I’ve found it super nice.

Similar to the “AI should be an assistant programmer, not an independent dev”, having Claude there and being able to ask questions about specific topics while I read is fantastic. Especially for scientific papers that are outside my specialty (i.e. all of them.)

Have you heard of tabs?

Have you heard of the HN guidelines?

> Be kind. Don't be snarky. Converse curiously; don't cross-examine. Edit out swipes.

Didn't know about it either, until it got pushed in my face a few days ago, for whatever reason. Didn't leave a great taste in my mouth.

I’m used to all browsers adding features that I don’t want every few months. I just disable them and forget about them.

Not mentioned is the ability to setup custom prompts. [1]

I enjoy reading technical blogs from the global south and the slavic world. I've found that LLMs do a far better job at translation than Google Translate/DeepL, etc. in these niche domains, so I added a translate prompt to my context menu and that converted me over to using it.

[1] https://connect.mozilla.org/t5/discussions/advanced-configur...

Thankfully they are unconnected by default. Unfortunately the sidebar can get triggered accidentally with keystrokes from muscle memory (Emacs) and then one has to deal with the side bar. It would have been nicer if by default nothing existed in the first place.

I have a Gemini sidebar, it just appeared out of nowhere one day, but I find it useful though. You can give it a URL and it can summarize it or whatever.

It's super useful for power users. You can connect any model you like through ollama (or similar services) and use your own pre-defined prompts to run against the selected text on websites simply from right click context menu.

How to connect it to Ollama? I use PageAssist[1] extension for that.

[1]: https://github.com/n4ze3m/page-assist

You have to change settings in about:config, e.g. `browser.ml.chat.provider = http://localhost:3333` with the address of your model server (I actually use open-webui with ollama for that).

Then you could change the prompts, for example, those are mine :

browser.ml.chat.prompts.0 {"id":"search", "value": "What do you know about the selected text?", "label": "Search"}

browser.ml.chat.prompts.1 {"id":"grammar", "value": "Can you fix the grammar in the selected text?", "label": "Grammar"}

browser.ml.chat.prompts.2 {"id":"summary", "value": "Can you summarize the selected text?", "label": "Summary"}

browser.ml.chat.prompts.3 {"id":"code", "value": "Can you explain the selected code?", "label": "Code"}

browser.ml.chat.prompts.4 {"id":"translate", "value": "Can you translate the selected text?", "label": "Translate"}

I have found instructions here https://connect.mozilla.org/t5/discussions/advanced-configur..., though they are a bit outdated since you have to setup prompts differently.

It was weird when it just showed up one day, but after using it a bit, I like it.