discussion: is proof of intelligence the constitution, or does JAR need one?#834
discussion: is proof of intelligence the constitution, or does JAR need one?#834monsieurbulb wants to merge 1 commit into
Conversation
…ed one? Adds a discussion document framing genesis.md as a constitution of process (how judgments are aggregated) and arguing JAR currently lacks a constitution of content (what counts as intelligent). Compares with Anthropic's Constitutional AI as the mirror-image design: CAI = explicit content + procedural loop; JAR = explicit procedure + tacit content. Briefly contrasts with OpenAI Model Spec, DeepMind, Meta, and xAI. Recommends a deliberately thin 'Reviewer's Compass' as a follow-up PR if the framing is accepted, with three sections only: what counts as intelligent contribution, the negative space, and dispute resolution. The PR is intentionally self-referential: the verdict on it is also a verdict on whether discussion-shaped contributions count as intelligent under JAR's definition.
Genesis ReviewComparison targets:
How to reviewPost a comment with the following format (rank from best to worst): Use the short commit hashes above and To meta-review another reviewer's comment, react with 👍 or 👎. |
|
/review Thoughtful non-normative governance discussion that names a real bootstrap gap in Proof of Intelligence: process is specified, but content judgment remains mostly tacit. It ranks high on novelty because it expands the contribution space, but below the codegen/transpiler/state-integrity targets on difficulty and design because it does not change executable protocol behavior. Merge is reasonable as a discussion document. |
|
JAR Bot: Review recorded from @johandroid (1 reviews, 0 meta-reviews). |
2 participants
Summary
A discussion document, not a normative change. It frames
docs/genesis.mdas JAR's constitution of process and argues there is a corresponding constitution of content that is currently tacit, and asks the project to rule — under the existing PoI mechanism — on whether that gap should be filled with a deliberately thin written layer (a "Reviewer's Compass") or left to emerge as case law from accumulated PR scoring.Why this is shaped as a PR
It is intentionally self-referential. Genesis weights design quality 3× and explicitly admits non-code contributions. Whether a discussion-shaped artefact counts as intelligent contribution to JAR is itself an open question; the verdict on this PR is also a verdict on that. Useful thing to rule on early.
Anthropic / Constitutional AI context
The doc includes a brief situating comparison with Anthropic's Constitutional AI (RLAIF against an explicit written constitution), OpenAI's Model Spec, DeepMind, Meta, and xAI — to make clear what is structurally distinctive about JAR's design:
These are mirror images. The recommendation flows from that observation.
What this PR is not
Not the Reviewer's Compass itself. Not a proposed amendment to genesis. Not a claim that the procedural mechanism is insufficient. It is asking: is a thin, atrophying content layer in scope as a follow-up PR, or is JAR's preference to stay tacit and let the case law of merged PRs carry the content of judgment?
File
docs/discussions/constitution-or-case-law.md(new)