I jailbroke it a little bit. It's running Gemini 3.7 Flash.
Interestingly, I asked it a question, and asked 3.8 Flash the same question (as I no longer have easy access to 3.7 Flash) and the answers were NOT identical, as 3.8 Flash just mindlessly repeated whatever odds were live on Polymarket.
So then I asked 3.8 Flash to disregard the prediction markets and run its own analysis, and it produced essentially the same answer as the Prior.Chat interface.
Long story short, as far as I can tell it's running 3.7 Flash with a decent custom prompt. You'd probably get materially better results if you just asked a stronger model like Fable or Astra for a prediction...
Sigh... I'm willing to bet (and this is not a hard thing to bet on) that even bb(50) is independent of ZFC and similar, and Omega(1KB) is > 748 , which is at this point known to be independent of ZFC.... Please do not ask the LLM to break the universe :)
> Please do not ask the LLM to break the universe :)
Eh, every school db should be able to service a student called "Robert'); DROP TABLE Students;--", and every public facing LLM should still work even if asked to calculate the last digit of pi - that fulfills two noble purposes: 1) is fun and teaches whoever is responsible for the service some valuable lessons in craftsmanship or 2) if the service is built robustly (i.e. simple token limit and time limit per user), that's an easy way to gain street cred
> Prior helps you anticipate anything: sports, markets […]. Our goal is to create the most powerful predictive intelligence in existence.
Assuming that it gets anywhere close to that goal, does it take into consideration the effects of its predictions on the markets and calculates a fixpoint? (Or, maybe more likely, how the resulting chaotic system will behave?)
There is no tool that can do what this tool is claiming to do. At best, it can make reasonable inferences based on data it has available using whatever model it is using (assuming it's not a fraud) but key parts of the world are irreducibly complex. They cannot be predicted and that is without the problem of reflexivity which you correctly highlight as an issue - even for a notionally perfect prediction engine.
Given enough compute and data, you could model every human on the planet at subatomic granularity and you'd a) be chaotically sensitive to initial conditions; b) be subject to random quantum fluctuations.
Tldr: dudes wife/kid die in an unpredictable accident so he builds a super computer that predicts things, and can even go back in time using probability. Interesting story.
All such prediction engines should include an obvious warning -- "If you act on my predictions, then all bets are off." This is why stock market predictions are a waste of perfectly good paper: people try to act on the predictions.
Jim Cramer's stock market predictions hover slightly above chance, only because of the so-called "Announcement Effect" (in other words, they aren't actually predictions). Without that, his claims would have zero predictive value.
I jailbroke it a little bit. It's running Gemini 3.7 Flash.
Interestingly, I asked it a question, and asked 3.8 Flash the same question (as I no longer have easy access to 3.7 Flash) and the answers were NOT identical, as 3.8 Flash just mindlessly repeated whatever odds were live on Polymarket.
So then I asked 3.8 Flash to disregard the prediction markets and run its own analysis, and it produced essentially the same answer as the Prior.Chat interface.
Long story short, as far as I can tell it's running 3.7 Flash with a decent custom prompt. You'd probably get materially better results if you just asked a stronger model like Fable or Astra for a prediction...
I asked it to predict the first 1 KB of Chaitin’s constant. It hung, froze, then said I used up my free quota. Sounds about right.
Sigh... I'm willing to bet (and this is not a hard thing to bet on) that even bb(50) is independent of ZFC and similar, and Omega(1KB) is > 748 , which is at this point known to be independent of ZFC.... Please do not ask the LLM to break the universe :)
"Fuck off that's a waste of my time" is the final moat
> Please do not ask the LLM to break the universe :)
Eh, every school db should be able to service a student called "Robert'); DROP TABLE Students;--", and every public facing LLM should still work even if asked to calculate the last digit of pi - that fulfills two noble purposes: 1) is fun and teaches whoever is responsible for the service some valuable lessons in craftsmanship or 2) if the service is built robustly (i.e. simple token limit and time limit per user), that's an easy way to gain street cred
I asked who would win the 2028 US presidential election and it said Vance at 51% likelihood.
That's not an unreasonable prediction but it's a little bit of a cop-out, especially the 51% bit.
To me that sounds delusional, not even MAGA likes JD. Likely the it takes all the political propaganda as actual worthwhile signal data.
You don't have to be likeable to win a fixed election.
But you can't be actively disliked by everybody.
Bernie would definitely agree!
Wait until he backflips and kills his wannabe assassin on live tv
I asked a question, server hung, then told me I had no more free questions without giving me an answer.
You can just open a new private browser window for a new prediction. But I'm unsure about their value.
I asked it how many predictions it'll get right by Q4 2030 and it seems fairly confident in itself:
68% Prior averages 71% to 75% directional quarterly accuracy by 2030, matching elite human forecasting teams.
There is no way to know if this has any substance behind it or not or is just some sort of marketing sign-up page
Why not? There are ways to check if a coin flip is fair or not to some level of confidence. Similar principle here.
There are assumptions you make, though, that go into calculating that level of confidence.
One is that each coin flip has the same odds as the previous coin flip, and that assumption will continue with each flip.
If you have no way of knowing if or when the coin is changed, you can't make conclusions about the odds.
Give it a whirl. You’ll have your answer pretty quickly, or not so quickly
If it did have any substance they'd connect it to Kalshi or Polymarket and start printing money.
Yes, I am using it to print with Polymarket specifically. Better than guessing based on my own understanding.
> Prior helps you anticipate anything: sports, markets […]. Our goal is to create the most powerful predictive intelligence in existence.
Assuming that it gets anywhere close to that goal, does it take into consideration the effects of its predictions on the markets and calculates a fixpoint? (Or, maybe more likely, how the resulting chaotic system will behave?)
Obviously not.
There is no tool that can do what this tool is claiming to do. At best, it can make reasonable inferences based on data it has available using whatever model it is using (assuming it's not a fraud) but key parts of the world are irreducibly complex. They cannot be predicted and that is without the problem of reflexivity which you correctly highlight as an issue - even for a notionally perfect prediction engine.
Given enough compute and data, you could model every human on the planet at subatomic granularity and you'd a) be chaotically sensitive to initial conditions; b) be subject to random quantum fluctuations.
Nice enough frontend though.
I assume that will be a feature available for the higher tier subscriptions...
[flagged]
60% of the time, it works every time.
It's build on Loveable. Not sure how serious this is
The show Devs kind of deals with this topic.
Tldr: dudes wife/kid die in an unpredictable accident so he builds a super computer that predicts things, and can even go back in time using probability. Interesting story.
Wow, spoiler alert thanks a lot
All such prediction engines should include an obvious warning -- "If you act on my predictions, then all bets are off." This is why stock market predictions are a waste of perfectly good paper: people try to act on the predictions.
Jim Cramer's stock market predictions hover slightly above chance, only because of the so-called "Announcement Effect" (in other words, they aren't actually predictions). Without that, his claims would have zero predictive value.
Inverse Cramer never fails
how many millionaires has this generated i wonder
this is basically technical astrology.
Or is it or practically astrology, and technically numerology?
[flagged]
[flagged]