Awesome repositories are no longer awesome
Once-useful GitHub “awesome” lists of curated tools and resources are increasingly seen as bloated, outdated link dumps driven by self-promotion, SEO, and star-farming. Maintainers describe being overwhelmed by low-quality and AI-generated pull requests, making genuine curation costly and leading some to abandon or heavily gatekeep their lists. Many still value well-maintained, narrowly focused lists as discovery tools, but argue that future solutions will need stronger editorial standards, automation for filtering obvious spam, or entirely new ways to surface high-quality software.
Perceived Decline of “Awesome” Repositories
- Many see most awesome lists as degraded from curated collections into sprawling, low-signal catalogs.
- Comparisons are made to early web directories and webrings: briefly useful, then overwhelmed by spam, self‑promotion, and outdated links.
- Several commenters say these lists were “always junk”; others say the earliest ones were genuinely curated and valuable.
Curation, Effort, and Burnout
- Maintaining a high-quality list is likened to gardening or editing: time-consuming, with content constantly going stale.
- Maintainers report:
- Large volumes of self-promotional and low-quality pull requests.
- People submitting projects after only hours of existence or with no users.
- Star‑farming behavior and “content marketing” disguised as curation.
- Some add strict guidelines, templates, CI checks (e.g., minimum stars, repo age) to keep quality up; others consider or have decided to archive their lists.
Impact of AI and Automation
- AI agents dramatically increase spam PRs, often ignoring contribution checklists and flooding maintainers.
- This creates “asymmetric warfare”: trivial to auto-generate 100 PRs, costly for humans to review.
- Proposed mitigations:
- Deterministic filters (repo age, stars, template adherence).
- AI or bots as pre-filters that only reject obvious trash, not decide what is “awesome.”
- Some reject using AI for judgment, arguing that taste and common sense are inherently human.
Ongoing Value and Alternatives
- Supporters say good lists still:
- Provide overviews of an ecosystem and “unknown unknowns.”
- Beat SEO‑driven listicles and shallow search results.
- Critics find them too long, unfilterable, or redundant with search/LLMs.
- Suggestions:
- Smaller, highly selective or “boring” lists with just one–two solid options per problem.
- Tools and sites that update automatically and rank repos semantically or by signals.
- Even “awful‑X” lists to warn people away from bad options.
Broader Reflections
- General agreement that high‑quality human curation is more needed than ever but increasingly hard to sustain.
- Debate over whether group curation can work reliably with good process, or whether incentives and scale doom most lists.