Noi: an AI-enhanced, customizable browser | Hacker News Reader

Noi: an AI-enhanced, customizable browser

github.com

55 pointsBluestein39 comments

Is it really AI-enhanced, or is it just a handful of features for AI interaction (like a snippet repository for "prompts" or sending the same query to multiple chat pages) invented by someone who does not have the resources to do their own AI but really wants to get in on the hype? Either way, seems like a bad and overhyped idea for a browser.

I don't want an AI browser, per se. I want an AI agent that slips into every pane of glass the platform companies own (Chrome, iOS, Windows) and works for me against the advertisers and attention stealers.

I want an advocate that detects and nukes advertisements. That filters clickbait and rage content. Something easy enough that everyone can install, so we can all be collectively free of this nonsense no matter what tech stack we use.

Imagine if AI became the ultimate anti-advertising, attention-preserving, sanity-defending weapon.

"No Google, you're not allowed to advertise to my person." Or, "these comments are toxic drama, so let's not expose our human to them."

This would be a great new technological era.

Mm. Turn the web back into something you could read on a command line, turn social media back into what I remember IRC being 20 years ago.

Good boi AI. GoodboAI?

Yesyes, the movie HER actually had it right in that you ran it on your own hardware at home.

Cloud or large company AI has no hope, only at home, user chosen censorship, working for my own interests not some advertiser, doing everything that the big companies all working on this DONT WANT is what would interest me

Plus: trustworthy to have my interests in mind instead of whatever tech company owns it

I might sound shallow or incognizant for saying this, but any time I see traces of Chinese in a product, my radars alert "Stay away from this." Maybe it's because of the constant worry about them spying on western users. And maybe it's because I am not exposed to high quality software from that part of the world yet.

I see this a lot in LLMs too. So many Chinese companies trying to game the benchmarks to show higher in the leaderboards without the fear of any backlash. Is it really a skewed perception of how software works in China, or is it really the case that their culture encourages finding shortcuts to success at all costs?

I must confess that is my first instinct as well.-

I guess - where posssible - careful code review is one's friend ...

(But, point taken).-

You really need to think about your biases. I've worked with teams (devs & non-devs) from China for years and I did not find them different from "western" ones. Some people will do stellar work, some will try to find short-circuits and lie to your face.

The culture is absolutely different, so it may take time to recognize behavorial patterns in one way or the other - as a European working in the US, it also took me some time to figure out the "US way" of working.

Chinese coworkers with code review and visibility are more trustworthy than Chinese open-source developers.

It's because you are brainwashed and you cannot think for yourself.

If you're worry is the Chinese spying on you, you can look at the code and double check. Furthermore the US govt spies on us on a massive scale as Snowden showed us.

Second when it comes to software quality they started much later than us but our culture has as much finding shortcuts to success as theirs.

Case in point: open AI killing off the team to make AI stuff to make a quick profit. Stealing copyrighted work on a massive scale to train llms.

So get off your high horse. Stop trying to hide your dumbness (lack of critical thinking) behind an intellectual sounding comment.

I see you've been on HN more than long enough to know all of what was wrong with your response to GP's comment.

> you can look at the code and double check

Can you honestly say that when a few months ago someone was trying to put a security hole in OpenSSH by making changes that looked harmless to an archiving tool (the xz backdoor)?

Your response is such textbook propaganda, instantly riffing into whataboutism with US surveillance, that I actually had to wonder if you had a real profile.

I then did see that you have actual posts, but imagine if I'd have to investigate the profile of absolutely every commenter I read. Isn't that practically impossible to do? And isn't auditing the source code of absolutely every project you use a few times harder than that?

On the topic of AI web browsers, a feature to audit the people whose comments I read would be an interesting feature. Also made me think of: https://www.reddit.com/r/OutOfTheLoop/comments/94638i/what_i...

I think you are conflating "american" users with western users. Most users around the world are aware of Americans intentions, not chinese. Only Americans think Chinese products are a problem and you guys are investing billions on making this a world sentiment. Without much success.

tldr: Americans are afraid of Chinese people possibly spying on them, the world hates Americans real world-wide espionage system.

I'm not from the US and I'm wary of both. The US is geopolitically aligned much closer to the EU, which is why I'm much more cautious about potential involvement of the Chinese government compared to the US.

It's not shallow, it's well documented across the internet and in their behavior with everything they do.

China is the land of shortcuts and facades. It's part of their culture. See: gutter oil, washing veggies in the toilet, concrete infrastructure with no reinforcement, etc. Even when they respond to natural disasters: it's not to help citizens but pose for pictures to show them helping then bouncing.

Chinese people don't even trust other Chinese to sell them things: so companies often bring in foreigners to sell Chinese products to china.

China is VERY CLEAR in who they are, and they show no attempts to be otherwise.

I'm sure some people think your country is full of lazy, smelly thieves, too.

I'm from there, and they are!

>Browser: Noi not only includes curated AI websites but also allows the addition of any URL, providing a tailored browsing experience

Wait, that's what a browser is meant for. To add any URL.

Careful, until the developers clarify. https://github.com/lencx/ChatGPT/issues?q=trojan https://github.com/lencx/nofwl/issues/102

Are there any LLM products yet that can navigate websites for you ?

Or perform operations on your local computer?

We've got some experts on browsing agents working w/ us on OpenDevin: https://github.com/opendevin/opendevin

For local automation, you might be interested in https://github.com/OpenInterpreter/open-interpreter

Still early days for both though.

LaVague https://news.ycombinator.com/item?id=39698546

skyvern https://news.ycombinator.com/item?id=39700013

TaxyAI[0] is not totally useless. Here's my fork using Ollama locally.[1]

0. https://github.com/TaxyAI/browser-extension

1. https://github.com/gavmor/browser-automation

I tried installing and running the arm64 macOS app and received the following error:

> “Noi.app” is damaged and can’t be opened. You should eject the disk image.

They explicitly mention that in the readme, its an apple security measure and you need to run that xattr command.

What a disingenuous error message. I like Apple’s approach to TCC/permissions but this is quite obviously profit driven and intended to drive developers to the App Store with all the associated fees

Yep, and Microsoft does exactly the same thing. I have an open source video review tool for esports teams that I give away. On both Windows and Apple I'm expected to pay a yearly tax (~$200 on Windows from memory) to distribute my software on that platform. The owners of these platforms make no allowance for open source authors to distribute without that fee.

This is done in the name of "security", as if malware authors couldn't afford to pay it.

> as if malware authors couldn't afford to pay it.

I think it's the identity verification bit that's dissuades malware authors, not the cost. Having your malware linked to your name and address isn't something most malware distributors want.

Nonprofit orgs and schools can get free dev accounts (at least for Apple), so for some open source projects distributed by one of the open source nonprofits, it is free.

> I think it's the identity verification bit that's dissuades malware authors

I mean, the whole LastPass trojan situation seems to have proven that Apple doesn't even drill down on that: https://www.malwarebytes.com/blog/news/2024/02/warning-from-...

But what are they doing that's requiring going through settings/xattr, when every other direct download app I've used has worked fine with the right click -> Open to suppress the default error about signing?

Ah, I missed that. Thank you.

Interesting. I was just thinking about a wallet-like app or extension for AI to hold my API keys and interactions/conversations. I think that might be something in the future.

Go for it. It sounds like a tangible, workable idea.-

Do you think in future we would have AI based browsers with fluidic UIs to make everything simple for the user. Websites or backend would then provide hints for the fluid UI flows with pre-made flows to select from. Could be the next iteration for browsers. Instead of Dom we would have AI based Dom.

It sounds cool and horrible at the same time ...

I think some sort of AI use is virtually unavoidable in UX/UI testing ...

AI-based-DOM sounds like an oxymoron ...

... perhaps only my bias, which considers everything AI to be unstructured, which is clearly wrong.-

NoI(ntelligence). Sounds like a good name for something that explicitly doesn't have any AI bs

This looks interesting but 4.2k stars for a js app that shows readily available LLMs is fishy.....

I mean it's nothing ground breaking. I guess the AI bubble is real :)

Reply on news.ycombinator.com