Surfaced while reviewing #232, which resynced docs/ with the current protocol spec. Two claims in the site FAQ describe mechanisms the protocol no longer has. Out of scope for that PR — AGENTS.md keeps docs/ and the website on separate pipelines — so tracking here.
What's wrong
lib/faq.ts:20 — "Account abstraction lets a publisher fund a payment channel for their audience".
Payment channels were replaced by a shared payment pool: one deposit per payer, opened once and reused across every node, with capped delegated signers and cumulative node-addressed vouchers. There is no per-pair channel to fund. The publisher-pays story itself still holds — it just runs on a pool whose capabilities are issued to the audience's keys, which is arguably a cleaner fit for the same answer.
lib/faq.ts:8 — lists "phantom availability, rate manipulation, blacklist violations" as the on-chain slashable offenses.
Phantom announcement was removed. There are two slashable offenses: rate manipulation and blacklist violation. Content corruption is absorbed at the wire (no voucher for an unverified window) and refusing to serve is explicitly not slashable, so neither belongs in that list either.
Why it matters
lib/faq.ts is prose-as-data per AGENTS.md, and app/llms-full.txt/route.ts serialises the same strings — so each error ships twice, once on the page and once in the LLM mirror. It now also contradicts docs.decdn.org, which #232 corrected.
Scope
Deliberately not in scope: content/blog/*.mdx. Six posts use the old terms, but they are dated artifacts and should read as of their publication date.
Surfaced while reviewing #232, which resynced
docs/with the current protocol spec. Two claims in the site FAQ describe mechanisms the protocol no longer has. Out of scope for that PR — AGENTS.md keepsdocs/and the website on separate pipelines — so tracking here.What's wrong
lib/faq.ts:20— "Account abstraction lets a publisher fund a payment channel for their audience".Payment channels were replaced by a shared payment pool: one deposit per payer, opened once and reused across every node, with capped delegated signers and cumulative node-addressed vouchers. There is no per-pair channel to fund. The publisher-pays story itself still holds — it just runs on a pool whose capabilities are issued to the audience's keys, which is arguably a cleaner fit for the same answer.
lib/faq.ts:8— lists "phantom availability, rate manipulation, blacklist violations" as the on-chain slashable offenses.Phantom announcement was removed. There are two slashable offenses: rate manipulation and blacklist violation. Content corruption is absorbed at the wire (no voucher for an unverified window) and refusing to serve is explicitly not slashable, so neither belongs in that list either.
Why it matters
lib/faq.tsis prose-as-data per AGENTS.md, andapp/llms-full.txt/route.tsserialises the same strings — so each error ships twice, once on the page and once in the LLM mirror. It now also contradictsdocs.decdn.org, which #232 corrected.Scope
lib/faq.ts:20— re-word the publisher-pays answer around the pool modellib/faq.ts:8— drop phantom availability; two offenseslib/copy.tsandlib/faq.tsfor other retired terms while in thereDeliberately not in scope:
content/blog/*.mdx. Six posts use the old terms, but they are dated artifacts and should read as of their publication date.