Show HN: Filtered GitHub Trends (gh-trends.nilsherzig.com)

A Show HN project presents a lightweight interface that pulls GitHub Trending and lets you filter out repositories by keyword so you can declutter the feed — options include language and timeframe (daily/weekly/monthly) and an addable list of filter terms. The screenshot shows a default blocklist of AI-related tokens (ai, llm, rag, agent, language model, openai, gguf, model context protocol, claude, gpt, gpts, voice cloning), a “Filtered out: 0” counter and a toggle to “Show filtered.” Note: the demo currently displays “Error: Could not load trending data,” so live fetching wasn’t working at the time of the post. This is significant because GitHub Trending is increasingly saturated with AI/ML projects and tooling, making discovery noisy for developers not interested in that space and skewing visibility metrics. A simple client-side or proxy filter helps researchers and maintainers identify non-AI work, measure AI-term prevalence, or create curated feeds. Key technical implications include the need for reliable trending retrieval (rate limits/CORS), support for pattern matching or regex to capture variants, potential integration with RSS/APIs for automation, and careful UX around false positives/negatives. If expanded, it could also support category-based filters, scoring thresholds, or public blocklists to study AI hype on open-source platforms.