06-reference

not boring lonesome dove llm defragmentation

2026-09-08·reference·source: Not Boring·by Packy McCormick
llm-as-recommendercontent-distributiondiscovery-mechanicsevergreen-contentlindy-effect

"Everyone is Reading Lonesome Dove" — Packy McCormick

Why this is in the vault

Packy traces a 5-year, multi-platform diffusion chain (podcast → death of author → BookTok → Yellowstone tie-in → 40th-anniversary reissue) for a 1985 novel, then notices the actual proximate cause for his own and his friends' pickup was Claude and ChatGPT both independently recommending it — and argues LLMs are becoming a distribution layer that resurfaces "best" (lindy) content rather than "newest" content, unlike search (intent-driven) or social (recency-driven).

The core argument

Historically fragmented discovery (McMurtry's death, a New Yorker profile, BookTok virality, a Taylor Sheridan-branded reissue) explains why a 40-year-old novel keeps resurfacing, but Packy's own path in was neither of those — it was Claude recommending it unprompted when asked for a page-turner, then ChatGPT independently recommending the same book on unrelated queries. He frames this as a structural shift: search optimizes for "what you're looking for," social optimizes for "what's being talked about right now," and LLMs optimize for "what you (or people like you) would think is great" — which systematically favors durable, well-regarded back-catalog work over novelty, because that's what has accumulated the most and best signal to recommend from.

Mapping against Ray Data Co

Direct tension with Sanity Check's current distribution posture, which is built around social/feed mechanics (X threads, LinkedIn remix, hook-driven virality per feedback_x_voice_mismatch and the remix skill) — all recency-optimized channels in Packy's framing. If LLMs are becoming a live discovery layer that resurfaces "best," not "newest," that argues for treating Sanity Check pieces as durable evergreen assets worth being well-cited and well-structured for LLM retrieval (clean canonical markdown, strong internal linking, no derivative filler per feedback_no_derivative_sanity_check_pieces) rather than optimizing purely for launch-week social pop. It's a complementary distribution channel, not a replacement — but it's currently unmeasured and untargeted for RDCO surfaces, which is a real gap given the founder's four-layer targeting filter (project_targeting_system_prioritization_filter): there's no instrumentation today for whether Claude or ChatGPT would ever surface a Sanity Check piece to a third party.

Sponsorship

No third-party paid block in this issue. There is a house self-promo CTA up top for "not boring world" (Packy's paid subscriber tier — bigger Weekly Dose, an American-engineer-in-China series, off-the-dome notes, founder-co-written essays), flagged per the skill's rule that self-consulting CTAs get disclosed even when not disqualifying. This is a solo Packy essay — no co-author byline, so the "co-written essay = founder pitch" gotcha in the README doesn't apply here.

Related