01
BUILD
Publish the ranking before the names
Combine multiple public rankings, freeze their dated inputs and publish the criteria and code. Commit the resulting list on-chain so later changes are visible.
soonami4good · concept v0.1 · September 2026
What do the field’s leading researchers actually think?
For every major argument in society, a published vote of the 300 most significant researchers in that field—selected through a reproducible ranking, verified by cryptography, and recorded where nobody can quietly edit it later.
SELECTIONReproducible
IDENTITYCryptographically verified
RECORDContent-addressed
OPERATORDesigned to disappear
01 / THE IDEA
Scientific Vote does not settle scientific questions. It measures what a field’s most significant researchers currently think—and makes that evidence inspectable.
If the panel splits, the split is the result. Dissent, abstentions, confidence and response rate are published as prominently as agreement.
02 / THE PROTOCOL
Four steps. Every one open to inspection.
The method separates the editorial choice of who belongs on a panel from the cryptographic proof of who controls each seat.
02
CLAIM
Let researchers prove their own seats
A scientist proves control of their institutional email through DKIM-in-ZK, with eduGAIN as the planned fallback. The operator never holds a secret that can claim a seat.
03
VOTE
Ask precise statements and publish everything
Five-point responses, a 1–10 confidence score and optional comments. Ballots stay encrypted until close; results show individual answers, dissent and abstentions.
04
REPEAT
Carry the protocol into the next field
The code, claiming mechanism and governance remain. Only the rankings and institutional domains are rebuilt—making each new panel cheaper to establish.
03 / THE RESULT
No headline without the evidence beneath it.
A published result always keeps the distribution visible—not just the majority.
Strongly agree
Agree
Uncertain
Disagree
Strongly disagree
Illustrative interface only—no vote data is shown.
A
Every answer. Named responses where participants opt in, with a pseudonymous route for safety.
B
Every confidence level. Uncertainty is measured rather than hidden inside a single percentage.
C
Every limitation. Response rate, demographic composition, conflicts and abstentions travel with the result.
D
Every version. Published records remain addressable and cannot be silently replaced.
04 / TRUST MODEL
Be precise about where trust remains.
Cryptography can verify that the right person controls a seat. It cannot decide who deserves that seat in the first place.
T1 · WHO ARE THE 300?
An editorial judgment, made contestable.
No system removes this choice. Scientific Vote makes it pre-registered, reproducible, committed, challengeable and declinable.
- Criteria published before names
- Open code and pinned inputs
- Corrections are bounded and timelocked
- Panel composition published with each result
T2 · WHO HOLDS EACH SEAT?
A cryptographic proof, without an enrolment desk.
Institutional email signatures become public, verifiable proofs. The operator cannot impersonate a panellist or quietly reassign a seat.
- DKIM-in-ZK as the recommended mechanism
- eduGAIN fallback where DKIM coverage fails
- Passkey-signed smart accounts
- Public claim and challenge window
05 / GOVERNANCE & FUNDING
Build an institution that no founder has to remain trusted to run.
PRESENCE
Content-addressed
An ENS name resolves to pinned site files and records, preserving a verifiable history rather than a quietly editable database.
TREASURY
Yield-funded
ETH is staked in a Rethink Finance vault. Operations use yield while the principal remains untouched.
CONTROL
Guardian sunset
A temporary guardian protects launch. Its powers expire only after public, measurable conditions are met.
CADENCE
Four-month cycle
Governance follows the soonami4good cycle, with timelocks and a deliberately narrow privileged surface.
06 / DESIGNED AROUND FAILURE
The difficult parts are social before they are technical.
01Nobody votesParticipation
Use two-minute ballots, begin at low frequency, build a committed standing panel and pre-commit the minimum response rate required to publish.
02“It’s a blacklist”Legitimacy
The list is an invitation, never a judgment. Anyone may decline. Non-participation is never reported as disagreement, and the ranking method remains open to competing implementations.
03The selection picks the answerBias
Use multiple co-equal inputs, including a non-bibliometric roster; publish code, criteria and the panel’s composition beside every result.
04Question wording becomes the storyFraming
Draft wording in public, use standardised effect-size language and require panel ratification before any vote opens.
05Lobbying or governance captureIntegrity
Encrypt ballots until close, disclose conflicts, audit a random sample and place every privileged action behind a timelock.
07 / ROADMAP
Each phase has a gate—not just a date.
FEASIBILITY
Test the identity layer
Survey institutional DKIM coverage, prototype the ranking, approach Carbon Brief and scope the legal wrapper.
Gate: DKIM threshold met or eduGAIN fallback confirmed.RANKING
Publish criteria, code and the 300
Freeze all inputs, commit the Merkle root and open the challenge window.
Gate: a third party reproduces the list independently.CLAIMING
Open seats and support
Launch wallets, recovery and a non-crypto onboarding path.
Gate: reach the pre-committed claim rate—proposed at 40%.FIRST VOTE
Publish three to five climate statements
Panel-ratified empirical statements, complete results and visible composition.
Gate: result published in full with response rate.SUNSET
Remove the founding operator
Transfer control to the DAO after one complete governance cycle without guardian intervention.
Gate: decentralisation demonstrated, not announced.SECOND FIELD
Prove the system repeats
Reuse the protocol for quantum computing or nutrition.
Gate: materially lower marginal cost than field one.08 / OPEN QUESTIONS
The decisions that still belong in public.
01Should the panel be selected by rank alone, or balance geography, discipline and representation?
02What minimum claim rate makes a vote publishable at all?
03Should individual ballots be named, pseudonymous—or decided by each panellist?
04Who may propose statements, and who ratifies the wording?
05Should Scientific Vote partner with Carbon Brief or build only from open inputs?
06What structural protection matters when a result is politically inconvenient?
THE HONEST ONE-LINE VERSION
A record, not a persuasion campaign.
Scientific Vote will not tell society what to believe. It will make one important claim checkable: what the field’s most significant researchers currently think.
Read the concept paper