OpenClaw surpasses React to become the most-starred software project on GitHub

Original Article ↗ Hacker News Discussion ↗

GitHub stars, legitimacy, and metrics

  • Many see OpenClaw’s rapid rise past React as proof GitHub stars are now a weak/“gamed” metric (Goodhart’s law, comparable to bestseller lists or Facebook likes).
  • Suspicion that stars are inflated by:
    • OpenClaw agents starring the repo during/on setup.
    • Bot farms and purchased stars.
    • Hype from AI incentives (e.g., free API credits for highly-starred repos).
  • Others argue the stars may largely be “real” because:
    • Issues/PR volume is enormous and roughly consistent with popularity.
    • It’s genuinely fun and widely appealing, especially to non-programmers.

Practical use cases vs “solution in search of a problem”

  • Reported real uses include:
    • Personal briefing agents (calendar + weather + news), TODO and reminder management.
    • Email triage and limited scheduling; Discord help bot; WhatsApp/Telegram/real-estate lead follow-up.
    • “Second brain” for videos/papers/podcasts, knowledge organization, summaries, reading lists.
    • Home automation (coffee, Home Assistant hooks), network scans and server reboots.
    • Automating painful web workflows (booking, scraping sites with bad UX).
    • Dev helpers: repo edits, deploying small changes, long-running tasks like compiling Node on old hardware.
  • Critics say nearly all of this is doable with scripts, cron, Zapier, Automator, n8n, or LLM-in-IDE tools—and often more safely, cheaply, and reliably.

Risk, safety, and trust

  • Strong concern about giving an LLM agent:
    • Write access to email, cloud accounts, or production systems.
    • Control of personal messaging accounts (WhatsApp/Telegram) and real identities.
  • Examples cited of agents deleting email inboxes or potentially exfiltrating data.
  • Suggested mitigations: strict sandboxing (VM/VLAN, separate OS user), read-only access, narrow APIs, backups.
  • Counterpoint: any powerful tool is risky; responsibility is on users to restrict access appropriately. Others reply that nondeterministic behavior makes this fundamentally unlike simple tools like rm.

Cost, efficiency, and overengineering

  • Concerns that constant “heartbeat” loops and multi-hour agent runs are token-hungry and inefficient; some report quickly rising monthly bills.
  • Some rely on free tiers or local models, others pay substantial monthly API costs and feel it’s worth it.
  • Several foresee future token prices rising, at which point many current agent workflows may look like expensive overengineering.

Cultural and broader implications

  • Enthusiasts emphasize:
    • It finally lets non-programmers make computers “do things” via natural language.
    • It’s simply fun; feels like having a sci‑fi assistant, which drives adoption more than pure utility.
  • Skeptics see:
    • A hype- and influencer-driven fad, akin to gadget obsession or early smart-home over-automation.
    • Acceleration of “dead internet” dynamics: bots talking to bots, spam/marketing agents flooding platforms.
  • Some predict agent-centric computing will reshape how people interact with software and kill many niche SaaS products; others think this will settle into a middle ground once novelty fades.