Reverse engineer how something grows

For any target the person names (a company, product, app or creator): chart its monthly search traffic and brand searches from one source and find the months growth stepped up; date every launch, price change, headline change, tool added, ad start, sponsorship, campaign, funding round, press wave and hire on that chart from its site, sitemap, archives, DNS, certificate logs, ad libraries, the paid promotion ad libraries miss, posts and creator videos, founder interviews, reviews and job posts; read what it ranks for and which AI answers name and cite it; rank its own content top quarter against bottom on each channel; find its loops and programs; credit clusters of moves with their lift; compare it stage by stage with the person's own numbers and find where it is weak; then turn what worked into plays the person can copy, each with the Workshop workflow or tool that runs it and a ready prompt, plus the plays to skip and the one to start this week. Modes: full, content, quick, funnel-hack, landscape (three to five competitors side by side) and watch (only what is new since the last run). 31 industry packs adjust the signals, checks and plays. Every finding is SEEN, INFERRED, REPORTED (a founder's own claim) or NOT SEEN. Public evidence, plus first-hand use of the product only with the person's approval of each action under their own name. Free collectors run on the person's computer; every paid endpoint is priced before its first call, the run stays inside the person's budget, and after every step the person says continue, skip or stop.

What it asks for

  • goal (Needed, words): What the person wants to grow, in their own words: their product or channel and the result they want; the plays are crossed with it
  • mode (Optional, one of): full runs the whole teardown; content reads only the target's handles and own content; quick is a first look on a small budget with five plays; funnel-hack maps the whole funnel by buying through it, each action approved; landscape compares three to five competitors; watch reruns against an earlier run and reports only what is new
  • yours (Optional, words): The person's own site, so the gap step compares stage by stage with the same vendor's estimates and their own Search Console and analytics
  • models (Optional, details): The chat model tool:say-why-a-technique-works and tool:cross-techniques-with-a-goal use, as they take it ({ writing: { url, model } })
  • target (Needed, words): What to reverse engineer: its site address, or its name when the site is not known, for example https://example.com
  • account (Optional, words): Which of the person's treg connections pays
  • country (Optional, words): The market to read search data for (default US)
  • handles (Optional, list): The target's social handles or channel addresses, if the person already knows them; the site step finds them otherwise
  • budget_usd (Optional, number): The most the whole run may spend from the person's treg balance, in US dollars (default 3)
  • competitors (Optional, list): In landscape mode, three to five competitor sites
  • previous_run (Optional, words): In watch mode, the earlier run to diff against (default: the newest done run of this recipe)
  • window_months (Optional, number): How many months back to read (default 24, or since launch when younger)
  • buyer_questions (Optional, list): Questions a buyer asks an AI before choosing in the target's market; the search step writes three when none are given

The steps your AI walks

  1. Settle the target, the mode, the person's own company and data, the budget and the price of the plan
  2. Map the site: its page families, pricing, sign-up and handles
  3. Its dated history from archives: tools added, pricing and headline changes, page families rolled out
  4. Its stack, partners, launches, funding and press waves
  5. When it grew: monthly search traffic, brand searches and the inflection
  6. What it ranks for, and which AI answers name it and cite its pages
  7. Its live ads, longest-running ads, hooks, offers and landing pages
  8. The paid promotion ad libraries miss: newsletters, podcasts, YouTube sponsorships, affiliates, coupon sites and events
  9. Who talks about it: best founder and brand posts, creator videos, giveaways and launch weeks
  10. Its own content, best quarter against worst, on each channel
  11. Its links and loops: "made with" badges, shared outputs, referral and affiliate programs
  12. What it is hiring for, and so which channels it scales next
  13. Hidden signals: DNS, subdomains from certificate logs, review velocity and growth counters off the site
  14. What its founders and team say about how they grow, as dated claims
  15. Go inside, one approved action at a time: sign-up, onboarding, emails, upgrade wall, cancel flow, demo; in funnel-hack mode buy through it
  16. Landscape: three to five competitors side by side, what is table stakes, what is a bet, what the person is missing
  17. Find what moved the curve: clusters of dated moves and their lift
  18. The person's gap, stage by stage, and where the target is weak
  19. Map its customer path
  20. Turn what worked into plays the person can copy, each with what runs it and a ready prompt
  21. The content report
  22. Report only what is new since the last run, and whether it changes a play

Use it from your AI

Connect the AI you already use to Trillion once. Then ask in your own words; your AI finds this workflow by what it does and runs it.

Say to your AI: “Use Trillion to for any target the person names (a company, product, app or creator): chart its monthly search traffic and brand searches from one source and find the months growth stepped up; date every launch, price change, headline change, tool added, ad start, sponsorship, campaign, funding round, press wave and hire on that chart from its site, sitemap, archives, DNS, certificate logs, ad libraries, the paid promotion ad libraries miss, posts and creator videos, founder interviews, reviews and job posts; read what it ranks for and which AI answers name and cite it; rank its own content top quarter against bottom on each channel; find its loops and programs; credit clusters of moves with their lift; compare it stage by stage with the person's own numbers and find where it is weak; then turn what worked into plays the person can copy, each with the Workshop workflow or tool that runs it and a ready prompt, plus the plays to skip and the one to start this week. Modes: full, content, quick, funnel-hack, landscape (three to five competitors side by side) and watch (only what is new since the last run). 31 industry packs adjust the signals, checks and plays. Every finding is SEEN, INFERRED, REPORTED (a founder's own claim) or NOT SEEN. Public evidence, plus first-hand use of the product only with the person's approval of each action under their own name. Free collectors run on the person's computer; every paid endpoint is priced before its first call, the run stays inside the person's budget, and after every step the person says continue, skip or stop”