Tyler Cowen links to Manus, with sources that claim it may be "another DeepSeek moment." Is it that?
https://marginalrevolution.com/marginalrevolution/2025/03/manus-assorted-links.html
If the answer is obvious this will resolve to YES or NO at that time.
If the answer to this question is non-obvious: This resolves based on a Twitter poll that @thezvi will run, with the wording "Is Manus another DeepSeek moment?" and the choices being Yes and No, with or without options for See Results or Unclear/Maybe, and it goes to YES or NO based on plurality between those two choices.
No tears. I will NOT overrule the poll results even if they're 51/49 and I disagree.
1,000Looks like Manus is apparently just a Claude Sonnet wrapper, so this is almost certainly a NO to me: https://x.com/jianxliao/status/1898861051183349870
I checked out the demo, and it’s decent, but it definitely doesn’t feel like a “DeepSeek moment.” What made DeepSeek special was how it challenged the assumptions of many people who’ve been involved with AI for years. Very few would have predicted that a Chinese company could train a model
At around 5% of the cost that everyone thought would be necessary.
That could seriously compete with models released by all the established players.
With Manus, I’m not sure what assumptions I’m supposed to be updating. Agents have been a hot topic for a long time, and it’s no surprise that more will keep coming out. Nothing unexpected there, in my opinion.
@Aleph yeah, deepseek was just such an outlier that it seems unlikely this could be considered comparable (unless it really blows up)
No idea whether it will be described as another DeepSeek moment, but it looks great. If it genuinely outperforms OpenAI's Deep Research and is so much more flexible and autonomous with regard to computer usage, I think it could be a huge jump forward. As a programming hobbyist, there are definitely a lot of projects beyond my ability that I wanted to do but couldn't because existing frontier reasoning models didn't have enough context; this seems like it might get around that issue.
@ProjectVictory faces doubts due to limited availability.
My bet is it's real and good and interesting.
Whether people will describe it as a DeepSeek moment, IDK.
The demos looks really impressive! https://x.com/manusai_hq/status/1897742115490480288?s=46
To me, a "DeepSeek moment" represents something that fundamentally changed how we viewed both computing requirements and how companies could gate SOTA performance. I don't see Manus having that style of impact. While it might be as impactful (I don't think this, but I'm open to the possibility), I don't think it will be comparable in the same way that will cause people to vote it another "DeepSeek moment"