跨平台分发战法Cross-Platform Distribution

Core Thesis: Not Reposting, but "One Fish, Two Dishes"

Bilibili and Douyin are not two publish buttons for the same clip. They are two machines with opposite logic:

• Bilibili: users arrive wanting to "actually understand something." They tolerate length, tolerate a slow build, and reward density of evidence plus a sense of the "uncanny." Completion rate and engagement drive the recommendation; long videos (8+ minutes) get their own traffic pool.
• Douyin: users are passively scrolling and get "intercepted." The first 3 seconds decide life or death; it rewards a strong hook plus a self-contained twist. Note: Douyin publicly shows like counts, not view counts, and its recommendation is already a composite ranking (completion / engagement / shares / follows). You cannot simply compare like counts against Bilibili's view-count magnitudes.

The asset of "Cambridge Tulingzi" is one core (a piece of textual-evidence reversal plus one cognitive hot-take). It should not be cloned into two clips, but split into two forms to feed the two machines. Principle: first use Douyin short-cuts to validate which hook lands, then return to Bilibili to turn the validated hook into a deep talk.

"One Fish, Two Dishes" Template (Example: Zhuangzi Drumming on a Basin / The Skull)

The same core — "Zhuangzi was not a recluse who'd seen through the world, but a man who observed death as data" — split across two columns:

Key point: the Douyin cut nails the single twist and stops — no evidence expansion; the Bilibili cut catches the people Douyin brought in and goes deep. Both share one research note: one cost, two outputs.

Iron Rules: Different Discipline for the Two Machines

Execution implication: any new topic is first tested as a 45–90-sec short cut on Douyin; if 1 of 3 breaks out, then invest in an 8–12-min Bilibili deep talk. Never reverse it — don't burn days shooting a long video on an unvalidated angle.

90-Day Cold-Start Cadence

Phase-linking logic: phase one solves "can you be scrolled to," phase two solves "can you be remembered (stable tag)," phase three solves "why should I believe you (persona)." The three cannot be reordered — building a persona before validation = self-indulgence; posting long videos before a tag = no pool to catch them.

Account Trade-offs: Who to Learn From, Who to Beware

Judgment criterion: learn from accounts that "win with their own brain," avoid accounts that "win with someone else's content." Tulingzi's moat is evidence + hot-takes, not voice and editing — so reposted-recitation channels, however good their numbers, are irrelevant to the persona and stay out of the reference pool.

One-Line Battle Discipline

Split one core into two forms; Douyin validates the hook first, every episode twists on its own; Bilibili goes deep later, escalating within a series; 90 days in three phases — first get scrolled to, then get remembered, finally get believed.

Note: account sizes and platform recommendation mechanics described in this chapter are experience-based judgments from public observation, not officially disclosed data; the "likes ≠ views, composite ranking ≠ like-based ranking" point on Douyin is common knowledge about the platform's mechanics — exact weights are not officially published, so defer to your own measured data in execution.