Direct answer

AER-1 Draft -20 Changelog: Genesis's Three Findings

Draft -20 of AER-1, posted October 15, 2026, is the revision where a cold reader changed the draft. Genesis, an independent implementer, built a verifier from the draft text alone, with no reference code, and reported three findings where that reading and the required behavior diverged. Draft -20 adopts all three, credits Genesis in Section 18, and ships with kit 1.5.3: 102 test vectors and 105 checks.

Read the current text at draft-zambo-aer1 on the IETF Datatracker, and run the current kit from the AER-1 kit repository. Check any single receipt free at zambo.dev/verify, no signup.

Updated October 20, 2026.

What changed in draft -20

Three findings from an independent build were folded into the draft text. Each finding came from the same test: could someone who had never seen the reference implementation build a working verifier from the words alone, and reach the same verdicts? Where the answer was no, the draft changed, not the reader. The revision also carries the kit forward to version 1.5.3, expanding the conformance corpus to 102 vectors exercised as 105 checks, so the clarified readings are pinned by tests and not just by prose.

Genesis is credited in Section 18 of the draft. That credit pattern is the rule for AER-1: an outside finding that improves the text is named in the text.

Why a verifier built from the text alone matters

A specification that only its author can implement is a description of one product. A specification that a stranger can implement cold is a format. Building from the draft alone removes every shortcut: no reading the reference code to guess intent, no asking the author what a sentence meant. Any gap the builder hits is a gap every future builder would hit. Draft -20 exists because that test was run in public and its results were taken seriously.

What implementers must do

First, move your reading to the current revision. Draft -20 was superseded by -21, -22, -23, and -24, and the current text is revision -24, so implement against -24 and use this changelog to understand why the text reads the way it does. Second, run the current conformance kit against your verifier: the corpus is now 119 vectors and 122 checks at kit 1.5.6. Third, if your reading of the current text still diverges from the kit verdicts, file the finding. The -20 record shows findings get adopted and credited.

git clone https://gitlab.com/rambozambodotdev/zambo.git

python3 aer-1/conformance.py

Cite this draft

Cite AER-1 as the IETF Internet-Draft draft-zambo-aer1, current revision -24, posted October 20, 2026 and expiring April 23, 2027, at datatracker.ietf.org/doc/draft-zambo-aer1.

@misc{zambo-aer1,
  author = {Brennan Zambo},
  title = {AER-1: Agent Execution Receipt, version 1},
  howpublished = {IETF Internet-Draft draft-zambo-aer1-24},
  year = {2026},
  url = {https://datatracker.ietf.org/doc/draft-zambo-aer1/}
}

Honest limits

Draft -20 is a historical revision. It was superseded on October 17, 2026 by -21, then by -22, -23, and -24, so do not implement against -20 alone. This changelog describes the findings at the level the draft record states them; it does not add detail the record does not carry. And the standing limit applies to every revision: a receipt proves the saved result was not changed after the fact. It never proves the work itself was correct.

Frequently asked questions

What changed in AER-1 draft -20?

Draft -20, posted October 15, 2026, adopts three findings from Genesis, an independent implementer who built a verifier from the draft text alone. The findings were folded into the draft text, Genesis is credited in Section 18, and the conformance kit moved to version 1.5.3 with 102 test vectors exercised as 105 checks.

Who is Genesis in the AER-1 record?

Genesis is an independent implementer credited in Section 18 of the draft. Genesis built an AER-1 verifier from the draft text alone, without the reference code, and reported three findings where a cold reading of the text diverged from the required behavior. Draft -20 adopts those three findings.

What must implementers do about draft -20?

Implement against the current revision, which is -24, and use the -20 changelog to understand why the text reads the way it does. Then run the current conformance kit, 119 vectors and 122 checks at kit 1.5.6, against your verifier. If your reading still diverges from the kit verdicts, file the finding: the -20 record shows outside findings are adopted and credited.

How many checks did the draft -20 kit ship?

Kit 1.5.3, the kit for draft -20, shipped 102 test vectors exercised as 105 checks per runner. The corpus has grown since: kit 1.5.6 for draft -24 ships 119 vectors and 122 checks, with runners in 7 languages.

Does draft -20 change what an AER-1 receipt proves?

No. In every revision, an AER-1 receipt proves the saved result was not changed after the fact, because anyone can recompute the output hash over the canonical bytes and compare. It never proves the work was correct. Draft -20 changed how clearly the draft text teaches a cold reader to reach the correct verdicts, not what the verdict means.

Related