秋田预测市场 · Discord 审计

Anthropic Mythos市场规则变更引发结算争议

2026-06-10 · Polymarket · 43 条相关讨论

residual 2026-06-09 20:25:36

earlier the rules said: Products labeled as Claude Haiku, Sonnet, and Opus 4.7/5.0 or similar will not count for this market's resolution unless they are confirmed to be the same model referenced in the leak. and now they added more context saying: Any model whose official name includes “Mythos,” or that Anthropic officially describes as a “Mythos-class” model or similar, will qualify towards this market’s resolution. now this means any model not mythos (the actual one that was leaked) just being a mythos-class (they most prob wont launch the actual mythos model that was leaked before) can be used to resolve to Y? this is a significant change in the rules.

earlier the rules said: Products labeled as Claude Haiku, Sonnet, and Opus 4.7/5.0 or similar will not count for this market's resolution unless they are confirmed to be the same model referenced in the leak. and now they added more context saying: Any model whose official name includes “Mythos,” or that Anthropic officially describes as a “Mythos-class” model or similar, will qualify towards this market’s resolution. now this means any model not mythos (the actual one that was leaked) just being a mythos-class (they most prob wont launch the actual mythos model that was leaked before) can be used to resolve to Y? this is a significant change in the rules.

epsilon 2026-06-09 20:55:12

This was a fine if not a good clarif

This was a fine if not a good clarif

epsilon 2026-06-09 20:56:03

Yeah one of the leaks literally says it’s the exact same model but w/ safe guards

Yeah one of the leaks literally says it’s the exact same model but w/ safe guards

epsilon 2026-06-09 20:57:23

There was always a chance that counted given they said the “model referenced in the leak” which is Mythos and made clear that things which were not called Mythos could count. It was just unclear if Anthropic describing something as “Mythos class” met the requirement of being “the thing mentioned in the leak”

There was always a chance that counted given they said the “model referenced in the leak” which is Mythos and made clear that things which were not called Mythos could count. It was just unclear if Anthropic describing something as “Mythos class” met the requirement of being “the thing mentioned in the leak”

epsilon 2026-06-09 20:58:22

The rules are explicit about models not named mythos possibly counting

The rules are explicit about models not named mythos possibly counting

cambridgeisaac 2026-06-09 21:01:45

why did mythos release for today just drop so much?

why did mythos release for today just drop so much?

nsc 2026-06-09 21:07:02

its normal on the day off, usually retards start blasting

its normal on the day off, usually retards start blasting

TakenOver 2026-06-09 21:19:20

Yeah welcome to PM. They did that for MTSR and Trump saying Iran too

Yeah welcome to PM. They did that for MTSR and Trump saying Iran too

TakenOver 2026-06-09 21:19:28

Clarifications can very well be just changes to rules

Clarifications can very well be just changes to rules

TakenOver 2026-06-09 21:20:36

This is one of the less bad clarifications, despite me being a yes holder

This is one of the less bad clarifications, despite me being a yes holder

nsc 2026-06-09 21:24:54

residual put it so well. "The same model referenced". Now they will resolve P2 on a diff model while Mythos will stay unreleased, it is very stupid.

residual put it so well. "The same model referenced". Now they will resolve P2 on a diff model while Mythos will stay unreleased, it is very stupid.

epsilon 2026-06-09 21:28:44

It’s literally the same model w/ guard rails

It’s literally the same model w/ guard rails

epsilon 2026-06-09 21:28:50

they just are calling it something else

they just are calling it something else

epsilon 2026-06-09 21:29:38

and the rules literally had a carve out for this anyways!

and the rules literally had a carve out for this anyways!

epsilon 2026-06-09 21:30:32

<:ICANT:1306491472417263636>

<:ICANT:1306491472417263636>

epsilon 2026-06-09 21:32:31

If mythos preview & mythos 5, or any Mythos X variant would count as well as some “non mythos” model that is confirmed to be mythos preview, it makes a LOT of sense that any differently named but “mythos class” model should count. (one could very reasonably argue this is the intent of the rules, albeit not explicitly listed)

If mythos preview & mythos 5, or any Mythos X variant would count as well as some “non mythos” model that is confirmed to be mythos preview, it makes a LOT of sense that any differently named but “mythos class” model should count. (one could very reasonably argue this is the intent of the rules, albeit not explicitly listed)

residual 2026-06-09 21:35:45

that means it is not the same model referenced in the leak. the model referenced in the leak was said to have exceptional ability to find and exploit zero-day vulnerabilities and they planned to release it to a closed set of clients (proj glass wing). the model they release publicly today or later wont be this model but it will still qualify as per the "NEW" rules.

that means it is not the same model referenced in the leak. the model referenced in the leak was said to have exceptional ability to find and exploit zero-day vulnerabilities and they planned to release it to a closed set of clients (proj glass wing). the model they release publicly today or later wont be this model but it will still qualify as per the "NEW" rules.

epsilon 2026-06-09 21:36:06

Ok but “Mythos X” would count and would not be model referenced in the leak either!

Ok but “Mythos X” would count and would not be model referenced in the leak either!

residual 2026-06-09 21:38:05

yeah but now even if its not named "Mythos" and not the referenced model it still counts!

yeah but now even if its not named "Mythos" and not the referenced model it still counts!

epsilon 2026-06-09 21:38:30

My argument is it is a natural extension of the rules that a mythos class model counts.

My argument is it is a natural extension of the rules that a mythos class model counts.