rameerez

Look at the commit history.

Out of 557 total commits in the repo, 510 have been done before the past 2 weeks. All those (minus 5 in 2023-2024) have been done on or before July 2022, months before ChatGPT had even launched.

Out of the 47 commits in the last 2 weeks, 26 were README updates / CI / docs. The remaining 21 commits are clearly cleanups, speedups, bugfixes, and tangential features like Blender import/export. Which leaves us with 505/557 commits (90%) if we're not generous --or 531/557 commits (95%) in the best case-- of non-AI commits to the repo.

OP clearly wrote most of the project by hand and has just been cleaning things up for public release the last couple weeks. Exactly as he disclosed in his comment.

Loads just fine?

It works just fine?

OP here. As others have said, it loads immediately for me (tested on desktop + on mobile data + incognito)

The entire site is cached + Cloudflare sits on top of everything. I just ran a couple performance tests under the current HN traffic (~120 concurrent visitors) and everything looks good, all loads under 1 second. The server is quite happy at an average load of 0.06 right now, not even close to start breaking a sweat.

Turns out you can get off the cloud and hit the frontpage of HN and your site will be alright.

I'm the thread OP – can confirm we're using Stripe Checkout and all CC processing happens in a fully-managed Stripe Checkout page.

We're not using any other API from Stripe other than that required to launch the hosted Checkout page.

I'm the OP of the Twitter thread – I've had the exact same experience: unrealistically low risk scores for most fraudulent transactions. There were plenty of red flags for each of them (400+ cards and 40+ names under one single IP, most payments got already flagged for credit card testing fraud early on before succeeding after many tries...) Even dumb heuristics would have blocked 90% of the fraudulent payments. I appreciate Stripe is fixing this quickly after making it public and refunding fees, but something is definitely wrong with their risk calculation algorithm.

Fixed! Thanks for the heads up. The cache was interfering with pagination in category pages.

Where?

Update – just fixed this!

Thanks for the heads up! It may be that the filtered query yields no results and I'm showing a blank page instead of a message. I need to add a nice info message like "No results – please remove some filters" or something along these lines to make it clearer.

Try with a common query word like "cat" – filters work fine for me with all models!

I'm not sure these models are good at generating great logos yet, but searching for "logo design" returns a few good ones for me https://prompthero.com/search?q=logo+design

Search anything – the model selection filter is at the top right

Thanks! It's Ruby on Rails with some vanilla JS here and there when I need it.

Hi guys!

Over the past two weeks I've been tinkering with NFTs and found out that most of them do not store their images on the blockchain: they're fully editable after mint. I thought this was against what most people believe NFTs to be, so I built a site to explore this problem. You can enter any ENS wallet or any Opensea NFT link and it will analyze it to tell you if the NFTs are immutable or not.

I also wrote a blog post [1] explaining the problem in a bit more detail.

Happy to answer any questions and hear your feedback!

[1] https://rameerez.com/no-your-nft-is-not-on-the-blockchain/

What an extremely talented individual you are. Thanks for sharing your work and your process in the YouTube videos.

As someone who comes from a web dev background and who's completely unfamiliar with developing large projects in C, I'm curious about your setup. What does it look like? In terms of what kind of IDE, OS, interesting tooling, etc. you use on a regular basis to work on this.

Also, I've seen the first commit goes back to Oct 2017. Did you work on this full time during the whole time / how many full-time months you would say you devoted to this project? Also, if I may ask – what do you do for a living?

Thanks again and congrats on the project!

Hi guys!

I was talking with a friend the other day [1] and he shared this idea, I think it comes from a Veritasium video [2], that if you took 1.000 random people and ranked them by their overall success, the top ones would almost always be extremely lucky (i.e.: you get to the top by being lucky, not by hard work alone)

Although the math behind it is extremely simple (just a weighted sum), as well as the code [3], I think the idea that hard work alone doesn't guarantee success is a powerful one. Plus, I think the app is fun to play with.

Hope this thought experiment helps reflect about the whole idea of randomness in life, and maybe help spark a few interesting conversations with friends! It'd be great to hear your feedback.

[1] https://twitter.com/rameerez/status/1334104302361702400 [2] https://www.youtube.com/watch?v=3LopI4YeC4I [3] https://github.com/rameerez/luck

Yes, but not Transformer-based like these two, rather LSTM-based like ULMFiT

A bug in the scoring algorithm that I need to fix

Thank you for these words :)

I'm most active on Twitter :) https://twitter.com/rameerez

Thanks for the heads up!

I'm most probably publishing a paper later this year detailing the process. Also, 80 pages of my thesis would have loved this wouldn't have involved AI to make the whole thing simpler :)

Some friends run an AI bootcamp and helped me finding the initial set of users to help me with labelling. Initial labelled data was generated mostly through them, both manually labelling and with the approach described in https://useguard.com/experiment Also, the model I'm using relies heavily in transfer learning and achieves very reasonable results with few labelled items (the paper in which the technique is described actually maintains that with only 100 labelled examples they reached comparable results to using 10x that data in models that use older approaches)

Yes, definitely. I mention it in Guard's own privacy policy, I don't like using it either, but reasons are: (a) it's the simplest and as far as I know one of the few free analytics tools available, (b) not having a measure of the website activity will make me effectively blind and unable to make decisions, (c) I don't send any personally identifiable event (and, for this matter, I don't send any events apart from page loaded events). I'm also open to suggestions to replace GA.

The scoring definitely needs some work – I think some factors should have more weight. Also services' data vary a lot so it's difficult to come up with a good measure for everyone. Ex: I try to take into account whether the service has had any recent data breach, so it penalizes a lot if it has but also scores rather low if it hasn't; privacy policies' length vary wildly and I think that also plays a large role... It needs some tweaking but I think with some improvements I'll reach a more accurate scoring

Yes, initially it will all draw from the same dataset, so a user in theory could definitely see all services' snippets. But, to increase statistical significance in the data it gathers, I've restricted the initial amount of items in the test so right now this will not be the case (otherwise, I'd be dealing with circa 3,500,000,000 different pairwise comparisons hehe)

Thank you! :) I've read some recent research and looks like this is actually measured: only 0.001% of all internet users start reading them (and even a smaller amount of people likely finish reading them). On top of it, if you had to read all the privacy policies you accepted only on the past 5 years alone, you would have to use 3.040 hours of non-stop reading. Crazy. Love your privacy-oriented username btw! ;)

This was actually one of the ideas to evolve the project! Another one is making an app that protects your digital privacy from these threats, kinda like an antivirus but for privacy threats instead of viruses (https://useguard.com/blog/future/) Would love to hear feedback on what should this project become next :)

It should be pretty close to 100%, though, they seem to be super privacy friendly!

Bug, one of the components of the score is not properly normalized I think. I'll fix it as soon as I handle the traffic overload :)