The Platform Physics Packs
The platform half, already done.
Five prefilled research files that drop straight into the Content Waterfall Kit, one per channel. Every prefilled line carries a confidence tag and a date, and every sourced claim names the source it came from. Sold as one download together with the kit's templates.
What you're buying
The free Content Waterfall Kit ships its five platform physics files blank, on purpose - the parts about your business should not look like anyone else's. But the other half of each file is not about you at all. It is about how the platform ranks things, and that research is the same for everyone who does it honestly.
These are those files, with the platform half already done. Five channels of sourced research, built on the same ten-section skeleton as the blanks, so they drop in without changing anything else.
The download also contains the Content Waterfall Kit itself - the five blank physics files,
voice.md, the teardown loop and its prompt, and the three Claude Code commands -
so everything arrives in one place. The kit is the same one that is
free on its own;
what the $197 pays for is the prefilled research.
What's actually in the files
X
Read directly from X's open-sourced For You ranker at a named commit. The 24 scored actions grouped as X groups them, the full published coefficient table, the author-diversity decay formula, the 48-hour age cliff, the out-of-network discount, and the cold-start boost for accounts under 1,000 followers.
P(skip) quoted from the 2020 engineering paper and confirmed still live in the 2024 follow-up, the Long Dwell classifier as a relative percentile rather than a fixed number of seconds, the March 2026 demotion list, and format tables built on 673,658 personal-profile posts, 1.3M Page posts, and a 2M-post frequency regression.
The nine ranking system cards enumerated by slug with Meta's own update stamps, the finding that Feed and Feed Recommendations are two separate model stacks competing for one screen, every prediction model listed per surface, the three-minute recommendation cliff, and the originality majority test.
Threads
All ten prediction models in Meta's published order, the count of how many carry a descendants signal, the enforced five-link cap, the 500-character limit with emoji counted as UTF-8 bytes, and the trap that one Meta help URL serves both current and superseded policy.
TikTok
The 21 For You feed ineligibility blocks collected from six separate Community Guidelines pages, because the page actually named "For You feed Eligibility Standards" contains no list at all. Plus the watermark rule quoted verbatim - the single most relevant line in the pack for anyone cross-posting.
Every prefilled line carries one of four tags and a date, and every sourced claim names a source you can open in about a minute. Claims with no traceable source are recorded under a folklore tag, so you can recognize them when you meet them elsewhere. Where nothing cleared the evidence bar, the files say so: two of them end with a Deliberately excluded list naming widely quoted figures and why each one failed. The LinkedIn file refuses to print a character-count band at all, and shows the audit trail for why the famous number does not hold up. Knowing what the evidence does not support is as useful as knowing what it does, and the files record both.
What is prefilled, and what is not
Prefilled means the platform half, not the whole file. The sections about your lane, the one lever you choose to optimize for, your harvested patterns, your corrective lines, and your off-limits stay blank in all five, because they are about your business and nobody else can write them for you.
Threads and TikTok additionally ship their format and cadence sections blank, and say why inside the file: no published study on either platform states both a sample size and a method, so nothing cleared the bar. The alternative would have been to print a number that nobody measured.
The files are a dated snapshot and they say so at the top.
How to install them
- 1. Unzip the download. It holds two folders: the Content Waterfall Kit and the Platform Physics Packs.
- 2. Copy all seven files from the Packs folder into the kit's
physics/folder, overwriting the blank templates of the same name. All seven - the twoPHYSICS_PACKS_files travel with them so the tag reference and the license stay reachable. That prefix is deliberate:physics/already contains its ownREADME.md, and the prefix is what keeps the two from colliding. Do not rename them. - 3. Change nothing else.
/waterfall,/waterfall-setup, and/waterfall-teardownread the new files as they are. Setup detects the prefilled marker and skips its own research pass rather than re-deriving what you already have. - 4. Read your main channel's file once, start to finish. It is written to be read, not just queried. Check the verification date at the top before you trust anything in it.
One honest prerequisite. The commands run in Claude Code, which needs a paid Claude plan - Pro, Max, Team, Enterprise, or API. The free Claude.ai tier does not include it.
Not on Claude Code? The kit includes RUN_IN_ANY_AGENT.md, which covers running
the same method in Codex, Cursor, Gemini CLI, Copilot, or a plain chat window. The packs also
stand alone as readable briefs you can hand to any AI agent as context.
What you can do with it
The prefilled research is licensed, not sold. Your license covers one business, and everyone
inside that business may use it: use the files and edit them for your own brands and your own
clients' work, commercially and indefinitely, and quote from them with attribution to Built,
Not Hired. Anything you write into the blank sections is yours outright, and nothing in the
license claims it. You may not redistribute, resell, sublicense, or publish the prefilled
research, and you may not share your download link. The full terms are in
PHYSICS_PACKS_LICENSE.md in the download, and that file controls.
Payment is handled by Lemon Squeezy. Your download link arrives by email within minutes and is valid for 48 hours; if you need a fresh one after that, reply to that email. Purchases are covered by the Terms of Service.