What will Kamala Harris tweet about before the election?
This time calculated by text embeddings, instead of Levenshtein distance, to better capture the meaning, instead of the specific words, of the tweet
I will:
take all tweets from @VP made before the election
clean them (e.g. removing hashtags and links from the end of the post)
calculate their text embeddings
calculate the text embeddings of all the options submitted in this market
resolve all options that closely match a real tweet as YES
How do we define a "close match"?
Refer to the following document:
https://docs.google.com/spreadsheets/d/16VT92CjXOdQDlIrl2IPwgi9HxrF6PRxIrza720IcOx4/edit?usp=sharing
If the embedding distance is LOWER than the distance between ANY real Kamala tweet and hypothetical alternative, it is deemed a close match.
Important caveats:
Only text written by the user "@VP" counts, for example a repost without any additional text is not eligible. Replies count. Any other types of posts count as long as there's text written by "@VP".
Text in images/videos/sound doesn't count.
If the account is hacked or Kamala didn't write the post himself, that doesn't matter as long as the post is created by "@VP".
If Kamala changes accounts, I will also start tracking her new official account
If the post is deleted before this market is resolved, it becomes not eligible.
I will use whichever publicly available embedder is at the top of the MTEB leaderboard at resolution time
I might add more examples to the reference sheet in the first week after market creation
If you want to include a longer answer that doesn't fit in the manifold answer field, create an option and leave the full answer as comment
To use this market:
think about the possible ideological space that Kamala might occupy. Will she attack Musk/Tesla? Will she back Palestine? Will she talk about reviving industry?
ask gpt/claude to generate plausible tweets in her voice on the topic
add them as a new option
See also the Trump market
1,000
1.00