Public data
The signal is computed from data anyone can read for free: an official register, a public API, a filing system, a chain. No paid feeds, no private data and nothing about private individuals.
Open Signals · Call for signals
Open Signals lets researchers propose new signals for The SI Observatory and publish them under their own name. A good signal is one number built from public data with a method anyone can repeat and a clear refresh rate. The team reviews every proposal in private, builds each accepted signal itself, and credits its author on the signal and in the list of Observatory Fellows, a title with no pay or reward.
The bar
Every live signal on the Observatory meets the same three tests. Proposals that meet them are the ones that go live.
The signal is computed from data anyone can read for free: an official register, a public API, a filing system, a chain. No paid feeds, no private data and nothing about private individuals.
One number, with a method short enough to write in plain words and exact enough that two people get the same result. Say what is counted, what is left out and why.
How often the number changes and how often it should be read: hourly, daily, weekly or monthly. A signal that cannot be refreshed is a study, not a signal.
Not accepted: code to run, price predictions, anything that scores people, and data behind a paywall or a login.
Examples
Each live signal fits in three lines: what is counted, from which public source, and how often. A proposal can be just as short.
share = documents saying Super Intelligence / (documents saying Super Intelligence + documents saying only the old name), over the recent window the tracker reads.
Data: Executive Order 14434, 91 FR 63129, Federal Register API
The full methodmeter = rogue SI headlines / all SI headlines over the last 7 days, times 100.
Data: RSI Watch method
The full methodA case is logged when at least one primary or reputable secondary source documents it. Severity follows the published scale; the highest level the evidence supports wins.
Data: Rogue SI incident log (JSON), METR documented agent incidents
The full methodA measure counts as in force from its effective date. Voluntary codes and guidance are never counted as in force.
Data: SI control map (JSON)
The full methodSum of current power (MW) for every watched site, as listed in the Epoch AI data center dataset.
Data: Epoch AI data centers (CC BY), NASA GIBS, Harmonized Landsat and Sentinel-2
The full methodCoinGecko category market cap for "artificial-intelligence"; the token count is every coin listed in that category.
Data: CoinGecko Artificial Intelligence (AI) category
The full methodPropose
Write the method in plain words and link the public data. The proposal is private: it is stored only for the review, and only the name you choose and the method text are ever published, after acceptance.
Send the form below. Your contact stays private; only the review team reads it.
The team checks the data sources, repeats the method by hand and may write to you with questions.
An accepted signal is built from your method by the Codeflex team. Code from a submission is never run.
The signal page credits you as written in the form, and you are listed as an Observatory Fellow.
Accepted signals
No proposal has been accepted yet. Accepted signals appear here with their authors, then join the live board.
Observatory Fellows
No Observatory Fellows yet. An author becomes a Fellow when a proposed signal goes live. Fellow is a title only, with no pay, reward or rights.
Questions
Anyone: researchers, students, journalists and builders. You publish under the name or handle you choose, with an optional affiliation.
No. Accepted authors are credited on their signal and listed as Observatory Fellows. Fellow is a title only, with no pay, reward or rights.
No. Describe the method in words. The Codeflex team builds every accepted signal itself and never runs code from a proposal.
They are stored privately for the review and used only to reach you about your proposal. They are never shown or shared.
The method text is published under CC BY 4.0 when a signal is accepted, with credit to you. The numbers follow the Observatory's open data license, CC BY 4.0.
On the three tests above: public data, a clear method and a refresh rate, plus whether the signal tells people something new about the SI era. The team may ask questions before deciding.