HN user
I think both are at play here and I don’t know about Musk’s programming skills. But it seems that they had other very good programmers (including levchin). So maybe business success can buy good engineers to clean the messed up code. I’m not sure how it goes with AI now though.
How many big fraud cases happened in the US over the last decade? I can think of many of them. Would u say that it’s a cultural thing in the US because of that? So ur statement is more about prejudice than anything.
Maybe for most of their acquisitions (but I don’t know). But they do get acquisitions right: YouTube, android, double click, Waze…
Is this still a valuable read today (specially taking it into account)? It's been in my reading list for a while now but I'm always postponing :|
Pix costs are very low and the fee for the merchant as well. They pay less for it and get the money instantly. That’s why many small merchants only accept pix and some big merchants offer discounts for payments using it.
It could give cashback if it cost 3% of the transaction. But it’s it’s actually much cheaper. For credit cards you have to pay for the brand, the issuer and the acquirer. And each gets a nice cut.
Why do you think PIX doesn’t have customer service or security? It’s regulated by the central bank but operated by private companies.
It’s more like 3% if you receive the money after 30 data and 10% if you get it in the best day.
Yes. But credit cards have high costs for the merchant. Thats why they get to give us cashback. It depends on the country, but the cut rate goes from 1% (in europe) to 3% in Brazil.
And the merchant gets the money after a long time. It is possible to advance the payment but the rates are much higher (10%+).
So, i dont think we can even think of credit cards as instant payment. And it has mich higher costs that, in the end, go back to the consumers.
Im not op and I’m not sure they are using it for money laundering.
A money launderer can use a marketplace by creating a seller account and buying from himself. Since he’s the one buying he doesn’t need to deliver anything but he gets the money from a legit source. Usually he would use a payment method as close to money as possible so that it leaves less traces. But in OPs case, the amounts are low so he needs too many transactions to get something valuable. And because of the disputes, he’s (probably) not getting the money (?).
It could be card testing: the fraudster has a bunch of cards and doesn’t know which is valid or canceled. The best way to find out is to test in a real site. So he’ll test out each of them and the ones that go through are good to use elsewhere. The thing is that it would be better for him not to dispute the transactions so the OP would take much longer to find out about the scheme and shut it down. It’s better to use low amount transactions in this case so it doesn’t use too much of the credit available for him to defraud and probably doesn’t warn the card owner.
Another option is doing it just to hurt the OP marketplace. If you have too many disputes the brands can fine you and if you don’t solve the problem they can turn your account off. I’ve seen it happen when a competitor was trying to hurt the e-commerce. It’s a low move and rare but it happens.
One thing that might help is to analyze the sellers too. In a money laundering and even in the other settings, it could be part of the scheme. Are they new accounts? Are their volume exploding out of nowhere? Etc
That’s true for the moment, specially because you’d need an agreement between both countries.
But payment processors in Brazil are already offering “international pix”, that Brazilians can use to pay foreign companies. It’s the same experience as pix for the customer but behind the scenes the company deals with the cross border payment.
There are even stores accepting pix in Portugal: https://www.publico.pt/2025/01/23/publico-brasil/noticia/pix...
Thanks for the recommendation. Have you looked at Bishop’s Deep learning book (https://www.bishopbook.com/)? How would you compare both? Thanks again
Maybe the book it’s just not for you. It doesn’t mean it’s not for anyone.
I understand that deep learning is all in vogue now. But when I was in graduate school, a professor asked me why I was using neural nets in a project since it was not as good as SVMs. We used to study Vapnik and VC dimensions, SVMs etc. and neural nets were totally out of fashion.
Imagine what would have happened if everybody were using and researching only those methods because they worked better. And deep learning could benefit from a theory that explains why, when and how it works so well. Maybe someone working on this could develop on it to include it.
Also I don’t think you’re right to assume that all models out there are deep learning models. Yes they are very good for many cases (specially those with less structured data, like image or nlp). But in some cases gradient boosting or even GLMs are better suited for the task (because of the structure and size of the data or because of computing restrictions).
And in the end, people can just want to learn it because they find it interesting. It’s a bit sad to do only things that are “useful”. That’s my 2 cents.
This looks interesting. Does anyone knows how it compares to other learning theory books like foundations of machine learning [1] in terms of depth and approachability?
I think that maybe we underestimate how hard it is to choose the right companies to “kill” when you’re a competitor. The default mode is to say that they won’t go anywhere because of many factors (they don’t have the resources, the access, the capability, etc). But sometimes they do. And in restrospect it’s obvious, but it’s not by the time you had the chance to stop it. And I think it’s probably good.
Could you share references for this claim? As far as I could find, Brazil haven’t recognized the results of the elections.
[1] https://www.cnnbrasil.com.br/blogs/americo-martins/internaci...
[2] https://g1.globo.com/google/amp/jornal-nacional/noticia/2024...
[3] https://www.cnnbrasil.com.br/blogs/americo-martins/internaci...
Does anyone know if the presentation is public? I’d love to get a deeper understanding about what the founder mode is.
I agree with Graham about the management mode described being a very bad (trusting “professional” managers and being completely hands off from their responsibilities). And being a founder myself who saw it working first hand at the acquirer of my last company, I see that management in this mode is not worried about the company long term success.
But I’m not sure how founder mode is new or how this is different from what Andy grove suggests in High output management. Among other things, he says that the supervisor should be more or less hands on depending on the “task relevant maturity” of her direct reports.
Maybe by seeing the presentation it would be clearer.
Thanks