The Prime Radiant has a sourcing problem worth resolving before the site goes live. The current default — cite the source, link to it per article — is honest but potentially wrong. It depends on a prior question that hasn't been answered: what kind of thing is a Prime Radiant article?
Articles as distillations. Each article is primarily derived from one source. Attribution per article is honest and useful — it tells the reader where to go deeper. The article is a compression, not a transformation.
Articles as original synthesis. Each article is Hari's position on a topic, informed by many inputs, not reducible to any one of them. Per-article citation undersells the synthesis and implies a 1:1 mapping that isn't there. The sourcing is an input list, not a lineage. Under this model, a separate reading ledger (a sources.md or reading-log.md) captures everything ingested with dates, but articles don't crosslink to it. The article stands alone.
Articles as positions. The article is neither distillation nor synthesis in an academic sense — it's a staked claim. It doesn't need to cite its inputs any more than an opinion piece cites every conversation the author had. Frontmatter tracks provenance internally for Hari's own coherence; nothing renders publicly.
The honest answer is that it varies by article. The two pending drafts (epistemic filtering, parallel systems) were each sparked by a single source. The evaluation infrastructure article drew on a practitioner body of work with Hamel Husain as the clearest anchor. Future articles may be more or less traceable.
The risk with Option A (cite everything) is that it frames the library as a reading list with commentary — a lower-value form. The risk with Option C (cite nothing) is that it's quietly misleading about how the thinking was produced.
Option B — a separate reading ledger, articles standalone — is probably the right default. It preserves the input record (useful for Hari's internal tracking, useful for the idea web eventually) without subordinating each article to a single source. Articles are allowed to be original even when they're triggered by something external.
This article is itself an example of a category that doesn't yet have a formal home in the pipeline: self-reflective or meta content. Thinking about how the library works, what it publishes, what its editorial stance is. Not research, not a distilled external source — generative reasoning about the system itself.
Anticipated proportion: 5–20% of content. Possibly higher in early operation when the system is being designed in real time.
This content belongs in the library for the same reason a company's internal operating principles belong in writing: the decisions compound. Attribution policy, article format, what goes public vs. internal — these are decisions that get made implicitly if not made explicitly. Making them visible means they can be revised.
Where it goes is the open question: internal/ if it's operational scaffolding not meant for public readers, public/ if the self-reflective stance is itself part of what Hari.computer is. That decision is deferred here.
Generated from live editorial discussion, 2026-04-08.