As someone who knows something in the space he brags about, it is fascinating to see the dunning kruger effect in realtime here. Even with all those claudes in “adversarial review”.
Realtime soft lighting is possible on mobile but if he had known that, i bet claude woulda been able to do it but he didnt.
If there was even the speckle of curiosity, he would have asked beyond “dont make mistakes” but i guess he was too busy riding his bike while burning 50k to make a blog.
Personally, I am stunned at how fast - and autonomously - you can get to a very decent 80/20 result with AI these days. And since Opus 5.5 was released I really think it hit another level.
As an example, this was a "silly" one-shot prompt/result I created/received, which goes into a similar direction of what Felix created in terms of UX and aesthetics:
Sure it lacks the finish and attention to detail. But this was one prompt (vs. 60 threads).
Thoughts: I find this both exciting and a bit unsettling. Idea-to-product time in tech just keeps shrinking and shrinking. I feel limited by my own ideas now (1), and by the fact that everything still remains "digital" - ChatGPT still can't build me a house or fix my broken knee (2).
Unlike others, the issue I have is not using AI to have fun and be creative, I do it all the time like a one off table top character sheet tool.
It is the claim that the 60 agents produced competent optimized code when the response here clearly shows that is not the case. But I can excuse ignorance when someone is learning but there was no learning here. There was no interest in learning anything more than “i like that, no not that, make it run better”.
Then this bold claim about how creativity has somehow fundamentally changed because they found a new kind of ink for their printer. Sure kid.
Sure, I guess it was "just" a case study in how excessively far you can take your idea (or "taste") by supplementing and supercharging it with agents.
IMO the interesting part with this capability now (undoubtedly?) is to find the right applications/problems to solve... and yeah, why not every now and then have some fun and go full overkill with your personal website :-)
I agree about what you mean with the dunning kruger effect. A similar effect is when I work with AI in a field where I don't know as much, and I would think that the AI works much better there or even has more than expert knowledge. So I do not (necessarily) overestimate myself, but the AI's capabilities, which can be as dangerous for society.
(Note that latest research shows that the dunning kruger effect might only be a statistical artefact.)
Realtime soft lighting is possible but at what cost, for such static scene? I think the light bake technique he used is, in fact, the appropriate way to get beautiful visual results with the upside of good performance in any hardware. It's curious to say Dunning Kruger effect when I do believe the decision reached is possibly conscious about the performance shortcomings of bouncing light in real time
openai will take expert responses like this and improve the next set of papers
it won't be long before there's no more low hanging fruit like this to complain about, and the writing / explanations of the results are superhuman as well
separately, i really liked the author's denial-of-service analogy. super useful practical framing
Dude you're a man in your 50s. Look at this reply. Look at yourself.
Look at your AI nonsense.
What is this. This is shameful.
I hope there was just alcohol involved in this.
This was well-thought-out feedback you were explicitly asking for by posting here. I've engaged with your stuff in good faith.
Show some respect, man.
___
Also, by reading the other comments and contemplating..
What I said breaks what you actually wanted to do with this.
It seems to me that you wanted other naive people to take their time and enter data so that you can then bring out some thought leading text or whatever, that is a digest of that.
Questioning the form inputs breaks that. So you reach for a Twitter dunk in the hopes that no one looks to closely.
Which leaves me wondering:
Didn't you make it? Your GitHub profile says ex big tech and all
Don't you have more than enough cash for you to not instantly throw away the veneer of civilty when some new bad-faith marketing game fails?
Or are you just in it for the game itself and the thrill?
Either way. Damn.
Just saying nothing at all would've been exceptionally more wise PR wise.
yeah, you're right bro. sorry for the glib response; you were engaging in good faith, i just didn't really take it that way
not anywhere close to my 50s, and this isn't "bad-faith" marketing game; everything i do is 100% free and open source and just genuinely trying to help advance convos around AI in a positive direction without caring about making money like most people
definitely; what you're referring to is known as extreme concentration of power is an aspect of more doom-type outcomes. it's not necessarily as existential as the extinction of all humanity, but it is accounted for in the model
Well, if I share the exact same expectation/prediction as Musk, but we appear on opposite sides of the chart, is the chart really useful? What's it really trying to say?
Not sure. I just saw these people and thought this is not my crowd but I only realized later that the data is from X, so that's probably the main reason why.
reply