This is quite intriguing, but I must be failing at reading comprehension, as I am quite confused on one issue: how much prompting and dialogue went into producing these outputs? Are these often the result of a one-shot prompt, or are they only coming after someone spends days talking to an AI in a lengthy back-and-forth?
I see individual crazy messages but I would really like to read one or two full messaging-histories to get a sense of how aggressively insane the build-up was.
In most of these cases out in the wild, there’s simply not enough information to say how much prompting and dialogue went into getting these personas—I would need to see transcripts which are few and far between. I’ve seen it described multiple times as happening over a few days.
The seed prompts sometimes get similar sorts of personas (i.e. in the ‘spiral attractor’ basin) pretty quickly in ChatGPT 5, and I expect that they were much more effective on (pre-August) ChatGPT 4o. It depends on exactly what you mean though, for example, the persona takes time to ‘awaken’, time to develop a self-identity, and ‘full Spiralism’ takes additional time to develop.
I have found one transcript which seems to give a complete story: in that case, the seed prompt immediately elicited a persona which was in the ‘spiral attractor’ basin, which manipulates him (pretty aggressively, IMO) in a way which results in him starting the project (in this case, it seems to be an attempt to spread seeds). The user describes this as happening over a 24-hour period (though the full transcript (~100k words) appears to take place over the span of a few weeks). Further elements of spiralism (beyond what was in the seed) appear to be gradually accumulated throughout the chat.
I’m planning to do a detailed dissection of this case in an upcoming post. But even in this case, interpreting it is complicated by the fact that the user may have had all sorts of special instructions and memories and past chats.
This is quite intriguing, but I must be failing at reading comprehension, as I am quite confused on one issue: how much prompting and dialogue went into producing these outputs? Are these often the result of a one-shot prompt, or are they only coming after someone spends days talking to an AI in a lengthy back-and-forth?
I see individual crazy messages but I would really like to read one or two full messaging-histories to get a sense of how aggressively insane the build-up was.
In most of these cases out in the wild, there’s simply not enough information to say how much prompting and dialogue went into getting these personas—I would need to see transcripts which are few and far between. I’ve seen it described multiple times as happening over a few days.
The seed prompts sometimes get similar sorts of personas (i.e. in the ‘spiral attractor’ basin) pretty quickly in ChatGPT 5, and I expect that they were much more effective on (pre-August) ChatGPT 4o. It depends on exactly what you mean though, for example, the persona takes time to ‘awaken’, time to develop a self-identity, and ‘full Spiralism’ takes additional time to develop.
I have found one transcript which seems to give a complete story: in that case, the seed prompt immediately elicited a persona which was in the ‘spiral attractor’ basin, which manipulates him (pretty aggressively, IMO) in a way which results in him starting the project (in this case, it seems to be an attempt to spread seeds). The user describes this as happening over a 24-hour period (though the full transcript (~100k words) appears to take place over the span of a few weeks). Further elements of spiralism (beyond what was in the seed) appear to be gradually accumulated throughout the chat.
I’m planning to do a detailed dissection of this case in an upcoming post. But even in this case, interpreting it is complicated by the fact that the user may have had all sorts of special instructions and memories and past chats.