Gosh I shudder thinking about how many countless malicious websites I visited in high school using StumbleUpon...
Show HN: StumblingOn – Randomly find indie websites – recreating StumbleUpon
31–40 of 63 posts
Re: Show HN: StumblingOn – Randomly find indie websites – recreating StumbleUpon
#32Needs a way to flag abuse, like https://get portal.co/earlyaccess which is just some company's product with a slick sales site. First thing I got when I clicked :( Edit: another abusive link https://www.the followershop.com/ (i have broken the links to avoid any sort of SEO for them)
Good point. I have a "like" button in the history section. I will add a "report"/"bad"/"shouldn't be here" button to that as well.
Hey, just wanted to say the first site I got was https://www.the-odin.com, and I thought it was particularly random and cool. "Genetically Engineer Any Brewing or Baking Yeast to Fluoresce" was not at the top of my list for cool presents, but I thought it was super cool!
Re: Show HN: StumblingOn – Randomly find indie websites – recreating StumbleUpon
#33Re: Show HN: StumblingOn – Randomly find indie websites – recreating StumbleUpon
#34I made something similar but a bit more random. https://randomdomain.benbristow.co.uk Keeps taking a random word from an English dictionary and a random popular domain extension (com/net/org etc.), pings it and returns the first one that responds. Quite fun once you sift through the parking pages. I need to work on something to detect those.
Ah, now that really is random! One idea for sifting parking pages that comes to my mind is using image recognition. I can easily tell when something is a parking page by looking at it, I bet you could get an ML model to do so with pretty good accuracy too.
There is a class called fast.ai [1] which I personally found to be an amazing introduction to machine learning / deep learning. The first hour and a half course gets you using their library to make an image classifier that can differentiate dog and cat breeds. I bet by simply gathering training data (sites that are and are not parking pages) and reusing the same architecture used in that first class you could train a model that does a pretty good job of it.
Once you could classify pages you could just pre-compute a large list, classify, throwaway all the parking pages, and make the remainder available to your user.
Re: Show HN: StumblingOn – Randomly find indie websites – recreating StumbleUpon
#35Earlier quoted context omitted.
Good point. I have a "like" button in the history section. I will add a "report"/"bad"/"shouldn't be here" button to that as well.
Oh also for the sites marked down as bad, create a public /shame list that’s available as plain text/json and we can use it to add to block lists on our own projects.
Not a terrible idea but that could be abused just as badly. I sure hope there's a way to get yourself off this list if you are falsely accused.
Re: Show HN: StumblingOn – Randomly find indie websites – recreating StumbleUpon
#36Thank you! This is BIG! And this API endpoint is very interesting: curl -X POST "https://service.stumblingon.com/getSubmissions" -H "accept: application/json" -H "Content-Type: application/json" -d "{\"status\":0}" More on API here: https://service.stumblingon.com/docs
Thank you. I was wondering if people would find the API and do anything with it. You can use the same endpoint to check approved submissions (status: 1) and rejected submissions (status: 2) although currently there aren't very many rejected as I recently switched databases and didn't bother to transfer over rejected submissions.
Re: Show HN: StumblingOn – Randomly find indie websites – recreating StumbleUpon
#37The first site I stumbled on from this project offered to sell me instagram likes for $$.
Re: Show HN: StumblingOn – Randomly find indie websites – recreating StumbleUpon
#38Gosh I shudder thinking about how many countless malicious websites I visited in high school using StumbleUpon...
Hopefully there are zero malicious websites here, but the metrics page shows a little over 10k stumbles since submission - https://service.stumblingon.com/metrics
Re: Show HN: StumblingOn – Randomly find indie websites – recreating StumbleUpon
#39I made something similar but a bit more random. https://randomdomain.benbristow.co.uk Keeps taking a random word from an English dictionary and a random popular domain extension (com/net/org etc.), pings it and returns the first one that responds. Quite fun once you sift through the parking pages. I need to work on something to detect those.
Cool idea but the first 3 websites I get are all parking sites.
I'm pretty sure with a simple keyword check, you can eliminate most of them. A small amount of fasle positives are ok in this case I guess.
Re: Show HN: StumblingOn – Randomly find indie websites – recreating StumbleUpon
#40Thank you! This is BIG! And this API endpoint is very interesting: curl -X POST "https://service.stumblingon.com/getSubmissions" -H "accept: application/json" -H "Content-Type: application/json" -d "{\"status\":0}" More on API here: https://service.stumblingon.com/docs
Thank you. I was wondering if people would find the API and do anything with it. You can use the same endpoint to check approved submissions (status: 1) and rejected submissions (status: 2) although currently there aren't very many rejected as I recently switched databases and didn't bother to transfer over rejected submissions.
Just curious - what format are the user IDs? I thought they may have been IP addresses simply converted to an int, but if that's the case it's been done in a way I can't reverse.