demibabs

5 ms·

1.
by demibabs 3d ago
Yet if you overestimated the mileage you would’ve been none the wiser!
2.
by demibabs 3d ago
As much as I dislike flock, this only seems like an indictment on the police.
3.
by demibabs 3d ago
> TLDR: AI is really good at detecting deceptive adverts - why isn’t Google using it? I have the same question regarding “why is Workday’s resume parser so shit?”
4.
by demibabs 3d ago
You can clearly see the side effects of the models being trained on Reddit content.
5.
by demibabs 4d ago
I’m not the biggest fan of Ed Zitron, but something he said that stuck with me is: OpenAI and Anthropic spend much time warning us about “what if powerful AIs got into the wrong hands?” But it’s already in the wrong hands.
6.
by demibabs 4d ago
> Each task comes from a private production codebase that we licensed from a real-world company How does that work?
7.
by demibabs 4d ago
Doesn’t it ultimately have to be this way, to prevent saturation?
8.
by demibabs 5d ago
Can you elaborate?
9.
by demibabs 5d ago
> the reasons are so they can track who you are and sell your profile advertising. What? Like they weren’t doing this before? Obviously Google’s telemetry is tracking every link you click regardless; there’s no extra tracking benefit to
10.
by demibabs 5d ago
JavaScript is required to use Google search at all, no?
11.
by demibabs 5d ago
I think it would be depressing in its own way if this technology just ended up being regulated out of existence.
12.
by demibabs 5d ago
But what’s stopping Google from doing this already?
13.
by demibabs 5d ago
Can someone explain why this matters? Not being flippant I just don’t understand why this would be important.
14.
by demibabs 5d ago
> Generation scales with spend, review does not, and that asymmetry is the whole design problem. I see you guys have a blog post factory as well.
15.
by demibabs 5d ago
They already told the NYT that they’ve made “substantial progress” on another one of the MP Problems (most likely the Hodge Conjecture). That said, that’s probably just because of the drama miring their most recent one. After 2 I don’t see
16.
by demibabs 5d ago
> I am proud to be among the list of 25 initial signatories — all Fields Medallists — to the declaration below I wonder if there’s a Fields Medalist group chat.
17.
by demibabs 5d ago
Thanks, I was wondering why they would be using Salesforce in the cockpit
18.
by demibabs 6d ago
I still don’t understand what a “software factory” is. Can someone clue me in?
19.
by demibabs 6d ago
Good article
20.
by demibabs 6d ago
"It's just" is usually an annoying way to start a comment.
21.
by demibabs 7d ago
I’m well aware of the dispute, and OpenAI’s conduct has definitely been less than respectable. But when the central claim of the argument (that OpenAI and the researchers solved the same problem) is completely false, it’s hard to take the a
22.
by demibabs 7d ago
This article is completely mistaken. The original researchers did not solve Navier-Stokes, while OpenAI did. Ironically, this seems like the type of mistake you'd make if you used AI to do your research.
23.
by demibabs 8d ago
Isn’t that one of the most salient and straightforward argument against LLMs?
24.
by demibabs 8d ago
Yeah but the mathematicians are claiming that the key insight that made the problem tractable for AI in the first place, came from them.
25.
by demibabs 8d ago
Lol is this page written by AI?
26.
by demibabs 9d ago
Right direction to move in, but probably won’t do much. “Drug dealers are now required to give their customers the option to purchase salad instead of cocaine”.
27.
by demibabs 10d ago
Oh my god, this is unbelievably accurate. Do you practice this?
28.
by demibabs 11d ago
In the article he explains he’s not signed into Chrome nor Google during the experiment.
29.
by demibabs 11d ago
I’m working on something like this! My issue is that using Pangram for it ends up being quite expensive.
30.
by demibabs 14d ago
Engineering buildings with vibecoded software? Sign me up

More ›