Signed proposals. Weighted consensus. No single jurisdiction.

Who has standing

Every contributor at alphabell signs the research conduct charter on entry and acquires vote-of-record at the 24-month tenure mark. Until then, contributors are full participants in their cells but do not vote on lab-level proposals. The 24-month threshold is not a hazing ritual — it is the empirical point at which we have, on the historical record, been able to assess whether a contributor's understanding of the lab's structural commitments is well-developed enough that their vote is informative.

Long-tenured contributors are 142 of the current contributor pool of approximately 410. They are listed at /people.

What a proposal looks like

A proposal is a signed text document, posted to the internal proposal queue, with a fixed structure: (1) the change being proposed; (2) the reasoning; (3) the immediate consequences and the parties affected; (4) the rollback procedure if the proposal is ratified and later wishes to be reversed; (5) any sunset clause or re-vote schedule. Proposals are typically 600 to 2000 words long. They are written in templated Markdown.

Proposals are reviewed in two stages: a comment period (default 14 days; some proposal classes have different defaults) and a vote (default 7 days). The author may revise the proposal during the comment period in response to objections; substantive revisions reset the comment period.

How votes are weighted

Every long-tenured contributor's vote is weighted as follows:

weight(c) = clamp(1, 4, 1 + log2(tenure_months(c) / 24))

That gives a contributor at 24 months a weight of 1, at 48 months a weight of 2, at 96 months a weight of 3, and at 192 months a weight of 4. The cap at 4 is intentional: long tenure earns more voice, but not unbounded voice.

Some proposal classes — those touching the charter, the publication policy, the RSI capability-evaluation methodology, or the federated compute scheduler's allocation algorithm — use a higher passing threshold (currently 67% of weighted vote) than the default (50%). The proposal class determines the threshold mechanically.

What governance decides

• Charter changes. Highest threshold. Rare. Three since 2017.
• Publication policy changes. High threshold. Five since 2017.
• RSI capability-evaluation methodology. High threshold. Reviewed quarterly; substantive changes vote separately.
• Federated scheduler allocation algorithm. High threshold. Two major revisions; v2 ratified 2025.
• Cell formation, fission, merger, dissolution. Default threshold. Formation roughly once a quarter; the rest as required.
• Axis stewardship rotation. Default threshold. Stewards rotate every two years; the vote ratifies the cohort of cell stewards within an axis.
• Physical anchor decisions. Default threshold. Four anchors approved; most recent: Bali (2025).
• Funding-source approval. Default threshold. Each new funder is approved as a proposal; sovereign-research partnerships pass a higher threshold than philanthropic grants.
• Interpretability pairings. Cell-level governance. The pairing itself is a signed record between the two cells, not a lab-level proposal — but the pairing-rules amendments are.

What governance does not decide

Governance does not direct research. There is no body at alphabell with the authority to instruct a cell to work on a particular problem or to abandon one. The closest the protocol comes is the axis-steward role, which carries convening power and the responsibility to maintain an axis's published research agenda — but not directive authority over cells.

Governance also does not decide on individual hires inside cells. Cells onboard their own new contributors. The long-tenured-contributor quorum can review a contributor's standing at the lab — including charter compliance — but does not vote on routine onboarding.

Failure modes we know about

How to read a proposal

The internal proposal queue is gated behind contributor authentication. Public-interest proposals — funding-source approvals, anchor decisions, charter revisions, publication-policy revisions — are mirrored to the public news index after ratification. See, e.g., the Bali anchor decision or the External Evaluation Cooperative membership approval.