern for or something like that and I feel like I’m wasting my time with these smaller projects, and this is all in my head, it’s something I have to get over, but it just feels like it’s my responsibility to make these larger scale projects but that’s also like creatively paralyzing right now, because it feels like the next idea has to be big and wonderful and good as opposed to something I can maybe just throw away

» It seems like if you tend to explore subjects for a few years at a time, it seems like the small experiments are what let you find what the next big thing is you want to do.

Yeah, exactly, and I have to develop the emotional fortitude to do that and also the get into the habit of having that daily practice of doing experiments, right?

» But you did recently finish a lot of work and it’s been a lot of good attention to the project, so it’s a good space to be in.

No, yeah, it’s a really good space. But I also feel like - it’s the first time I’ve had a book like this that I’ve been promoting and this extended period of work and I feel like I’m just making people sick of it - I’ve done so many different readings of it. I just feel like every time I’m reading someone’s going to be like oh it’s this bullshit again.

» No no, and I like that you present it a somewhat different way each time, because I’ve heard you present it a few different times

But I am very eager to move on to the next thing, but I have no idea what it is.