Isn’t Manus just based on Claude plus browser-use?
I can’t get beyond the first few statement knowing that Manus is just a wrapper around “American counterparts”
Here's the thing - we failed.
Why? Three things.
- service providers like Expedia, Lastminute.com, Zwillow did the data integration by hand. They were able to afford to do this (the bit we didn't expect) because they were able to attract adverts and charge a transaction fee. They had a business model.
- Specific applications like the HMRC self-assessment were able to offer a user experience that gave close control and easy completion. The agents we built were able to do easy completion but they took control off the user and could not perform at 100%. Users were not confident and so didn't use them.
- We did most of the work pre iPhone, so we didn't have the right interface. Of course that has changed... but the Appstore also changed things and user expectations are sky high now.
It was also fersum difficult to program proper MAS. The issue is creating a system based on the expectation of behavior of many other systems which is of course a combinatorial mind bender. We mitigated this a bit with ideas from economics and game theory, but I think we were a long way off solving it.
LLM's do provide extra capability for solving some of these problems, and you young people are so clever and hard working that maybe you can just blitz through some of the others. I can't get at Manus yet, but I don't really expect that it will be a proof of this.
A couple of years ago, every company added a shitty chat bot to their website. Imagine that without adding a ton of complicated integrations and writing a ton of rules, you can make this chat bot actually become somewhat useful. That would take a ton of work away from customer contact centers.
Imagine that software testing can be as simple as writing: "Create 50 concurrent users that are signing up to my Android app." In stead of writing a shit ton of tests in Robot Framework that need to be updated for every change of the code.
There's some real benefit in AI Agents, even in the short term... But it might not be for end users.
And all those cooperation, that are supposed to buy AI labor- they know that those mafia companies trying to get in via subscription, will be the worst union of all. Software eating the worlds bread butter indeed.Subscription is up, and its now (costOfEarnings -1$)
People can see the takeover and the faces of it for what they really are and boy is it ugly.
a more perfect union indeed.
Maybe the foundational companies follow suit and make startups like this obsolete?
It makes we wonder now if what we really have is a creativity deficit.
Meanwhile I'd pretty much spend the time on the beach.
In all seriousness if such a system were to exist, why wouldn't the creator simply use it to make great returns for themself instead of scraps for everyone else?
Ignore the first sentence in Tyler Cowen’s post here, where he asserts that Manus is ‘for real, and ahead of its American counterparts.’ That’s a rather silly way of summarizing the situation, given everything we now know.