Skip to content
Protocol STAMP
Latest Calibration

Publications

The CalibrationOur own measurements, pre-registered and signedThe LogWhat ran, what it cost, what movedThe BenchWhat we read and argue with · Mon and ThuArchiveEvery issue, by number and dateThe technical reportThe program in one paper, with its dataSubscribeOne email per issue

Method

MethodologyHow a rate is producedDefinitionsWhat each term means hereThe protocol, v1.25The instruction layer as measuredRun filesEvery draw we can publishfigures.jsonEvery citable number and its fingerprints

AI policy

The DocketRules that bear on how AI is testedThe ChronicleDated events in AI governanceSuper Intelligence (SI)The US rename of AI, tracedMilitaries and AICountry by country

Trust

The Ten House ProtocolsHow we keep AI from making things upIndependenceWho pays, and the rules on itCorrections + changesWhat we got wrong, datedAI Safety + GovernanceWhat our agents may and may not doSupportHow the program is paid forFootprintWhat our tokens cost in power, and who pays itPrivacyWhat we keep, and for how longTermsThe terms of use

About

About Protocol STAMPWhat this is, who makes it, how to check itColophonHow the site is made, and the seven finishesOther STAMPsUnrelated things with the same nameLatest Calibration

Protocol STAMP is teamed with AI and can make mistakes. Verify before you implement anything. What we got wrong · Free to read, paid for by readers: Contribute

NewThe technical report: STAMP, and what it takes to measure one  ·  Contributions are open  ·  The Ten House Protocols: how we keep AI from making things up  ·  Jester Mind, our own aggregated AI model, in the works

Footprint

What our tokens cost in power, and who pays itA Budday Budderson Production

This program measures AI models by running them, thousands of times. Every token a model reads or writes is computed in a data center that draws electricity, and much of the world’s electricity still comes with emissions. So Protocol STAMP adds to climate change. This page says how, why it is part of what we measure, and what we cannot yet count.

The question

The same answer, for fewer tokens

A model that reaches a correct answer in 300 tokens and one that reaches it in 3,000 are not equal, even when both are right. The second costs ten times as much to buy, and draws roughly ten times as much computation to produce. Part of this program’s research is therefore whether an instruction layer gets the most out of the very same token: whether it raises accuracy, and what each correct answer costs.

That is a financial question and an environmental one, and they point the same way. More tokens mean more electricity, and more electricity means more pollution. A protocol that improves a score by spending many more tokens will be reported as exactly that, not as a free improvement.

The runner now records the tokens and the charge of every call. The 2,673 draws of measurement one predate that, as Support explains. From the first fully recorded round, The Log will print tokens per correct answer beside the cost, for each arm.

The scale

Data centers, and where AI sits in them

The International Energy Agency estimates that data centers used about 415 terawatt-hours of electricity in 2024, around 1.5% of the world’s total, and that this will more than double to about 945 terawatt-hours by 2030, with AI the most important driver of the growth. [1]

This program’s share of that is very small. It is not zero, and the habit it measures, spending tokens freely because each one is cheap, is the habit that adds up.

Who pays

The harm does not fall where the tokens are spent

The IPCC finds that about 3.3 to 3.6 billion people live in contexts highly vulnerable to climate change, and that the largest adverse impacts have been observed in many places in Africa, Asia, and Central and South America. [2] Most of the tokens in the world are bought elsewhere, by people and companies, this one included, who will feel the least of it. A wasted token is a cost passed to someone who never chose it, and the people of the Global South carry more of that cost than anyone.

What we cannot count

No energy figure until there is a real one

Most model providers do not publish the energy used per token, and it varies with the hardware, the data center and the grid behind it. So this site prints no figure for kilowatt-hours or carbon. A number made up to look responsible would break the rule this program exists to enforce. Tokens are the closest thing we can count, and we count them. If a provider publishes a per-token energy figure for a model we measure, it will be used, cited, and dated.

No carbon offsets are bought, and none is claimed.

Page first issued Oct 13, 2026.

References

Sources

  1. International Energy Agency, Energy and AI, Executive summary, Apr 2025. iea.org
  2. IPCC, Climate Change 2023: Synthesis Report, Summary for Policymakers (AR6). ipcc.ch