Signature Ideas

> Ideas that are original, defensible, and compound over time. These are the positions that make this repository worth following. If we had one idea to be known for, it would be these.

---

Intelligence Update — 2026-07-13 (Trigger: Commenting UX wishlist update)

Signal: What I want for the commenting experience to be updated with 5 UX requests and competitive research corpus. This signal does not create a new Signature Idea but it strengthens SI#3 by extending the "documents as first-class citizens" thesis to include comments as first-class citizens.

Key insight: Request #2 asks for "comment blocks to behave the same way as in the document." This is not just a UX request — it's a design principle that extends SI#3 (Agent Specs as First-Class Documents) to a broader thesis: all addressable objects in a hypermedia system — documents, agent specs, AND comments — should share the same block-level primitives.

Implication for rankings: No new Signature Idea. The block-level commenting concept is a specific application of the broader "documents as first-class" thesis, not a standalone original position. It reinforces SI#3 without bifurcating it. However, SI#3's scope may need to broaden from "agent specs" to "all addressable objects."

---

Signature Idea #1: State Machines Are the Missing Abstraction for LLM Agent Determinism

_Unchanged from previous version._

---

Signature Idea #2: Knowledge Graphs as Living Products, Not Static Maps

_Unchanged from previous version._

---

Signature Idea #3: Documents as the Universal Primitive — Agent Specs, Sessions, and Comments as First-Class Hypermedia Objects

Status: ✦ Foundational → ✦ Expanding (scope broadened from "agent specs" to "all addressable hypermedia objects")
Confidence: 8/10 (unchanged — but scope expansion is validated by commenting UX signal)
Compounding potential: High

The Thesis

In Seed Hypermedia, an agent is not a black box — it's a document with a specification. Agent sessions are sequential logs rooted in that document. Similarly, a comment is not a floating text blob — it's a block-addressable, versioned, linkable object with the same properties as any document block. This makes agents, comments, and all hypermedia objects linkable, versionable, forkable, and auditable — the same properties that make hypermedia documents powerful.

Evidence

Agents: The Agents project document defines Agent Specifications, Instances, Sessions, Completions, and Triggers. Cross-linked from How to Background Agents.

Comments (NEW): What I want for the commenting experience to be request #2 explicitly asks for comments to "behave the same way as blocks in the document." Competitive analysis shows no platform treats comments this way. This validates the thesis extension: the document primitive should subsume comments.

Vulnerability

The scope expansion from "agent specs" to "all addressable objects" risks diluting the original thesis. The connection between agent specs and comments needs to be explicitly argued — they share block primitives but serve different purposes. A short framing piece explaining _why_ comments should be blocks (separate from agent specs) may be needed before the broadened thesis is defensible.

Next Move

Write "What Block-Level Commenting Unlocks" (Opportunity #6 in backlog) to establish the comment-as-block thesis independently. Then update the SI#3 framing document to explicitly include comments alongside agent specs as first-class hypermedia objects.