Facts For Sale
April 2, 2026
Knowledge is the most valuable thing in the world. So why doesn’t it transact? To answer that question, let’s suppose I know something that I think you’d like to know,1 and examine how I might sell it to you.
I Trust You
One option is for me to trust you entirely, tell you the thing up front, and then ask you to please pay me. Maybe you will! If we have a longstanding relationship, and I am a reliable source of valuable information, you probably will. This is the informant model. You could also decide that my information wasn’t particularly valuable to you, and that as a result, you don’t owe me any money. I will be sad, because whatever monopoly I may have had on my information is now gone. Even if there exists another buyer for whom the information might be valuable, the cat is out of the bag. Last but not least, you could refuse to pay me even if you do value the information, and I lack meaningful recourse.
You Trust Me
This sounds better. I tell you how much I think the information is worth, you raise your eyebrows, we negotiate for a bit, and you pay me. And then I tell you the thing. Much better for me, much worse for you. This is the consultant model. If the information turned out to be useless or overpriced, you’re still out the cash. If we have a longstanding relationship, I might give some of the money back by way of apology, or promise to do better next time, but now we’re back to the informant model. This won’t work in a faceless marketplace in which we’ve never met before and will never meet again.
How Did We Meet Again?
If we really are in a faceless marketplace, it’s not at all clear how we found each other. If you had to seek me out, I presumably had to say something about the information I was selling, which is obviously rather delicate. I could also have branded myself as knowledgeable about a particular topic, but at that point, we’re running an expert network with extra steps. If I sought you out, then presumably you had to tell me the type of information you were in the market for. While it might be easy to indicate a general domain of high value information that would interest you (e.g. you are a tech VC in SF, you would like to hear internal announcements from foundation labs), it’s virtually impossible to enumerate the set of things you would like to know before you’re inferencing them. Most of what you’d pay for is valuable at runtime but unknown at build time.2
Google Is Your Friend
You probably don’t want to pay me much for anything that you could have found out for yourself. And in 2026, a decent proxy for what you can find out on your own is what a web search enabled LLM can tell you about the world. This should go without saying, but if Claude knows whatever I do, I should pack up and leave. This is, in a sense, why information retrieval and AI forecasting are so central to Talarion as a business: AI sets a (high) floor for valuable information.
Liar, Liar, Wallet On Fire
In the words of Robin Hanson, if people were always open and honest with themselves and others, and believed that others were this way as well, we would have no need for information markets. Unfortunately, I could be lying to you. Credibility, or the lack thereof, is a foundational problem for any information market. Expert networks address this problem by establishing personal credibility — verifying the information itself is less important if the source is trusted. But in a true marketplace, resume vetting is far too much friction. Prediction markets keep incentives compatible by asking participants to trade on their beliefs: you bet on what you think2, and earn when you’re right. There’s no point in lying to the market. If only you could forecast on … anything.
Next Up
In my next article, I’ll explain how Talarion solves these problems, and our vision for the world’s first true information marketplace.
1 I presumably do in fact think this, or I would not be writing this blog.
2 Consider this analogy foreshadowing for the next post…