Title: locally controlled recommendation engines

2025-03-16 22:59:39 +0800

-----

2025-03-16 07:54:26 -0700

(chatgpt says:)

Integrating LLMs into Firefox extensions is still relatively niche compared to more traditional extensions like ad blockers or privacy tools. However, interest is growing as LLM APIs (like those from OpenAI) become more accessible. Here are a few points to consider:

- **Emerging Trend:**

While only a handful of Firefox extensions currently leverage LLMs, developers are increasingly experimenting with them for tasks like summarizing web content, generating responses, or even assisting in writing.

- **Platform Dynamics:**

The majority of early LLM-powered extensions have tended to appear on platforms with larger user bases (like Chrome), but Firefox’s commitment to privacy and open standards makes it an attractive platform for innovative projects.

- **Barriers to Adoption:**

Cost of API calls, performance considerations, and ensuring data privacy are some of the challenges that may slow widespread adoption.

Overall, while not common yet, we’re likely to see more Firefox extensions integrating LLMs as developers continue to explore their potential.

--

so I have this idea to create a firefox enxtensions that lets a user run their own recommendation engine that's highly private and gives exclusive control to the user. that way the user can interact with the engine, tweak and refine and also ask for the opposite, or even possibly orthoganal recommendations with some kind of control over the degree of divergence.

this is a big project though, the first layer is to get the thing to compile data. I guess I'd start with catching data on just youtube videos. second step would be to process the data and call to an llm for recommendations. third step would be to create interactivity, possibly up and down votes on recommendatins and the option to ask for opposites.