2026 really was a year where agents began demonstrating their ability to conjure up something out of nothing, so I would implore people on Bluesky and Mastodon to earnestly consider the possibility that 2027 is going to be a bonkers year in terms of AI capabilities. https://blog.aifutures.org/p/q35-timelines-update-looks-like-the
0 Boosts
0 Likes
@paul Based on my day to day experience over the last 3+ years. 🙂
@mergesort @paul Same. But also even if we don’t get quite to that metric, stuff is going to keep improving assuming you look back even 6 months; the rate of improvement is something I couldn’t even imagine a year ago.
@kyleve @mergesort it has improved, it’s useful now. Saying its going to keep improving at the same rate or even a faster rate is where I have problems with. Stuff is too new and too complex to be able to do any kind of prediction of where its going to be in 6-12 months. Saying its going to take over all white color jobs next year makes it sound like someone has drunk way too much koolaid.
@mergesort I’m talking about the article you posted, they make their conclusion on data that I’d call questionable mostly because there’s so much incentive to tell a huge growth story.
@paul These aren’t numbers from Anthropic, they’re from the independent and well-respected organization METR. They were embedded within Anthropic for a few months and were given full independence, ultimately publishing a report on what they observed.
@mergesort based on cherry picked data from an AI company going public any day now.