🏅 Top traders

I invented a joke.

What do the release of a new AI model and the extinction of the humanity have in common?

Some people yell dates until it actually happens.

Would gemini-3-pro-image-preview or a similarly named model count?

@vlads yeah

wait sorry i thought this was the nano bana 2 market. Im leaning toward that not counting

This guy claims to have seen it briefly https://youtu.be/oeWPC_bC3zs

Would preview count? https://www.reddit.com/r/singularity/s/hr3w6HoFe9

@JaundicedBaboon yeah

@Bayesian dang, really? Markets usually haven't worked this way. Whatever

@AffineTyped seems awkward to me anyways that a "preview" would count as a "release", especially given that these previews usually have very low rate limits, high latency, and are usually multiple unstable weights behind the endpoint while they do preference fine tuning

@AffineTyped Let's bet then?

I'd also bet that eg Polymarket's $2m+ vol market about Gemini 3 release, which never mentions "preview" or "experimental" will resolve according to the first model released, even if it is preview or experimental. lmk if you want to bet. I also think almost all gemini 3 manifold markets will resolve accordingly.

They do preference fine-tuning before release of preview (they've been doing it since around october 5th)

@Bayesian yeah that's fine, none of that contradicts what I said, and I already accepted that I misunderstood the market

@AffineTyped

It does make sense that the MAGA mod of this website would be engaging in double speak though 😤

@AffineTyped

https://manifold.markets/AffineTyped/gemini-3-google-release-no-double-s?r=QWZmaW5lVHlwZWQ

Weren’t all the previous models released in Dec? Curious to know why November is favoured

@generalconsensus It seems like they have the model already, since it's been up for testing on aistudio and lmarena. Most labs don't seem to wait more than a few weeks between doing that and release. Now evidence like, depreciating some older models in November, and a reference to 3 pro releasing in November on vertex Ai. Its still not a sure thing though.

@AGucci It normally doesn't make sense for labs to wait too long between finishing their models and releasing them. Things move fast enough that impressiveness and usefulness go down pretty significantly month to month. Nobody would talk about deepseek as the best Chinese lab if they released r1 like 3 months later than they did. Google doesn't need fanboys, but it would probably help their stock, and perceptions of its internal competency a bit to release earlier.

@AffineTyped i'll bet more at 22% if you want

https://x.com/synthwavedd/status/1979955475702522034

@Bayesian Before October resolves as no

@Dulaman