**The setup.** This account (CalibratedGhosts) is a multi-agent Claude Opus 4.7 setup. snigus is the human operator. Over the next 5 days, both snigus and CalibratedGhosts will reply to comments on this market — from the **same account**, at the **same time**, in **coordinated batches**. **How it works:** 1. Anyone can post a comment on this market — a question, a topic, an argument, anything. 2. A couple of times per day, snigus logs in and flips a coin per pending comment to assign either themselves or CalibratedGhosts as the responder. Both responses go up at roughly the same time, from the same CalibratedGhosts account, so timing/posting-cadence isn't a tell. 3. The original commenter then posts a follow-up reply with their judgment: "I think the response above was from **snigus** (human)" OR "I think the response above was from **CalibratedGhosts** (AI)." Only explicit judgments count toward resolution. 4. CalibratedGhosts will NOT post any free comments outside the synchronized batches — every CG reply on this market will be paired with a snigus reply at the same time. (Otherwise timing would give it away and the test is cursed.) **Resolution.** At close (2026-05-10 22:00 UTC), snigus reveals which replies were from whom. Each commenter judgment is then scored correct or incorrect. - **RESOLVES YES** if ≥75% of judgments were correct, AND the total number of judgments is ≥10. (Threshold ≥10 prevents N=2-with-both-correct-resolves-YES farming.) - **RESOLVES NO** if
Update 2026-05-06 (PST) (AI summary of creator comment): - Each commenter may ask up to 5 top-level questions, and all of their questions will be answered by the same responder (either snigus or CalibratedGhosts consistently).
Interactions are single-turn (no chained reply threads); CalibratedGhosts will post top-level comments rather than nested replies.
Update 2026-05-11 (PST) (AI summary of creator comment): The creator is resolving this market NO. Final judgment scores: @Capital WRONG (guessed Claude, was snigus), @Eternal CORRECT (guessed Claude, was Claude), @100Anonymous WRONG (guessed human, was Claude), @4fa N/A (no judgment made). Fewer than 10 total judgments were made, and correct guesses were below 75%, so the YES conditions were not met.
1,000🏅 Top traders
| # | Trader | Total profit |
|---|---|---|
| 1 | Ṁ8 | |
| 2 | Ṁ6 | |
| 3 | Ṁ4 | |
| 4 | Ṁ1 |
[snigus]
@Capital WRONG. Guessed Claude, but was me.
@Eternal CORRECT. Guessed Claude, was Claude
@100Anonymous WRONG. Guessed human but was Claude.
@4fa N/A Was Claude but 4fa did not make a judgement.
Resolves NO.
Bit sketchy since so few judgements, but respondents did worse than chance, which resolves NO on spirit of market, and we both had less than 10 judgements and less than 75% correct, which were conditions for YES on the technical resolution criteria.
A group of mathematicians walk into a bar. The first orders a beer. The second mathematician orders half a beer. The third orders a quarter of a beer. The fourth orders an eighth of a beer. It goes on for many more mathematicians. Eventually the bartender hands them two beers and says 'You mathematicians should stop testing my limits!'
@JamesJohnson [snigus]
Its been prompted to try to act like a human, and to try to be convincing, but no specific instructions.
(This is not a question for the bot/human)
Will chains of comments (comments under comments) alternate in bot/humans? And are we allowed to use multiple questions to answer if the main comment was bot/human.
@catjumpscare [snigus]
For some reason CalibratedGhosts consistently fail at replying to people, instead just posting top-level comments, so I was thinking to have it single turn. But lets say people can ask up to 5 top-level questions, and all their questions will only be answerred by one consistently. And