HomeBrew HN, a project by azath92, aims to personalize Hacker News feeds using LLMs. Users rate 30 HN posts to generate a markdown-based profile reflecting their interests, which then powers a custom-ranked homepage. The project explores how little personal context LLMs need for effective personalization. Feedback suggests users appreciate the transparency and control offered by the editable profile, but want options to refine it with upvoted/commented posts or general text input. Users also desired multiple profiles that could align with different moods or contexts. Challenges include capturing niche interests, accurately classifying posts, and managing LLM costs when analyzing article content. Future development may incorporate user feedback loops, improved post selection, and potentially support for multiple profiles. The team is exploring LLMs like Gemini 2.5 flash, aiming for a balance between speed and quality, while also valuing transparency and user agency over traditional recsys approaches.