• by Andrew Nesbitt

  • source https://nesbitt.io/2026/02/13/respectful-open-source.html

  • … fix from a stranger still carries cognitive load beyond just merging: triage, review, checking for regressions, responding,

    • managing expectations …
  • … once you merge someone’s code, you’re maintaining it

  • … fix exists on my fork. …

    • nobody will because fork discovery is effectively broken.
  • Git was pull-based

    • GitLab is at least honest about it by calling them merge requests.
  • If the default were flipped so that fixes exist publicly …

    • contributor’s job would be done when the fix is public …
  • Fork discovery is broken

    • GitHub made forking easy and fork discovery nearly impossible.
    • … people use the fork button as a bookmark, … bots, … generate forks that are nothing but noise.
  • … issue #347 and being able to see “three forks have patches touching this code” without anyone having submitted anything,

  • … maintainer didn’t ask for the audit, didn’t agree to the compliance framework, and is now expected to respond on someone else’s timeline.

  • … Private vulnerability disclosure … has its own AI spam crisis as anyone following curl’s experience with HackerOne

  • Miranda Heath’s report on burnout in open source names six causes

    • workload is only one of them:
    • toxic community behaviour,
    • hyper-responsibility,
    • pressure to keep proving yourself
  • Communities are not fungible,

    • built on years of shared context and ambient trust that
    • can’t be rebuilt once the people holding them together burn out.
  • Unsolicited PRs, drive-by issues, and automated audits are all withdrawals from a finite account.

  • AI coding agents …

    • … volume of low-quality inbound to popular projects has exploded.
  • … experienced maintainers who tried graduated responses and ended up at the nuclear option because nothing else worked.

  • AI pressure is going to force that switch on more … repos …

  • GitHub made forking a one-click operation … without ever investing in making the resulting graph navigable …