Who speaks for Nixpkgs: core burnout and the Stack opt-out

Two threads on the NixOS Discourse have converged on a single institutional question. The Nixpkgs core team has stepped down after ten months, citing burnout and a governance design that never quite worked. Almost in parallel, a dispute over opting the project's repositories out of Hugging Face's The Stack has turned on whether any individual, or even a Steering Committee member acting alone, may speak for the whole. Both episodes are about authority under volunteer conditions: who may bind contributors, how much real power a steering body holds, and whether a package set of this scale can absorb modern supply-chain and consent pressures without structural change.

The core team's exit was framed as honesty rather than drama. qyliss announced that the group had "unfortunately decided to disband," listing concrete work from the past ten months: reforming committer delegation, onboarding nineteen new committers, extending the merge bot, restoring contact with GitHub and securing an Enterprise Cloud upgrade, helping triage a GitHub security advisory, and establishing an initial automation and AI policy. The pride was real. So was the diagnosis. The role "has sadly not turned out to be the lightweight role compatible with active technical contribution that we had originally hoped it would be." Attrition had already thinned the team. A call for new members produced one active applicant and a mixed response to outreach. With a Steering Committee election imminent, the remaining members concluded that dissolving the team was preferable to pretending it could continue.

The sharper institutional claim sat in the experience report itself. "Our experience is that the Steering Committee as an institution lacks a native instinct for the delegation envisioned." In other words, the body meant to hold strategic authority had not developed the habit of handing operational power downward in a way that made a consensus-focused core team sustainable. Reactions on the thread largely accepted the reasoning even while mourning the outcome. yuuko called the decision "entirely sensible, regrettable though the outcome may be," and looked toward a future environment "more amenable to something like the core team comfortably existing and not burning out." Others thanked the departing members and asked the practical question stefan-hacks put bluntly: with the core down to a handful of people, what are the next steps, and why was the unsustainability not confronted earlier? The graph of commit activity shared in the thread served as a quiet reminder that the same people had been carrying both governance load and technical work until the combination failed.

The Hugging Face fight poses the authority problem from the opposite direction. nrabulinski opened by describing The Stack as "an enormous dataset, composed almost entirely out of stolen data people have not consented to," and asked someone with organizational privileges to submit an opt-out for NixOS repositories. Individual contributors lacked the power to remove organization-owned trees even when they rejected the inclusion of their own work. hexa added the downstream technical stake: the data would train models used "to generate code that lacks any licensing indicators that it should as derivative works." Consent, license provenance, and the practical reality of LLM output collided in a single request.

What followed was not primarily a debate about machine learning. It was a debate about who may act. samueldr had moved to opt out repositories connected to their contributions and then confronted what they read as an override. They stated they "never retracted or intended to retract the request for opting-out" and objected that another party's message implied "the authority to make the call on my behalf." The procedural line they drew was precise: adding the matter to the Steering Committee agenda would have been proper; acting proactively was not. "Now an action was taken in the name of the project and contributors. Despite the wishes of at least one contributor." Even SC weight, they argued, would at least have come "from a source of authority" rather than from an individual unilateral step.

The counter-position, pressed by ile-du-fantome among others, inverted the overreach claim. The proactive step, on this view, was the original opt-out of entire repositories by someone who was "not the sole contributor or owner." Organizational repositories, even when one person wrote most of a given tree, belong to a collective; inclusion or exclusion "should be up to the leadership of that organization, not any individual contributor." Licensing status and sole authorship were treated as orthogonal to the question of who may bind the org. The result was a mirror image of the core-team complaint: in one case the center would not delegate; in the other, the periphery could not be allowed to decide for the center.

Cafkafk, speaking from the Steering Committee side, reported the interim state with little comfort for either camp. People had reached out with questions, but the committee had "not discussed this yet in actual committee quorum." There was "isn't a clear direction from the SC yet," with the topic possibly landing in a later meeting after a board sync. The body whose authority both sides were invoking had not yet taken a formal position.

Read together, the two threads describe one crisis with two faces. The core team was created to embody bottom-up, consensus-focused governance inside Nixpkgs and discovered that the surrounding institutions did not know how to feed it work, protect its time, or recruit its successors. The Stack dispute shows contributors and informal actors reaching for levers (opt-out forms, public pressure, individual copyright claims) because no settled process exists for deciding when the project as a whole is in or out of a training corpus. In both cases the Steering Committee is named as the proper locus of decision and, at the same moment, described as slow, incomplete, or lacking the instinct to delegate. Volunteer infrastructure built for packaging and code review is being asked to absorb burnout risk, security-incident load, AI-policy formation, and mass non-consensual data collection without a clear map of who may speak when the answer binds thousands of contributors.

Nothing in either thread resolves the design problem. The core team is gone; its achievements remain, and its warning about unsustainable load is on the record ahead of an SC election. The Stack opt-out remains contested, with at least one contributor insisting their personal request stands, others insisting only organizational leadership may decide, and the SC still without a quorate direction. Whether a future core-like body can exist, and whether consent and supply-chain questions of this kind will be handled by policy rather than by improvised individual action, are left open. The community has a sharper statement of the authority gap than it had a month ago. It does not yet have an institution that closes it.