Field note
Cross-domain links are the product
A page by itself is a summary. What makes it worth anything is what it points to - the claim in a podcast that turns out to connect to a claim in an article nobody would have thought to put next to it.
The page is not the point
It is tempting to treat a compiled page as the deliverable - a clean write-up of one concept, sourced and readable. It is not nothing. But a single page, isolated, is just a better summary of one source. You could have gotten most of that by reading carefully.
The value shows up one layer up: in what the page links to. A page that names the other concepts it touches, and points to their pages, is doing something a summary alone cannot - it is placing one piece of knowledge next to another.
A concrete shape
Say a podcast guest makes an offhand claim about how a technique fails at scale. Separately, an article you read months ago makes a claim about a specific failure mode in a different system. Neither source mentions the other. Neither claim, alone, is the insight.
The connection is the insight. The podcast claim and the article claim describe the same underlying mechanism from two unrelated domains, and only because both got compiled onto pages that link to a shared concept does that connection ever surface. Read either source alone and you never see it.
Only a linked network surfaces this
A folder of transcripts cannot make that connection - nothing in it points anywhere. A single summary per source cannot make it either, because a summary stays inside the boundary of its own source. The connection lives at the link, not inside either page.
This is why the method insists on one concept per page and mandatory links to related concepts, even when it would be faster to just write a paragraph and move on. The link is where the cross-domain claim gets built - skip it, and every source stays an island next to every other island.
The non-obvious link is the one worth having
An obvious link - two episodes of the same podcast on the same topic - is useful but not surprising; you'd have found that by scrolling the feed. The links worth building the whole method around are the ones between sources that share no context: a podcast and a research paper, an interview and an old blog post, two domains that never cite each other.
A compiled, linked network is what makes that kind of link findable at all. Nobody sits down and manually cross-references every transcript against every article. The network does it because every page carries pointers to every related concept it touches - and the link that matters is usually the one you weren't looking for.