Hacker News: JawsOfALion

New comment by JawsOfALion in "Visual Basic 6 rebuilt in C# – complete with form designer and IDE in browser"

JawsOfALion — Mon, 11 Nov 2024 11:24:04 +0000

"And why is this maniac putting everything into the browser that does not belong there? Is the reason putting there is just to have it there?"

Haha, relax a bit. If it wasn't a browser application i wouldn't have tried it at all (most people here would not have either). I see it more of a toy than a development environment - but the fact that it seems fully functional and in the browser is impressive to me.

Anyways, 99.99% anyone using VB6 today is doing it for nostalgic purposes, not to develop real useful code - and that's probably what the project had in mind. Just a fun project.

If you did want an accurate no-changes recreation of vb6 on the desktop, just download the original vb6.

New comment by JawsOfALion in "Visual Basic 6 rebuilt in C# – complete with form designer and IDE in browser"

JawsOfALion — Mon, 11 Nov 2024 11:14:51 +0000

Lol, that's amazing, as soon as it loaded up i was laughing at how accurate it was . So many memories from simpler times coding as a kid.

New comment by JawsOfALion in "IMG_0416"

JawsOfALion — Mon, 11 Nov 2024 11:01:34 +0000

I was going to explain the way i saw it, but i erased it and decided it's probably best not to give my thoughts in case he or someone in his life came across the comments out of respect.

New comment by JawsOfALion in "Ask HN: Local Image Classification in the Browser?"

JawsOfALion — Sun, 10 Nov 2024 11:44:58 +0000

Thanks for sharing this project, it indeed shares much similarity with what i want to do with it, so i'll have a look. But it should be possible to achieve the same thing on chrome - but it might not be a straight port (my chrome extension can work at hiding images based on the classification i get from the tensorflow mobilenet, but the accuracy is so poor currently it's not worth using)

New comment by JawsOfALion in "OpenAI reportedly developing new strategies to deal with AI improvement slowdown"

JawsOfALion — Sun, 10 Nov 2024 11:25:15 +0000

So is the age of raising at a 5B valuation on a few month old ai startup with a silly name, that has nothing built, not even a product demo, but just founder name recognition, ending? How much longer before the AI gold rush is over, or even a pop happens, I wonder.

New comment by JawsOfALion in "Ask HN: Where to put a static page that would last forever"

JawsOfALion — Sun, 10 Nov 2024 11:12:46 +0000

If he's hosting just some textfiles (i.e. a few KB), he won't have any aws bills that he would need to worry about prepaying. (I have a static s3 website, low traffic and low storage size, and i have not paid anything on it for years).

At least that's the case today, but the policy may change tomorrow. It's not easy to guarantee anything 10 years from now, let alone 100 years.

New comment by JawsOfALion in "Ask HN: Where to put a static page that would last forever"

JawsOfALion — Sun, 10 Nov 2024 11:04:27 +0000

I'm pretty sure if you put a file on s3, make it public, it will be accessible for quite a while, with no cost (if it's just text like you're describing). I have an aws account that has the html/js stored in s3 for a static website i created, I haven't paid a bill in years on it, and the site is still running today.

I don't expect it to last 100 years though, they may very well change their free policies a year from now, hell, i don't know if AWS or even Amazon will be around 100 years from now.

Ask HN: Local Image Classification in the Browser?

JawsOfALion — Sun, 10 Nov 2024 10:57:54 +0000

I want to create a chromium extension, one of the main components of the extension is classifying images (think dynamic content filtering, a few different categories, one of which is recognizing inappropriate content).

Originally I wanted to use a multimodal llm to classify images, because they tend to do quite well at classifying images with little dev effort, but it seems like it won't be possible to my knowledge to get a local model working with a Chrome extension, and an api call for each image will be too expensive as my goal is for it to be free to use.

So next I looked into tensorflow mobile net, and tried this specific example:

https://github.com/tensorflow/tfjs-examples/tree/master/chrome-extension

It looked promising and while it technically worked, it seemed to do very poorly on categorizing most things(except tigers, it seemed to consistently recognize them well). Accuracy was far too low.

Anyways I would like to hear opinions of people who are more knowledgeable in this field, what's the best solution to do a rough, but accurate classification of images with the least dev effort and runnable on a browser? Should I invest time experimenting with other tensorflow mobilenet models, or should I expect fairly low accuracy in them too? (I would like to as much as possible avoid investing in the effort of training my own custom model at this stage)

Comments URL: https://news.ycombinator.com/item?id=42099554

Points: 3

# Comments: 3