[GUEST ACCESS MODE: Data is scrambled or limited to provide examples. Make requests using your API key to unlock full data. Check https://lunarcrush.ai/auth for authentication information.] [@trevposts](/creator/twitter/trevposts) "Then: - FrontierMath turned out to be easier than I realized and it turned out OpenAI had access to some of the problems beforehand. - o3 actually got released and didn't seem that much better than o1 (though it is better) - Time just passed without anything crazy happening"  [@trevposts](/creator/x/trevposts) on [X](/post/tweet/1946358217543680047) 2025-07-18 23:55:35 UTC 3177 followers, 1026 engagements "To be clear I already thought these were pretty unlikely: maybe *right* after o3 my median date for when the types of capabilities described in the AI 2027 scenario (before the "choose your ending") would arrive got as early as 2030 with XX% on 2027 or earlier"  [@trevposts](/creator/x/trevposts) on [X](/post/tweet/1946357821697847356) 2025-07-18 23:54:00 UTC 3177 followers, XXX engagements "Back for my monthly few minutes on Twitter and the main thought I want to share is that it seems like AI progress in H1 2025 was much slower than I feared after the crazy o3 benchmark numbers in December so I now put less weight on very short (e.g. AI 2027-type) timelines"  [@trevposts](/creator/x/trevposts) on [X](/post/tweet/1946355961800905018) 2025-07-18 23:46:37 UTC 3178 followers, 7858 engagements "@brandonwilson @Jsevillamol Maybe but idk we've been in the regime of enormous amounts of money going into grabbing those gains for a few years. It's clearly working to some extent e.g. o1 and o3 but still seven months later nothing THAT crazy"  [@trevposts](/creator/x/trevposts) on [X](/post/tweet/1946484273462321305) 2025-07-19 08:16:29 UTC 3177 followers, XX engagements
[GUEST ACCESS MODE: Data is scrambled or limited to provide examples. Make requests using your API key to unlock full data. Check https://lunarcrush.ai/auth for authentication information.]
@trevposts
"Then: - FrontierMath turned out to be easier than I realized and it turned out OpenAI had access to some of the problems beforehand. - o3 actually got released and didn't seem that much better than o1 (though it is better) - Time just passed without anything crazy happening"
@trevposts on X 2025-07-18 23:55:35 UTC 3177 followers, 1026 engagements
"To be clear I already thought these were pretty unlikely: maybe right after o3 my median date for when the types of capabilities described in the AI 2027 scenario (before the "choose your ending") would arrive got as early as 2030 with XX% on 2027 or earlier" @trevposts on X 2025-07-18 23:54:00 UTC 3177 followers, XXX engagements
"Back for my monthly few minutes on Twitter and the main thought I want to share is that it seems like AI progress in H1 2025 was much slower than I feared after the crazy o3 benchmark numbers in December so I now put less weight on very short (e.g. AI 2027-type) timelines" @trevposts on X 2025-07-18 23:46:37 UTC 3178 followers, 7858 engagements
"@brandonwilson @Jsevillamol Maybe but idk we've been in the regime of enormous amounts of money going into grabbing those gains for a few years. It's clearly working to some extent e.g. o1 and o3 but still seven months later nothing THAT crazy" @trevposts on X 2025-07-19 08:16:29 UTC 3177 followers, XX engagements