Ajeya Cotra

This interview is part of the “A Peek behind the Curtain” interview series.

Ajeya Cotra is a Senior Research Analyst at Open Philanthropy where she worked on a framework for estimating when transformative AI may be developed, as well as various cause prioritization and worldview diversification projects. She joined Open Philanthropy in July 2016 as a Research Analyst. Ajeya received a B.S. in Electrical Engineering and Computer Science from UC Berkeley.

In this interview, Ajeya and I discuss her thoughts on research approach, including research phases and flow. We also touch on self-confidence, sustainability, and rest. 

Note: This interview is from early 2021 and some parts may no longer accurately represent the guest's views. The transcript has been edited for clarity and readability.

——————————————————————————————

Lynette: Let's dive into the key AI timelines project. I'm curious more about the process of this. What kinds of things went into a project of that size?

Ajeya: A lot of it is that Open Phil is rarely setting out to do long-term projects. In fact, we try really hard to set up to do the shortest projects we can, but they do pretty systematically turn into long projects. One genuine answer is that I did not realize it would take 18 months in calendar time and 10 full-time equivalent months when I started doing this. I think I would have been more daunted by it if I had realized that. There was a lot of wandering around to get into that place.

By six months in, I had a suspicion that it would be at least another six months, but I was still repeatedly wrong about how much longer things would be. This is partly because I think there were three different ways to be wrong or something, which I was wrong about to varying degrees. One is when people talk about the planning fallacy, they're like “things take longer than you think.” For writing a report, there are at least two ways that can be true.

One is that you underestimate how many hours you'll take per page you output, and the other is that you underestimate how many pages you'll be outputting. I tried to be quite empirical about hours per page but was still consistently falling prey to the planning fallacy because I didn't realize how much I would end up writing. Then there's another piece of the “didn't realize how much I'd end up writing”, which is, I didn't really account for the feedback back and forth, and changing things, and calendar time delays when people got feedback to me, and what I'd be inspired to push on and do more of. That's a big thing. I feel like for research, at least for long research or accidentally long research, a lot of where the planning fallacy bites is you realize you have a lot more to say. Or you need to say more to clarify something than you'd thought going in. It’s not so much that you didn't realize that it'll take you 10 hours to write a page or something. 

Towards the end, I ended up being empirical about that too. It's harder to gather data about it, but I would try to notice when I decided I needed to articulate a concept or write a new section, how long the average section was. It would feel like it would be five pages long, but actually, it was 15 pages. In general, I gathered a lot of data on that project and did a lot of forecasting at intermediate stages, trying to be empirical and base rate based from the past. It was still relatively tricky to try to do that.

Lynette: Do you think you have takeaways from that experience of continually seeing it grow, for which projects you choose to take on to begin with?

Ajeya: Yes, I think I do. For me psychologically, I think writing a report is a bad frame and encourages length. I've been trying to move in a direction of writing Google Docs for the consideration of other people at OpenPhil or writing EA forum posts or LessWrong posts. There's something where when the thing you're doing is in the reference class of something more casual and just traditionally short, I think there's just a bunch of beneficial things your brain does where you're like, "I don't need to explain this, I can gesture at it." One thing that really took lot of time in the timelines report -- and I'm not saying it wasn't always worth it, but it took a lot of time -- is precisifying concepts, that you could totally get away with hand-waving at in a conversation. You could also get away with hand-waving it in a LessWrong post. The goal, which I ended up backing away from, was to make it defensible to persnickety researchers or something, and that's a lot of work. A lot more work than you might think naively or than I thought naively. 

So I'm really trying to be in a frame of “I'm not doing projects, I'm not doing reports, I'm saying things.” When I'm saying things, I try to have a really specific audience of literal individuals in mind rather than “needs to be defensible” or some standard that's not tied to really particular audience members. That's a mode that I've been in for the last few months that's been helpful.

Lynette: What do you think you gained by going the full complete route that you would miss if you did a short version for everything? A 8020 version?

Ajeya: Well, a lot of it is just that a big goal of ours was to precisify concepts and defend ourselves to skeptics and stuff. Basically, that just always takes a lot of time and is sometimes worth it. I'm not doing stuff that meets that standard right now because I'm doing basically very inside baseball work. Where it's like “here are people who are on the same page as me about timelines, broadly how things will look and I'm trying to make suggestions to them about what to do.” That's just a very different goal. They're both legitimate goals. I think at the time that I was doing the timelines report, we really felt pain from not having justified our really unusual views or not having like put in a strong, good-faith effort to do that. Now we're feeling a lot less pain around that, especially there a number of other reports that are just about on their way out the door. We're more happy with our body of work in terms of external justifiability.

We're entering a bout of like, "Okay, let's talk with the people who buy this broadly. Let's figure out what to do about it. What grant programs have been up, et cetera." That's work that has a different pace and a different audience. I don't think I could have pulled off like fast-paced, short write-ups that serve the original goal. That's not to say that I did the original thing as efficiently as possible. It's just that, even at its most efficient, it would have been something that takes six-plus months.

Lynette: How do you go about this research-wise? 

Ajeya: I think there's phases. The first phase of big open-ended research problems for me is like the “wandering the desert” phase, which is a relatively low output, blank page, brainstorming heavy phase that lasts a few weeks or several weeks, depending on the size of the project. Where I'm just trying to poke at things and find an angle that speaks to me. I think you can dip in and out of the wandering the desert phase. I'm currently trying to do a project to get a sense of the last dollar benchmark we should use for OpenPhil, like what should we compare our grants against. How cost-effective is a grant for it to be worth it? 

I'm definitely in wandering the desert phase of like, maybe I'll try this gross modeling or maybe I'll try this functional form. Maybe I'll talk to this person. It's generally, in terms of hours turned out per week, quite low and very foggy or something. Then sometimes an approach that would take 20 hours to explore pops up. Then I explore it for 20 hours and that's a higher output. That's my favorite phase of research, where you've just thought of something that seems exciting and doable. Your audience is like, no one but yourself for now. That's a very productive phase.

I often feel like-- I don't know. Often in this phase, I don't know what to call it. It's just like I'm in a mildly manic mode where I frequently stay up at night and can't sleep because I have ideas or something. Then there's, if you make it through that, if something catches and you're like, this is the thing now, and I've done it, there's a second, sometimes even worse wandering the desert phase. If you were trying to present it to an external audience, how do I justify this? I know it's right or feel it's right or something. What exactly do I mean, and how do I say it? Which was a really tough and pretty long phase for the TAI timelines report.

In October of 2019, I had a whole report that was like 80 pages long. That was an internal report that I thought was almost done. Then, from November 2019 to February 2020, I was wracking my brains, trying to write it up. I had the thought that I was just “cleaning it up” or something which was extremely deceptive because it's just like there are so many assumptions that you can lean on. So many things you can handwave away when you're talking to people who are on your same wavelength. That's wandering the desert part two. Then there's cranking out the pages. There's some interleaving there. 

There was a lot of interleaving in the timelines report. I would wander the desert on one thing and then crank out some pages, how to explain concept X. Then that would happen all over again concept Y. For shorter projects, I think it's a one-and-done thing. You spend some time stressing about how are you going to say what you're going to say, then you say it. Then that final phase also involves a lot of like, for me, trying to sleep, but then realizing I have the perfect way to phrase something, getting up, and working through the night. That's how I would frame it. I think internal-facing projects tend to have only phases one and two and external-facing projects have basically two cycles of low productivity wandering and stress followed by cranking, flipping out.

Lynette: What happens after you've cranked out the pages and you have a draft?

Ajeya: It further depends on how seriously or how externally facing this thing is. Usually, I'll send it out for reviews. Usually, it feels easy for me. I don't know, it's rarely a game-changer. After I've been through the two cycles to get comments. It's sometimes tough to deal with them. Sometimes you have to do a bunch more work, but it doesn't feel like another wandering the desert phase usually. It feels more mundane.

Lynette: When you're in the first wandering the desert phase, when you're saying you're poking at things, what kinds of processes might that be? Do you have any strategies to help you get up interesting bits more quickly?

Ajeya: I don't think I have particularly novel strategies. I talk to people who might have interesting ideas. I just do a lot of talking to people and ask them how would they approach this thing. Then I try to look up past projects at OpenPhil that are vaguely similar to this project, and try to see if I can co-opt any of that. There's a lot of blue-sky thinking and a lot of soliciting input for the most part. I keep a list of ideas. It's generally, in terms of technology and technique, it's very simple. I haven't found really reliable ways to make it go a lot faster or a lot better so far.

Lynette: Sure. Do you have particular strategies for taking notes?

Ajeya: Not really. Let me think. When I'm thinking things through and thinking of ideas, I like taking notes with a pen on paper which is the classic advice. I think it just forces you to slow down and that's good. I don't generally read notes again. I think this is also fairly classic. 

When I write something up that I want anybody including myself to ever read again, I try to just write in prose. I generally don't have a long phase of having messy notes. I try to instead chunk things into short Google docs that stand on their own and make up points and they're written in paragraphs. I find that satisfying because I am personally a lot more likely to read something well-written. I don't necessarily have the patience to trudge through messy notes, even if I was the one who wrote them. When I really want to remember something, really want to remember why I believe a thing, I try to just write that down as if I were telling, not someone who is skeptical, but as if I were telling a friend and not expecting them to necessarily understand the messy notes. Notes are very much like an ephemeral thinking tool for me.

Lynette: Makes sense. I think there's a phase, it might be what you described as the second wandering the desert phase, but where you've already done the intellectually curious bit of figuring it out for yourself and now you're polishing it and making it presentable so others would find it compelling. A lot of people don't find that as interesting. Do you do anything to stay interested and engaged there?

Ajeya: Personally, I don't think uninteresting is the way I would describe it. It feels its own extremely daunting challenge. I feel a lot of negative emotions in this phase, but I'm generally not bored. It just seems, for me, there's been a lot to do to make things clearer. I also think of it as polishing, but I think that's often not empirically right. It's more another round of disentanglement and sharpening. At least that was the case with the TAI timelines report. Then there's a phase where you're not even doing that and you are just cranking out the pages.

For me, that's short enough that it's satisfying usually. I'm not writing a thousand-page tomes of things that I already believe. My report was long but it was like 150 pages. When I was really actually going on it, it was an hour or two a page. After a long period of racking my soul for “how do I put this point? How do I define this concept?”, I tend to find just writing a relief and generally positive. I crank up the tunes and I make it happen. It's not a large fraction of my experience. I haven't yet had the thing where it's getting boring.

Lynette: Sure. What strategies do you have for dealing with rabbit holes? Including when you're not sure if this is a rabbit hole or a useful topic?

Ajeya: I don't think that I've been historically that good about it. I think recently I've been trying to be a lot more mindful. Part of my strategy for dealing with rabbit holes is the thing that I said earlier.  Where when I have an idea, I want to just get it off my chest in a presentable form so that it exists. I feel, “yes, I've done a nugget of work on this thing.” So if I have an idea for, I don't know, a type of AI safety research that might be good. I think there's a temptation -- you feel a lot of excitement about your idea. I think there's a temptation to make the outlet of the excitement research and making the idea better.

I try to make the outlet expression and just saying “I'm just trying to tell another person why it feels exciting to me.” That often actually relieves a lot of the psychological pressure for me, so trying to do that more. I haven't done much of this so far, but I'm also trying to up that satisfaction by more making little posts publicly too. I suspect that would help too.

Lynette: It sounds what you're doing is you're breaking this down into questions and taking a question that you're trying to answer and writing up something on that. Rather than some broader holistic approach to doing the reading and research phase?

Ajeya: Yes. I think that's right. I think it's better psychologically and for efficiency to be trying to have something to show for every week. Like a thing that somebody could read. They're not trudging through a bunch of notes and it's not a work in progress. I'm happiest when I can make that happen. That's definitely not always the case, and not all research projects are shaped that way for better or for worse, but I try to be in that mode.

Lynette: Sounds good. What do you do when you get stuck and you're not sure where to go? What strategies do you have for dealing with that?

Ajeya: I think I'll often just turn to writing about it at the meta level and be like, “here’s what I've done. Here's why it feels bad.” I track my time, so I'll often also just go back over time logs and be like, “what did I try and for how long?” Then I'll be like “should I drop this project? Should I try this other track?” Sometimes I just think about that myself, but often I'll share it with a bunch of people at OpenPhil and be like, “what do you guys think?”

That's generally my go-to in terms of when I'm stuck and it's a way to have something to show for a week, that I wouldn't otherwise have something to show for, like “here's the document that someone can read that explains what's been going on.” I get some satisfaction from producing that kind of thing. Then often somehow something occurs to me as I'm doing that, that gets me unstuck.

Sometimes that's not the case and I'm just like “maybe I should drop this project.” Recently, I dropped a project because I had been wandering the desert. I'd written up a couple of these retrospectives or reports on what I've been up to and why it's been hard. Then we were just like “do something else.” It can be really helpful to avoid rabbit holing and spending too much time on things, to feel like you have a lot of options and there's lots of things you could do that would help. You don't have to cling to this particular project. 

That's also something that I have historically been pretty bad at and have been trying to more adopt this attitude of “there are lots of good things I can do.” There's no law of the universe guaranteeing that if I ask a question, I should be able to answer it. So I should expect that some of the time I'm going to ask some unanswerable questions and I'm going to discover that, and then I'm going to do something else good instead.

Lynette: Sure. Do you have particular triggers or processes for deciding which ideas you should drop and which ones you want to pursue further?

Ajeya: I think the main triggers are just very introspective and emotional. If I'm feeling crappy about something for a long time and just notice that I'm sad, I try to notice and think pretty hard about whether I should drop it. I don't think I've been able to find more early-stage or more objective triggers so far.

Lynette: Cool. Maybe focusing on shorter projects and more data points, how do you divide your research time between reading and thinking, talking to people, and writing up the short notes? Is there a rough breakdown in there that you tend to see?

Ajeya: Yes. Reading is a very small fraction of my personal research, just because I'm in areas where there isn't a lot of existing literature. If you lump reading and talking to people and thinking together into learning/research or whatever and then writing the docs into internal presentation, then I think I spend a third of my time doing the research and two thirds of my time writing stuff up roughly. Then maybe for the phase where I'm actually trying to write things up, I still spend some of my time doing the thinking and talking to people and racking my brains, but it's maybe more 20% or 25% of the time.

Lynette: Do you think that those other skills or processes that you bring to research? Things that you've learned to do that seem particularly valuable to you?

Ajeya: Yes, I feel like a lot of my work involves trying to understand people who have big ideas and are kind of confusing and have left a lot of holes in their argument. Trying to figure out or reconstruct a full argument and then critique it. There's a lot of extraction from conversations that I do and I think I've learned to be good at. I don't know that I have super specific strategies for it. Sometimes I think this is something that I had aptitude for and was doing this in some ways before it was my job. I like digging into people and I think I'm good at trying to restate what people said and really doggedly trying to make sure that I get it and can pass their intellectual Turing test. That is something that has been pretty helpful with the work that I've done recently.

Lynette: Does it seem like having a solid foundation or really understanding some ideas will help you build out other ideas more quickly or provide that foundation for learning new things better?

Ajeya: I think there's definitely major impacts on related fields which is pretty unsurprising. Thinking a lot about ML and AI for the timelines report has been helpful for-- right now, thinking about AI alignment strategies and how they could work, so just building out a tech tree of thinking about topic X in the course of project A, you probably want to capitalize on that by doing a bunch of further smaller projects that explore angles on it and exploit the fact that I invested in learning this thing and it's fresh in my mind. I would guess there's a more diffused learning things in general effect too, although that one's harder to introspect on and I don't really know how big it is because I feel like in one way or another, I've been trying to learn things in general for a long time, so I don't really have a lot of good causal attribution.

Lynette: Sure. What have you found are the best ways to solicit feedback?

Ajeya: I think that specific questions that you want your feedback givers to answer is nice and giving them particular deadlines is nice, and if they're external people, offering to pay is also valuable if you can do that. I generally feel like the larger the document and the vaguer the ask, the less likely people are to be helpful. Trying to anticipate objections and listing objections that you've already heard and asking people to say if they agree with them or if that sparks any thoughts, is pretty helpful and will get you a lot more responses.

Lynette: Sounds good. Kind of a catch-all wrap-up for general research here, are there things that you think you've learned to do unusually well?

Ajeya: I guess the thing about extracting information from conversations applies. I also end up doing a lot of back-of-the-envelope calculations or not so back-of-the-envelope quantitative models of things. That's been something that's been a little bit something I've specialized in relative to other researchers at OpenPhil and that's something-- I don't think I'm super good at it in an absolute sense, but I think I'm good at it in the relative sense. Those are probably the two big ones.

Lynette: How did you develop the skill for getting information from people? 

Ajeya: I think I'm naturally inclined to really make sure that I have understood something in an intellectual conversation. I think it's helped over time to become a little bit less self-conscious about not knowing things. A lot of the game for me is to stop people and be like, "It sounds like you're saying this, is that right?" and really paying attention and not letting things go until I'm satisfied that I've got it. I think that's probably the biggest thing. There's also just getting to know particular people and their communication quirks. I'm often talking to the same small set of people who have ideas that we want to vet or stress test or build off of. Just literally getting experience with them is helpful.

Lynette: Sounds good. Now I have two questions probably about wellbeing. What do you think are the critical things for you for mental wellbeing, happiness, making sure that your work is sustainable in general?

Ajeya: Well, for work sustainability, I think that I've noticed these cycles that I describe of wandering in the desert and then a manic phase of consolidation and writing. I generally try to make deals with myself where I'm just like I'm not stressing about how my hours are low when I'm wandering the desert, but I also don't try to force myself to sleep when I have ideas. Just last night I worked until 5:00 AM writing something because I went to bed and then I couldn't sleep because I had thoughts. That really works for me.

It's very stochastic and it's not going to happen every week. There'll be some weeks where I really want to write stuff down and I really know what I'm doing, and there are other weeks where I really don't know what I'm doing. The process of coming to know what I'm doing isn’t generally improved by putting a lot of pressure on myself to have a high tonnage of output because it's a lot easier to turn out a bunch of pages on something that is simpler and less thorny, but less central, which is something that I did a lot of in the timelines report where I just spent more time than I needed to talking about technical topics that were mildly interesting, but their central feature was that they were easier to talk about and less speculative than the main points.

I've become wary of the idea of like, "I want a lot of hours in a day or in a week. I want a lot of pages done." I try more to just hold myself to a standard of every week I hope to have some product to show for myself that's related to the central thing that I was supposed to be thinking of. Many weeks, that product will be like, "I tried these things and they didn't work.” And that's okay. A lot of weeks will be relatively low hours and that's okay. Sometimes I'll be really into something, and I shouldn't try to just make myself go to bed at a reasonable hour.

Lynette: Are there things that you do outside of work to rest or rejuvenate?

Ajeya: I live in a group house and I really like hanging out with my housemates. That's the main thing. Especially during COVID times, we've had dinner together almost every night and that's really nice. I also had been really into reading and then stopped for a long time and have very recently picked it up again, and that's always pretty absorbing and relaxing for me. TV with my partner is also really big. Those are my main deals.

Lynette: How much of a difference that is between things that you do for rest when you're tired versus things that you do to rejuvenate and energize yourself that make you excited to keep going?

Ajeya: I think tired activities are more like reading, watching TV, being on social media, passively hanging out with housemates. I think rejuvenating activities for me, a lot of which are harder now, I like to host people, I like to throw dinner parties, go on hikes with friends. For me, a lot of the rejuvenating activities are high activation energy social activities with people I don't live with, where I'm planning and hosting and making it happen, which is a big part of what I really love about life.

Lynette: Do you find that, for you, loving what you do is correlated with or necessary for doing it really well?

Ajeya: I think that having an inside view vision for research is very important for getting it done right. It isn't enough that somebody argues to you that it would make sense to do a certain type of research. It needs to grab you and be generative in your brain in a way that's hard to force. That definitely feels important for a research role. With that being said, I don't think every day or week feels like you're grabbed with something exciting and really like it.

Like I said, there's long periods of being for me, frustrated and lost in a fuzzy domain. I think it helps to just have patience to ride that out and be like “I am probing a lot of different directions. I am waiting for the feeling of traction.” That often is accompanied by joy, and excitement, and energy, and being in a mildly manic state, but that's not the whole job. A lot of the job is systematically and patiently poking around trying to cause that to happen. Then also a lot of the job is riding it out and following it through. I don't generally tend to think in a frame of loving the whole job but I do think just having energy is almost pathologically valuable for getting a lot of good stuff done.

Lynette: Are there particular things that you find draining to do?

Ajeya: Yes, it’s funny because I do love research but I find research more draining than, say, having a lot of conversations with people. I can put in fewer hours a day of it without tiring but that doesn't mean the process of it is a continuous negative. It just means literally that I don't have as much of it in me. That's like, other things being equal a disadvantage but it doesn't feel like it is dominant consideration for me.

It's more like what do I get done and the amount of energy I can pour into this. It's going to be very variable and there are going to be long periods where the amount of energy I can pour in is pretty limited. There are going to be some bursts where I can really put on the gas because I have a direction I like and there's a long way to go in that direction.

Lynette: Okay. Have you ever had doubts about whether you're good enough to pursue your career?

Ajeya: Yes. I've definitely had doubts like that and sometimes reality is not graded on a curve, so I don't know if I'm good enough in some absolute sense, even now. I feel like the stress that I feel from that has gone down over time. Just being here longer and having more things under my belt basically and noticing. I think when I'm in a wandering the desert phase, I’m especially likely to feel like I'm not good at the role. When you're new, you don't have very many cycles of that under your belt. It doesn't feel like a thing that's part of a cycle and that can be very demoralizing.

I feel like I'm more chilled out about it now because I'm like, "This happens and I'm probably going to be in this place for three months," but I've had so many three-month periods now and I kind of expect to see the light at the end of the tunnel. That was much less true when I was new and I didn't know if there would be this other higher energy, higher conviction period at all.

Lynette: Does guilt play a role in motivation for you?

Ajeya: I think guilt is a feeling I have reasonably often, not too intensely, and not all the time. It doesn't generally cause me to be motivated. It's more usually a symptom of being in an unmotivated place. Or being in a place where there's not so many hours I can pour into my thing and I'm wandering in the desert. I'll often feel, less frequently now but still sometimes, I'll feel guilty and frustrated that I haven't moved on something yet; things haven't moved forward; I haven't put in a lot of hours.

It's like by its nature it's not the sort of thing that-- Guilt can get you to do some things, like call your parents or maybe even run an extra lap or something. The bottleneck for finding traction in research isn't something that I find I've been able to usefully address by feeling guilty about it. It's more like an effect of being in a place where I'm for whatever reason not that productive.

Lynette: On the flip side here. What do you think are the benefits of self-confidence, maybe even what some would call arrogance?

Ajeya: I'm going to distinguish between a couple of different types of self-confidence or arrogance. A thing that I have found very helpful is to be like, "I am not married to my project. I'm not married to this idea." Other ideas will come to me and other projects entirely I could do and be helpful. That is really, really valuable for pulling back from the temptation to rabbit hole. I think rabbit holing often comes from a feeling of like, "Oh, I have this idea, it exists. I've been seeking and seeking and now here's the thing and I can sink my teeth into it." Being like, "I don't need to sink my teeth into this, it's not the only one. I don't need to cling" is quite valuable. 

There's another thing, like one of the other questions on your list was “how much am I intuition/curiosity-driven versus more systematically thinking about whether a research direction will be valuable.” I think, at least for OpenPhil research, it's very helpful to have humility about whether your research ideas are going to be valuable and take a lot of input from other people about that and be really willing to shape your original idea into something other people are on board with too.

I don't think that arrogance about how good particular ideas are, it's a flip side of the arrogance about “there are lots of ways I can contribute.” I think that's good and arrogance about “this idea I had is really great and I have a vision and I want to do it” is less good. Even though like I said before, having energy and an inside view for something is pretty valuable for getting a bunch of stuff done. It's like a tap dance.

Lynette: Sure. What about the aspects of self-confidence that are more emotional? It's like some combination of, “I know I can do this” or almost “I have permission to do this” or “I don't need permission from someone else to do this” that sort of thing.

Ajeya: Yes. I find that stuff quite valuable, and I've been developing more of it. In general, I think it's good. I feel like one angle there that's gotten easier for me over time is I don't need to figure thing out on my own. These problems and projects weren't given to me by my teacher, so there's no guarantee that they're solvable. 

I should not be surprised if I've set impossible tasks for myself, and it should always be a hypothesis that I have. It should be a salient hypothesis. That kind of self-confidence is valuable. Maybe it's the task's fault. I don't know, this isn't an exam. It's not like you "should" be able to do this thing. Earlier on, I had a lot of “I have to figure it out on my own” feelings and these days I’m much more likely to get somebody else's help or drop things. I think that's been good for efficiency.

Lynette: Are there examples of times when you've gotten someone help, and it made a big difference?

Ajeya: Yes. There was a project recently that I was thinking about investing in and how to invest optimally. I was just like, "I don't feel I'm excited about this, or I'm having ideas here, and I want somebody else to do it." Eventually, somebody else just did it. It was shuffled around. Priorities were shuffled around. I've just recently become more vocal about that and more vocal about being like, "I don't think this is for me." It's easier to do that when you feel you've been at a place for a while. There's more than one thing you can do there.

Lynette: Sounds good. Do you have an opinion on flow?

Ajeya: Yes. Flow goes back to the rabbit hole thing I was saying. I think flow in research, when you're not writing stuff up and constantly driving toward a product, is dangerous for me. I think for a lot of people more so than people appreciate, your attention and energy takes the path of less resistance. It flows downhill and that's not always going to be the thing you should be most thinking about.

In the timeline report, the most important stuff to get across to skeptics was almost entirely conceptual in answering conceptual objection, but I'm much more likely to enter into a state of flow when I'm researching something purely technical because that's easy. It's rewarding and so you enter into a state of flow. It's not easy. It's the flow thing. It's the right kind of challenge that give you reward every 20 minutes or something.

I spend a lot more time than I should have flushing out purely technical aspects of the report that were much less central. I should have spent more time in an unpleasant state of trying to express something fuzzy. That kind of flow can be dangerous. It can be really good when the thing you're in a flow about is expressing a thing. You just feel like you know something that isn’t written anywhere on the internet, and you want to say it. You're just like banging away and trying to say it. You're not reading a bunch of papers. You're not learning things. Your flow is just expression. I think that happens to me too, and that I'm very positive on but a lot of flow for me leads into rabbit holing.

Lynette: How do you scope within a project to determine what's most important, and what deserves your time while you're trying to be efficient?

Ajeya: I think it really comes back down to writing down and rewriting down and thinking about the high-level goal of the project. Having something that's an easy handle, that's like a sentence or two, and just checking in with myself a lot about where I'm at with respect to the high-level goals. Writing something every month about like, "When I do this last month and what was the goal? Have I moved toward the goal or not?"

——————————————————————————————

Enjoying the interview? Subscribe to Lynette’s newsletter to get more posts delivered to you.