00:14:34 · Alistair Smallwood
You're not trying to look for a 0.95 R squared correlation between two data sets, because mosaic theory is often qualitative, not quantitative, I. E. You want management commentary saying all has improved in this business unit and whether the second indicator had a similar direction. And so really explaining how to map those things and what matters and what doesn't matter. We also then have another skill which talks about the specific industry that we were looking at and how it constantly writes to that skill and updates it. And then also thinking about how you would think about actually doing this as an individual, economic transition and how you would qualify whether these two companies are related, and how you would check that through either management commentary or trying to understand whether there's any relation between their inventory build and ultimate revenues of a different business. Those types of things that you would do ordinarily as an investor, really trying to codify those into a series of skills so that the agent has access to those consistently. So you now have got the kind of data access and you've got the engine plus the skills. And then the third thing is arguably probably one of the most important is the evaluation engine. So every time the agent that's running thinks that it's found a relationship, you have to have a really high quality evaluation engine that sits underneath, that tries to qualify these leads, rechecks them, runs adverse aerial agents in the background to really understand these things. Ultimately, once all of those things are qualified, we actually then used JEV that sat underneath. It's obviously been huge hype across multiple platforms, but a really key thing when you're doing mosaic theory is ultimately you want to be asking a very simple question. If I had all of the information just from the company, what is the probability of the next quarter that I haven't shown you yet being positive or negative in terms of growth? And then I show you all of the information that I've collected from a mosaic basis, and then does that change the answer from Jeb suggesting that forecasts or the next quarter's growth would be more positive or more negative than the previous measure?
00:16:46 · Alistair Smallwood, Brett Caughran
And the fact that that was essentially deterministic massively unlocked our ability to do that work at scale and run that evaluation engine in a loop until we found some really exciting examples off the back of it. In terms of the second question, sorry, go, Brian. [00:17:01] S1: I want to just add one. I've had debates with a number of quants over the years that have made the claim that you can systematize fundamental investing with the technology. And I think the technology is there. The pushback I always have that no one has convinced me of yet is the fundamental investors or traders decision point is based off a concept of completeness of research, right? So there are times where maybe there's a signal in a transcript that this number is conservative, and that might be where I'm hanging my hat. But more often it's, I met with a management team at a conference and they're backing off of guidance, or I met with a competitor who's telling me this end market's slowing down and I can connect the dots. It's context that is gathered in meetings on the phone in messy sort of non agent legible ways. How do you think about that? Like if we really want to scale Mosaic theory, how do you think about scaling that completeness of context that is so critical to do that consistently?
00:18:14 · Alistair Smallwood
[00:18:14] S0: Yeah, I think it's a great point. I think the first point that I would make is that I wouldn't view this as a single end to end idea engine, because ultimately, there's a huge amount of stuff that you layer on top, which is all of those softer points that you talk about could be technical, could be crowding based, ultimately could be your fundamental relationship with the management team. But ultimately, really what we're trying to do is generate as many possible signals as possible in order to drive further investigation that then sits on the top, rather than replacing, oh, I've now got a seven quarter previous testable signal that I think you should trade on. Ultimately, I think that any fundamental investor that's been doing this job for any time won't feel that that's anywhere near enough to build the level of conviction that they would require. Ultimately, what it's trying to do is to add to lines of inquiry and fundamentally add to the potential conviction that they could have over an existing idea. In the same way that I've been to a conference and I've met a management team, they've mentioned something, I've then been at the same conference and a peer of theirs has then mentioned something else. Ultimately those two things together has forced me down a particular line of inquiry that might then lead to an idea in a few days time once I've done the work. Ultimately systematising that with public information is the same process, and because these things run autonomously, it then allows the investor themselves to have more time to layer on their pieces over the top to drive that holistic view of the business and potential investment idea.
00:19:50 · Brett Caughran
[00:19:50] S1: Yeah, yeah, that makes sense. And one other question I had, it's, we talked about the bitter lesson and prompt engineers and certainly trainers of prompt engineering were fairly rapidly bitter lessened, your chain of thought prompting and all these other duct tape fixes got very quickly abstracted into the skill. I've sort of been trying to understand, I've found actually increasing alpha in being really thoughtful and detailed about skill creation, but I'm curious your thoughts. Do you think that is durable or does this skill file become better lesson into the harness or into self learning or into post train models, sort of eval loop? What's your thought?
00:20:39 · Alistair Smallwood
[00:20:39] S0: Yeah, it's a great question. I think that the self learning piece is the most interesting angle here. I think ultimately, we start to run an increasing number of systems in a fully evaluated loop, whether that's human as the last stage of the loop or whether it's agent as last stage of the loop, I think the skills will be continuously improved autonomously as a result within a system that they work within. In the same way that if you as an analyst were doing consistent amounts of work on a given company, your understanding and ultimate thesis on that company improves over time. And I think that that's what we'll see from particularly loop driven agentic systems, where they are essentially self improving as they gather more context. What is particularly interesting about markets is they are essentially their own evaluation system. Ultimately, if you're not trying to call share prices and you're trying to call expected earnings, the learning loop that an agent can have as a result of that is really, really powerful. So let's say I ran my Mosaic theory at scale system for three days, found four really interesting relationships which would suggest that there is a beat coming in the next earnings for those three companies. If my loop continues to run, it will then see the eventual earnings that were then reported, try to understand why it worked or didn't work, and improve the skill accordingly. It might have missed something or misunderstood a transmission mechanism between the two companies and you just get this slightly self fulfilling process that happens as a result. In terms of the kind of bitter lesson build around skill writing, we still think the most effective way of skill writing is initial draft with whatever AI system you're using, but the handwork around them is still producing much, much better results. I would expect that advantage to slim over time. And ultimately, will come down to how well you can explain yourself to the AI in the first place to write the first draft.
00:22:37 · Brett Caughran
[00:22:37] S1: Yeah, what advice would you give to funds who are walking down that skill writing path right now? Has worked effectively for you?