Hmm. There's still no way for me to turn it off in general, but I can do it per market. It is not exactly convenient for me to go through 1000+ markets and add this line of text into each of their descriptions, plus remember to do it for all new markets. But I can at least do it on markets that get lots of comments. So the "stops annoying me" part of the criteria has not been met, while the rest sort of has. Annoying edge case I didn't consider.

Gonna punt this decision for later, will see how annoying the AI continues to feel to me in practice now that this feature exists.

Ok, it's definitely still annoying me. I just have too many markets for turning it off once per market to be a significant help.

Does anyone have an argument that it should count based on the (IMO ambiguous) literal wording of the description, despite it not solving the underlying problem or satisfying the title?

And as usual we get our incorrect AI update.

@IsaacKing I was trading up to 95% after the update but before you acknowledged it based on the wording:

This resolves YES if there's a way for me to turn this off

It was a method to turn it off, so I felt confident in my aggressive YES trade at the time. It stings that my read didn't quite match the spirit you were going for.

However, it's your market, and I'm not going to make a huge ruckus over it. This is just my experience of the situation.

———

Maybe we can scrape together some automation script that will auto-apply the magic words across the board to your markets?

It is also notable that this is "OR" language.

This resolves YES if there's a way for me to turn this off, or it's removed entirely, or the AI gets smart enough that it stops making detrimental edits, or it otherwise stops annoying me.

@IsaacKing If it's still annoying for you, it's still annoying. This was super obvious to me at least.

....but unlike other traders, I only bought Yes to 28%, then did a huge amount of selling that and buying No when traders went above 50%.

Maybe I'm biased, but I was definitely reading this as "What will Isaac think about this feature at the end of the year" rather than "Did this feature exist"

I did a rudimentary vibe-code with Claude to make a script that automates the process of putting the magic words on all open markets of yours. I have not tested the code, so I have no idea if it runs right, and maybe don't run arbitrary code provided by random people on the internet, but it's the approach I want to suggest more than the actual output here. https://claude.ai/share/184ac21c-7010-4502-a28d-f31c0e410e8f

Maybe you can think of a better way to do it?

@IsaacKing explain why ai clarification is not correct pls.

According to your initial description, the solution to turn of ai exists..

@Quroe

@MyDreamIsHere2018 I am voicing the state of mind I have, and I make no claims that I am unbiased here. I have YES shares that I want to pay out, and I have a vested interest in finding a solution to this problem statement.

That being said, I do take market descriptions as literally as I possibly can.

@Quroe no bias needed , the market description clearly says: "This resolves YES if there's a way for me to turn this off".

@IsaacKing not in the market, but I feel like it unambiguously should resolve yes. “This resolves YES if there's a way for me to turn this off, or. . .” There is a way to turn it off. The talk about being annoyed came after the or.

@TonyBaloney but just decided to buy some shares because it really does seem an unambiguous yes.

@IsaacKing yeah. I’m interested in what the argument for no is. I get that this isn’t what you intended. But you can turn it off. And this was enough to resolve to yes per your rules. “In the event of one of my market titles+descriptions containing resolution criteria that clearly and unambiguously conflict with the spirit of the market, I will follow the criteria in the title+description.”

Nobody is denying that there's an "or" in that sentence. There is also an "otherwise".

If somebody says "buffalo or phoenix or other city", the word "other" means that the descriptor that follows also applies to the previous items, thus clarifying that "buffalo" and "phoenix" refer to the cities, not the animals. By the same principle of English, "otherwise stops annoying me" means that the previous list items are of the "stop annoying me" variety.

In this case, there is now a way to partially - but not entirely - turn it off, and this did not remove the annoyance. So taking the sentence literally I do not believe it has been satisfied. The market title backs this up, as the title is highly vague and subjective and is clearly about my personal impression, not objective criteria.

I do wish I had thought of this possibility and written the description to account for it more clearly, so apologies on that front. But it seems to me that both the spirit of the market and its literal criteria are pointing towards this not counting, and nobody has yet presented a better argument in favor of resolving YES.

@1bets That AI clarification was incorrect because I was asking for arguments, I had not yet made a definitive ruling on the subject.

@Quroe I appreciate the script and may use it, but it's not provided by Manifold, so it can't count for this market any more than if under the old system a third party had set up a service to automatically undo any AI edit right after it's made. And this solution will not fully remove the annoyance, because I still have to remember to add the line to any new market I make.

@IsaacKing Entirely, completely disagree with you on the English side of your argument, but I feel it would be sorta tedious to argue the point further. . . . If only there was some kind of objective(ish) machine that could somehow transform the words of your description and make a ruling on the best way to read the sentence.

@1bets I'm not really sure what this kind of comment is supposed to achieve, but I hope you can understand:

• Bullying market creators is frowned upon

• Weird comments that don't make any sense are also frowned upon

• If you keep making comments like this you'll probably lose the privilege of commenting

@Eliza trying to help topicstarter to overcome the problem, hopefully it may help with the annoying issue

@IsaacKing and I’ll stop after this, I promise. But are you saying that this was always just a vibe based market? There was never any objective criteria, as they all hinged on whether you would still be annoyed? If Manifold had a dedicated button to disable this when you were making a market, would this still resolve NO because your were annoyed by having to remember to hit the button? If the AI description stopped entirely for all new markets, would it still resolve NO if you were annoyed by it continuing on 1k older markets you have? This seems to be your argument. So why not simply make the market about whether you’ll feel annoyed by manifold? The last clause shouldn’t negate the previous clauses though. . . . I don’t know. It’s just a really, very strange reading to me. . . I generally avoid purely subjective markets, but I guess you’re saying that this is one of them?

@Eliza Yeah, already has. I take a pretty hardline stance on LLM slop, otherwise it just takes over everything.

@TonyBaloney ...No? I'm saying what I actually said, which is that the literal text has not been satisfied, and neither has the spirit, so I really see no argument to resolve YES. You've indicated you disagree, but haven't explained why.

A button on the market creation page, hmm. Would mean I don't have to go look up the exact string every time, so would be a big improvement, for sure. That might be enough.

@IsaacKing totally understand that the spirit has not been met. my point is that the wording in your description has clearly been met. Your phoenix, bufffalo example was (imo) trivial word play. If I made a market that titled “Will someone feed me tomorrow” and in the description I said, this will resolve yes if someone gives me breakfast, lunch, or dinner, or I’m otherwise not hungry tomorrow.” I think it would be very strange to then argue that me having breakfast doesn’t resolve this to yes as I was still peckish afterwards. The clause at the end doesn’t negate the other clauses. It might modify their meanings, as in your phoenix example, but only partially. Because why even include them as example of a yes resolution when the last clause is all that really matters? And if you don’t believe me, go on any LLM and paste the description and ask it for the most straightforward interpretation. It’s only a vibe based market if it all hinges on whether youre annoyed. But they clearly made a way for you to disable this feature. It’s just not a method you like.

@TonyBaloney here’s a prompt I used for ChatGPT 5.1 and Gemini 3, and both said it was a fairly unambiguous yes based on the wording of the rules. If you get the same results, I suggest reconsidering your priors:

Here are rules to a market I’m participating in:

“Manifold recently added a feature where AI can edit my market description without my consent. Naturally these edits are terrible, frequently containing incorrect information and not understanding what the market is about at all.

This resolves YES if there's a way for me to turn this off, or it's removed entirely, or the AI gets smart enough that it stops making detrimental edits, or it otherwise stops annoying me.”

If manifold includes a feature where you can disable the ai by adding a phrase in a market’s description, would this be enough to satisfy the conditions, even if it was an annoying to implement this.

Be an objective and neutral judge. Describe any ambiguity in your decision.

@TonyBaloney According to Gemini (by far the most reliable AI model, apparently):

ChatGPT says YES, Claude says lean towards YES but only 65/35. And I didn't provide any of them with the title.

@IsaacKing weird. I used the prompt I gave you in Gemini 3 pro and I got whats below. Did you change the prompt?

@TonyBaloney btw. I get the irony of suggesting that an llm should be used to resolve this particular question . . . Still they’re pretty good at assessing meaning to simple phrases. I can get Gemini to say NO, but I’ve had to really make it clear that I want it to resolve NO for it to approach your argument.

@IsaacKing but yeah, maybe we’re getting different results with the same prompt. It’s hard to avoid AI just telling you what you want to hear, despite best efforts. Maybe this is a case for that. In any case, I’m thinking I’m not going to be able to conince you on this, and it’s your market.