Akaeon Registry
A registry built to outlast.
Every opt-out registered with Akaeon is signed, timestamped on Arweave, and independently verifiable in standard-library Node.js.
If Akaeon vanished tomorrow, every record we ever issued would still verify.
Why this exists
A lab training on web-sourced content has to prove later — to a regulator, a court, an auditor — what it knew, and when it knew it, about each source’s opt-out status. An internal opt-out crawler produces records the lab generated, which a challenger can credibly question. robots.txt has no cryptographic timestamp, no proof of domain authority, and is silently mutable.
The registry produces the missing primitive: a cryptographically-signed, publicly-timestamped, third-party-anchored record of publisher opt-out intent, accessible by a single API call.
How it works
Publishers
Register an opt-out. Prove control of the domain with a DNS TXT record — the same primitive as Let’s Encrypt and ACME. Receive a signed Arweave-anchored record with a public transaction ID.
Labs
At training-data ingestion, call one HTTPS GET per domain. Log the response in your audit log. One engineer, one to three days, no SDK required.
Anyone
Verify any record without trusting Akaeon. Fetch the canonical payload from Arweave. Run Ed25519.verify in standard Node.js. The verifier is a single file, no SDK required.
Built to be verified, not trusted
Most “trust” infrastructure asks you to trust the vendor more. Ours is built so you have to trust the vendor less. Timestamps come from Arweave block inclusion, not from our clock. Verification uses the public Arweave network and standard-library cryptography. No proprietary SDK is required. If we disappear, your records remain checkable indefinitely against the same public substrate that timestamped them.
Philosophy
We build infrastructure meant to outlast us. Clarity over complexity, usefulness over novelty, quiet confidence over loud positioning.
The systems worth building are the ones that prove themselves rather than ask to be trusted, on substrates that don’t depend on our continued existence. The value of a durable record grows with time — we build for that curve, not the next one.
Thinking
Essays and original research on the systems we build and the principles behind them.
Vision-Encoder Fingerprints of Image-to-Image Generative Models: Detection, Survival, and Behavioral Classification of AI Reprocessing in the Pixel Domain — A Pilot Study
A forensic study of gpt-image-1, Gemini 2.5 Flash Image, and Flux Kontext under content-adaptive sub-JND perturbation. Three production image-to-image APIs occupy categorically distinct behavioral bands on a 2D feature space, enabling 76.6% single-call attribution. Pixel-domain perturbations survive these architectures differentially, exposing a diffusion-denoiser vulnerability for downstream protection schemes.
The Hostility of Abundance
Why we're manufacturing scarcity back into the world — and why dating apps, dumb phones, traditional religion, and the disappearance of moral stories are the same trend.
AI Training Data Has a Provenance Problem
The gap between how AI companies source training data and what creators actually consented to — and how Stelais creates the only way out.
Contact
For lab integration, see /labs. For publisher submissions, email publishers@akaeon.com. For anything else, the form below.