I hope all 7997 of you are having a great week so far š„
If you're enjoying my writing, please share Terminally Onchain with your friends in Crypto and join the /toc channel on Farcaster š¤
Back in February, I wrote a 4000 word deep dive on the history of prediction markets. In the post, I expressed the need for different form factors supporting the long tail of local, community based markets.
"As prediction markets enter the mainstream by demonstrating forecasting accuracy for major upcoming events such as the 2024 elections, there will be more capital to incentivize liquidity for the long tail of local markets.
People naturally care more about events and outcomes directly related to them than general ones. Prediction markets will need better marketing and UX in order for there to be increased participation in markets past national politics and culture."
And with the launch of @0xsmallbrain's latest drop tmr.news this week, it's clear we're finally seeing early glimpses of this happening now.
Today @sofianeflarbi and I are launching TMR.NEWS" target="_blank">TMR.NEWS: the first natural language prediction market. Judged by AI & secured by crypto
See tomorrowās news at https://t.co/PUU6p5h468
522
TMR news is a simple, yet amazing concept: you place a bet on what you think tomorrow's NYT front page headline will be. And the next day, once the NYT website updates, AI will determine the winners based on which submissions were closest in meaning of the actual headline. Key word there being meaning - it's not as simple as throwing in buzz words.
The launch only happened a few days ago and people are already coming incredibly close to predicting the actual headline itself.
What I love about the product is how freaking simple the frontend is. Type your answer in a text box, decide how much you want to bet based on your confidence level, and hit submit.
But with that being said, the backend and mechanics of this idea are quite intense.
For starters, this concept merges 4 of the most interesting consumer tech applications in the space right now:
Layer 2s & Stablecoins: deployed on Base allowing users to make sub-cent bets and receive payouts straight to their wallets. The product uses eth right now, but can easily incorporate USDC to make the rewards appealing for a larger audience.
zkTLS: using Reclaim Protocol takes away the need for standard oracles used in prediction markets today. This works by fetching soft HTTPS data and then cryptographically verifying the information to make it hard onchain data. In this product, the fetch is on the NYT website for their headline which is then verified using zkProofs. Note: I'm purposely not getting into the technicals here, will save that for a long overdue zkTLS deep dive.
AI: using GPT, the AI will serve as the oracle and make judgements on which submissions come closest to the meaning of the actual NYT headline of the day. If the user is in support of the LLM used, then great. If not, you can use the smart contracts and choose to make your own forked version with a separate model to decide winners.
Prediction Markets: if you're even slightly online then you've probably seen the hype around using Polymarket data as reference points / sources for US elections and macroeconomics (i.e. rate cuts). TMR news adds a new flavor to current PMs you can participate in.
ā¢ @indexsupply for chain indexing
ā¢ @reclaimprotocol for bringing the internet onchain
ā¢ @privy_io for silky smooth onboarding
ā¢ @base for cheap txn fees
ā¢ @OpenAI, @NYTimes
Now, what makes this tech stack so interesting?
Well, it enables analog prediction markets. Meaning, it's now possible to make open-ended bets with a clear method of market judgements to determine payouts.
Today, prediction markets are simply binary questions. Do you think Bitcoin price will be above $70k by the end of the month? Will Trump pick JD Vance to be his running mate? And so on.
Why? Because these are closed-ended questions that a single, "trusted" person in charge can check and verify. But these types of markets limit the design space of meaningful questions that can be asked.
In other words, the current yes/no model of prediction markets are just scratching the surface of what's possible.
Robin Hanson, one of the prediction market pioneers, gave a talk at the Manifest 2023 Conference in which he emphasized that most people get caught up in the flashy narratives of sports, culture, and political betting.
But true mainstream adoption will come when application level prediction markets (PM) is solved for. The example he gives is how in 2005 Google implemented internal PMs to help understand employee sentiment on a variety of topics within the company.
These application markets are like niche newsletters - they have a concentrated set of true fans that deeply care about a given subject and want to explore the depth of that topic with their peers. The purpose is not to have a large amount of liquidity and awareness but rather to focus on producing deep insights where well researched individuals can have skin in the game for their takes.
In fact, Hanson's talk was called The Search for Gold. And what he meant by that was using the concept of prediction markets as a sourcing mechanism for the most nuanced and impactful opinions at a level that 99% of people wouldn't feel comfortable putting skin in the game for.
@Sofiane, co-creator of TMR news, explains the thought process on intelligent prediction markets in his launch post:
"To unlock liquidity for the long tail of interesting questions we need to move away from the paradigm of āone question one marketā. Bootstrapping liquidity on new markets is expensive, especially for the zero sum prediction markets live today...
However, moving from markets where participants can only meaningfully engage on a small set of questions with deep liquidity to ones that reward the long tail requires markets to be able to express information in higher dimensionality than the scalar markets of today.
In short, we need markets with semantic awareness, able to judge the semantic similarity between predictions and trusted sources of truth."
Now that this model of intelligence markets is out in the open, it'll be interesting to see what spinoffs pop up in the coming months and how the builders reach the right target audiences.
The most underrated part about the TMR News launch that I think people are not talking about is the fact that "the contracts are extremely simple and versatile. They allow anyone to permissionlessly create and resolve intelligent prediction markets for any piece of future content on the internet."
My guess is that creators with prolific communities will be the best starting point to quickly test out adjacent concepts. As a creator, I can use these intelligent prediction markets in creative ways to engage my true fans. Up until now, the use case seemed boring. What's the fun in asking things like "Will YB's next post have more than 2000 words?" or "Will YB's next Zora mint have less than 10 collects by the end of the month?".
But now, there's room for engagement to get community members meaningfully involved in my creative process. For example:
One interesting example that Small Brain gave in this Seed Club interview is enabling key opinion leaders (KOLs) to source their content from avid followers. For example, let's say you set up the same NYT model on predicting Elon's next tweet. Elon goes through the responses and sees 1 that he finds funny or intelligent and posts it. Then that person gets the payout for literally creating content for Elon. Of course, this would have to be anonymous to prevent any bias in the process.
Before I wrap up today's post, I wanted to discuss two points:
As this product continues to evolve, I think it would be so cool to see a UX similar to the new Zora mobile app that I discussed in depth a few weeks ago. Instead of this being strictly on web, a more "normie-friendly" user flow would be to have them download an app, enter their e-mail, use Apple Pay to load funds, and start betting. Abstract the crypto away with embedded wallets and paymaster. Also, the daily notifications on mobile would be a no brainer to emulate "BeReal" type daily habits. And it would be really cool to make the sharing & onboarding easier by enabling a sleek iMessage / sms UI that you can send to friends and have them participate in their first market right off the bat.
In regards to the NYT headline game, there is an obvious flaw in the sense that anyone at NYT who knows what the headline is can easily game the system. Even though the the terms of service mentions that anyone affiliated with the company can't participate...how reliable is that? I'd need to think more about what possible solutions here can be but I'm guessing it involves some sort of identity / attestation verifiability (i.e. no anons, need to KYC for certain markets, etc.)
That's all for today's post, I hope all of you have a great weekend!
If you have any thoughts on this topic, I'd love to hear them in the TOC Farcaster channel.
