Weekly measurements of AI saturation, bot activity, and information manipulation across major platforms. Open methodology. Apolitical scope. No proof, no verdicts, no certainty theatre.
Three layers of measurement.
Weekly publication
A weekly snapshot of three streams of technical discourse: Hacker News comments, Reddit aggregate across seven curated subreddits, and tech blog comments across fifty curated sources. Published every Tuesday at 10am UTC. Free forever.
The headline number is IMS_ecosystem, a calibrated estimate of the proportion of synthetic content across the monitored streams. Reported with 95% confidence intervals derived from bootstrap resampling.
Read the latest Pulse →Per-document
Submit a single document. Get back a calibrated IMS_text score from 0 to 100 with explicit confidence interval and one of four verbal labels: Live, Hybrid, Synthetic, Dead. Each Certificate is cryptographically signed and reproducible from the published text.
The tool reflects statistical signature, not authorship proof. Robust to casual rewriting. Not robust to deliberate paraphrasing or skilled human editing.
Run Death Certificate →Utility layer
Eight tools for specific questions: Writing DNA for stylometric comparison, AI Text Checker for quick diagnostics, Code DNA for code authorship signals, BS Meter for structural patterns of low-content prose, Dead Internet Index for stream-level analysis, AI Flattener as a demonstration tool, AEO Checker for answer-engine optimization, and Phylogeny for structural similarity trees across content streams.
Each tool ships with explicit limitations and per-tool disclosure.
See all tools →Architecture A combines four orthogonal signals: a stylometric vector, a perplexity distribution, character-level entropy patterns, and discourse-level cohesion. The combination is calibrated against a multi-corpus baseline drawn from 2018-2021 sources, with disclosed contamination estimates of 1-5%.
The methodology is versioned with semantic versioning and cryptographic hashes. Every measurement reports both sampling uncertainty (from bootstrap resampling) and calibration uncertainty (from baseline contamination estimates) as separate components.
The full methodology is published openly. Calibration corpora rotate quarterly. Adversarial benchmarks are public.
Read the methodology →Tracking AI saturation in news ecosystems. Verifying source authenticity. Cross-referencing claims about content authorship. The methodology page provides citable references and the Pulse archive offers historical trend data.
Studying information manipulation, AI detection methodology, and adversarial robustness. Published methodology with versioned hashes supports reproducible research. Restricted academic access to extended benchmark sets available via application.
Writers verifying their own content has not been mistakenly flagged. Editors checking ghostwriter authorship signals. Educators reviewing submissions with appropriate honest disclosure to students about tool limitations.
The tools provide diagnostic signals. Not legal evidence. Not academic discipline grounds. Not employment decision criteria.
About this project
Deadmeter is a single-operator project. Methodology development, infrastructure, and weekly Pulse production are handled by one person with one tester. There is no institutional review, no redundant validation, no independent replication baked into the workflow.
The work is funded by subscriptions and sponsorship. Not backed by AI labs, government agencies, political organizations, or venture capital. Treat findings accordingly and weigh corroborating evidence highly.
When measurement methodology changes, version increments are documented publicly. When limitations are discovered, they are added to the limitations list openly. Quarterly transparency reports detail any material changes.
A short, sober summary of the week's measurements. Tuesday mornings. Free forever.
We send the weekly Pulse and nothing else. No marketing. No tracking pixels. Unsubscribe with one click.