HN user

mska

341 karma

Twitter @RealBrianHi

Posts18
Comments27
View on HN

When views are low the math doesn't make sense but it is very possible to get a lot of views through AI generated + human reviewed content.

We're trying to do that with PulsePost (https://pulsepost.io) and the biggest challenge is unique content. Given a keyword or a niche topic, AI models tend to generate similar content within similar subjects. Changing the temperature helps to a degree but the biggest difference comes from adding internet access. Even with same prompt, if the model can access the internet, it can find unique ideas within the same topic and with human review it becomes a high value article.

Nice! I'm also building an online IDE [1] and was working on adding Chat feature for a while. The problem of making adjustments to an existing component is much more difficult than generating a new one from scratch.

Also, the tokenization for raw HTML code produces too many tokens which drives up the cost. Optimizing the tokenization process for HTML could potentially reduce costs and enhance performance.

[1] https://divmagic.com/studio

I'm currently working on an extension as well ([0]) and share the same concerns many have mentioned about extensions here. I'd like to highlight another dimension concerning the Browser APIs ([1]).

Handling the permissions necessary for certain API functionalities and the corresponding warning messages can be somewhat confusing. For instance, our extension uses "chrome.devtools.panels" to open a new window within DevTools. This API doesn't require any permissions by itself. Yet, for messaging across the popup, content, and DevTools windows, we're required to use activeTab and sendMessage APIs. The DevTools window operates in its unique context, almost like a tab within another tab. For example, updating the URL in the active tab doesn't directly update the DevTools window but triggers an event.

Messaging across these different contexts requires the "https://*/*" host permission, without which Chrome and Firefox won't send the messages between these isolated windows.

We made this permission optional, the DevTools Panel is activated only upon receiving explicit user consent. However, the permission prompt's messaging is something like "This extension requires access to all your data," which sounds very alarming. We don't access any data nor that we want to, but requiring that permission is mandatory since the message APIs won't work without them.

This is just one example of the many undocumented complexities within Chrome's documentation. Similar pitfalls exist with message exchanges between the background service and content scripts. Sometimes you don't know why your API call doesn't work even though you think you have the required permission and asking for more permissions show very alarming messages to users.

I think that a more granular permission approach, made specific to API functionalities rather than broad permissions that cover a list of APIs, would significantly help user experience. For example, requesting permission for the "sendMessage API" with a clear explanation would be far more informative for users than the general "All host https:///" permissions.

There's also the issue of building for different browser. The same browser API calls can have different permissions requirement on Chrome and Firefox which makes the development process more difficult and more confusing for users since the same extension requires different permissions on different browsers.

[0] https://divmagic.com [1] https://developer.chrome.com/docs/extensions/reference/api

Google Sidewiki 3 years ago

I had created a browser extension that does the same thing: https://everycomment.top

The biggest reason why I built it was to be able to get insight from a community on every site. Most web sites don't have comments and some filter a lot so I wanted to have a platform where the website owner can't modify or change the comments from people.

I didn't work on finding users though and it has about 10 people using it.

Out of curiosity, how do companies offering self-hosted/on-prem solutions monetize their offerings?

Do they rely on legal contracts to prevent customers from using the software for free or modifying it for their own purposes?

The biggest part in my opinion is data. Both quality and quantity.

Then, the deep learning model architecture and the hardware (GPUs).

I'm trying to figure out how much better GPT-4.5 will be compared to GPT-3.5

It is not possible to give a ratio with certainty. Since it will be trained on more data and probably on better hardware, all we can say is that it will be better.

I didn't know about ThirdVoice, thanks for letting me know.

The addon sends a request to the server every time the URL changes, without the user pressing a button (such as: a new tab is opened, the active tab changes)

The only reason for that is to get the comments for the current page. I could make this happen when the user clicks on a button but I thought that would make the experience slower and/or worse.

I do agree with the privacy part and I care a lot about that. There is absolutely zero tracking right now. There is no way to link or match requests to any particular user.

You can also verify that by checking the Network tab. The request will only include the current URL without any additional data.

I wanted to make this open-source to provide visibility but it will make it very easy for anyone to just copy it and submit as a new extension. If you have an idea about how I can make this open-source without people easily copying it, I'd be more than happy to know.