Bitcoin Desire Hits 8-Year Low | Michael Sullivan SLP756

**Michael Sullivan** (0:00)
Right now, Desire is actually at the lowest it's ever been in eight years. People are really not talking about Bitcoin the same manner that they have. And this gets kind of weird and conceptual because psychological moods and that kind of thing. But these moods do change in different ways, and they kind of have different relationships to price, at least historically.

**Stephan Livera** (0:17)
Hi everyone, welcome back to Stephan Livera Podcast. Joining me on the show today is Michael Sullivan. Michael has been doing some really interesting work on sentiment analysis, and I believe you actually also authored a book as well.
I don't know as much about that one, but definitely wanted to chat with you about the sentiment analysis and what you're doing there. But yeah, first of all, welcome to the show, and yeah, let's hear a little bit from you on your journey on this sentiment analysis.

**Michael Sullivan** (0:44)
Yeah, thanks so much for having me, Michael Sullivan, pumped to be here. So this sentiment analysis thing has been a huge rabbit hole for me, but a little bit of my background is that I've been an engineer for like 13 years, I've been a novelist and an author for about a decade, and I've been really obsessed with language. Like I've always kind of noticed the words that people choose to use, how it evolves over time, and these kinds of things. And then during my time in Bitcoin, I also have kind of realized like how much narratives play such a huge role, and especially like as the price swings up and down, the kind of narratives that evolve around tops, the kind of things that people get angry about and latch onto around bottoms. I just always thought this was interesting. So as of about like six months ago or so, I decided to just like really analyze some of my own language, specifically looking at my X data over time, and see if there's any trends there, see if I said anything interesting, and kind of look at some of the dumb stuff I'd said around tops, some of the things I was talking about around bears, and figured out it was actually very fascinating. And I've kind of evolved it from there and developed this entire large system for looking at like general emotions and moods of the crowds. We can kind of get into if you want to, but it's been really interesting.
And just tons of depth there that I didn't anticipate when I first started doing the work, but it's been really, really fun.

**Stephan Livera** (1:54)
So what was counterintuitive? What was something that you didn't expect?

**Michael Sullivan** (2:01)
So the granularity of different emotions has kind of been something I've gotten much deeper into as of late. Right away, when I first started doing it, I just kind of looked from like a broader, like the way you do sentiment is like with machine learning. Basically, you can see if an individual chunk of text is more optimistic, more pessimistic, these kinds of things. So I did it in a really basic way initially, where I was just kind of looking at people who are happy or sad, sort of like the fear and greed index, where it's really rudimentary. It's like just a binary yes or no, and there's just like one emotion. But what's been so fascinating was kind of counterintuitive, was how granular different emotions can get, and how they don't always interrelate in ways you might expect. And to give a more concrete example of that one, the conviction is one of the most interesting emotions or moods that I've looked at.
Because it's non-directional, whereas people can be really convicted around tops, people can be really convicted around bottoms, people can use more hedging language and be less convicted around these same time periods, and how those things interrelate with other moods is something that was not obvious to me right away, but it's been one of the most fascinating parts of this work.

**Stephan Livera** (3:09)
So I guess, let's put that in practice. Does that just mean people might get overconfident in the bull and overbearish in the bear market?
What are some typical things that you will notice when you analyze that data?

**Michael Sullivan** (3:26)
So there's a reflexivity to sentiment or emotions or narratives, where when people are optimistic, they'll disproportionately latch on to things like the Strategic Bitcoin Reserve. It was one of the bigger examples of this around the end of 2024, 2025, where people talked about there being a persistent nation-state buyer.
And this is the kind of thing that people really latch on to and have higher conviction around. During those times when price is good, everybody's more optimistic. And then the same kind of narratives, even though that's still moving forward right now a little bit, there's still stuff coming out, Besant's talking about it occasionally, is almost dead. There's like negativity around it. The same exact narrative, even though it has changed slightly, it's in a slightly different context now, but like there's a ton of negativity around it. So that kind of thing has been really fascinating, where you can actually see the same narrative evolve based on where the crowd is at mood-wise and how much the crowd's mood does relate to the price, because there's definitely a huge correlation there.