Mantic Monday: Let Me Google That for You
astralcodexten.substack.com
astralcodexten.substack.com
For the Solana ignition hackathon I played around with building a search engine based around a per-query prediction market. It’s fun to try to structure different problems in this way since it’s totally different to how we typically solve things in CS.
I agree that insider concerns are a problem, but I think there's a very large space for questions that, as Hal Varian says, (a) are of interest to the company but (b) do not reveal financially critical information.
Consider OKRs tracking engineering milestones. Estimating your team's progress a quarter or year out is basically forecasting, and participating in those forecasts does not generally make you a financial insider. (Though some OKRs deal directly with finances and would in fact do so.)
Examples: Real names and Google+. People objected and questioned these, and others, without a prediction market and they still got pushed through. There are products that are still alive but barely hanging on that were questioned too.
A successful gleangen would eliminate many PM and VP roles and they aren't going to let that happen.
(1) Ask "Will Google+ institute a real name policy by X date"? [This may not be useful to know.]
(2) Ask "How many users will Google+ have on X date with a real name policy?" and "How many without?" [Potentially powerful, to the extent that they're accurate and decision makers trust them.]
(3) Ask "What will the app rating of Google+ be on date X?" [Prices in all factors, so can perhaps reveal the effect of policy changes before they happen.]
I don't see prediction markets replacing OKRs or strategy planning, but rather complementing them, kind of like A/B tests and user studies. There's still a lot to figure out!
I'm creating a faster search engine for programming queries and this topic is pretty interesting. Could you share more info on how your engine works?
The concept is to build a search marketplace that feels like a normal search engine to the end consumer. The hope is that you'd build an ensemble of expert vertical search engines that each can get traffic because they're part of the larger exchange (rather than needing to build a general search engine to get any traffic at all).
You can look at my cringe video / code for the hackathon here: https://devpost.com/software/askbid-a-decentralized-search-m...
The basic idea is:
1) The user posts a search query to the marketplace. This starts a time-limited market around the question "which result will the user find most relevant for this query".
2) Search engines decide if they have a relevant result to that query. If so, they post their result(s) to the exchange, and mint and hold "yes" tokens or mint and sell "no" tokens against their result.
3) Other search engines / information traders do the standard prediction market thing and trade "yes" and "no" tokens for the various results posted to the exchange until price converges.
4) The user views a SERP that shows the results (probably ordered by descending price) and picks one.
5) The "yes" tokens become redeemable for escrowed crypto if that result is chosen. The "no" tokens become redeemable if that result is not chosen, or if the market expires.
I have a shitty prototype on Solana's Devnet where I need to fix and tune the rewards to each participant. I figured crypto folks might be into trying something like this, even if it really is just a convoluted way of calling Bing until I can convince people like you to join.
Right now I've been faffing about trying to find ways to simply explain this to folks to try to find people who are interested. Would love any feedback you have. :)
Once you get a hang of it, you can definitely use for better decision making inside a corporation.