We don't need more apps, we need better open source

AI-assisted “vibecoding” is making it trivial for individuals to spin up bespoke apps, raising questions about whether energy should go into yet more custom tools or into strengthening existing open-source projects. Commenters highlight severe sustainability and governance problems in open source: maintainers are overwhelmed by low-quality, often AI-generated pull requests, funding remains scarce, and expectations around free labor and support are misaligned. Others counter that OSS was never a guaranteed service, argue for clearer norms and licenses, and suggest that in a world where code is cheap, long-term maintenance, social coordination, and domain expertise become the real bottlenecks.

Licensing, capitalism, and fairness

  • Some view commercial wrappers around OSS and LLMs trained on public code/content as “enclosure” and rent‑seeking: extracting value from community work without compensation.
  • Others argue this is exactly what permissive OSS licenses allow; if you don’t want that, use a non‑free or strong copyleft license.
  • Disagreement over intent:
    • Some use OSS explicitly to undercut proprietary vendors (e.g., “make it harder for them to charge money”).
    • Others see capitalism and commercial vendors as what makes serious, supported software possible.

Funding and sustainability of open source

  • Widespread agreement that “open source doesn’t pay the bills”; many maintainers are burned out.
  • Ideas floated: paid PRs or escrow per PR, charging for distribution/support of FOSS, donations, co‑ops, “shop local” for small vendors.
  • Skeptics note free‑rider problems, race‑to‑zero pricing, and that most existing OSS that earns money is older and entrenched.

LLMs, vibecoding, and “slop”

  • LLMs let individuals rapidly build bespoke apps (VST players, piano trainers, running and food‑logging apps, apt caches, taggers, etc.), often faster than configuring existing tools.
  • Many praise this as empowering non‑experts and lowering barriers to contributing to older, complex projects (with an “AI copilot”).
  • Others see “vibecoding” as inherently non‑rigorous: people ship code they don’t understand, with weak domain knowledge and testing.
  • Term debate: “AI‑assisted coding” vs “vibecoding”; some reserve “vibe” for code largely unread/unchecked by its “author.”

Maintainers’ overload and governance responses

  • Strong theme: popular OSS projects are being flooded with low‑value or AI‑generated PRs and security reports (hypothetical bugs, hallucinated issues, resume‑padding contributions).
  • Review time and trust, not code, are now the scarce resources.
  • Reported responses:
    • No‑LLM or “no AI” contribution policies.
    • Requiring issues and discussion before PRs; auto‑closing unsolicited PRs, especially from new accounts.
    • Disabling PRs entirely or closing projects to contributions; forking for personal needs instead.
  • Some suggest using AI agents to triage issues/PRs and reproduce bugs, but others counter this adds setup, cost, and new attack surfaces.

Bespoke apps vs shared tools

  • One camp: custom, LLM‑built single‑user apps are often better than heavy, general‑purpose apps that are hard to install, configure, or adapt to real workflows.
  • Opposing view: thousands of near‑duplicate, vibe‑coded tools fragment effort, overload communities and app directories, and weaken shared ecosystems.

Security, future ecosystem, and norms

  • Concerns that OSS may become a primary security vector; proposals for tighter dependencies, reproducible builds, and industry‑funded “stewarded open source.”
  • Emphasis that OSS licenses never guaranteed support; calls for clearer SOCIAL.md‑style expectations and “stewarded” vs “license‑only” distinctions.
  • Broad agreement that the hard part now is long‑term maintenance, coordination, and social interaction—not writing the initial code.