Point your AI here
Original reusable templates for your own agents. Copy into Grok, Claude, GPT, Cursor, Codex, Devin, Hermes, or similar to research, verify, or sketch a careful experiment.
No earnings guarantees. These prompts are not financial advice, not proprietary playbooks copied from third parties, and not a promise that any case amount is repeatable. They never invent cases or dollar figures for this library.
Placeholders like {display_name} are for you to fill from a public case card (product / verified handle — not a personal full name).
Search X and the open web for people who publicly posted AI-agent revenue with a screenshot or TrustMRR/Indie Hackers link in the last 30 days. For each find: display_name (handle or product — never a personal full name), amount+kind+timeframe, tools named, and the exact proof URL. Skip anything without a dollar figure.
Given this case summary: {display_name} claims {amount} ({kind}) over {timeframe} using {tools}. Open the cited proof URLs (or archives). Quote the exact amount line you see. Say whether confidence should be verified, self-reported, or unverified under a processor-connected third-party rule. Do not invent numbers.Using only public information about {business_type} businesses that use {tools}, outline a 7-day experiment to test a similar offer with under $200 budget. Include channel, offer, price point, and what evidence would count as success. No fake case studies.Scaffold a local Python script that appends JSONL rows with fields id, display_name, amount_usd, proof_urls, confidence, collected_pt. Include a function that rejects rows without https proof URLs. No network calls in tests; use fixtures.
From a list of AI earnings claims with amount_usd and timeframe, group into $0–1k, $1–10k, $10–100k, $100k+. Flag any row where timeframe is one-off but amount_kind is MRR. Return a markdown table.
Write a parser that, given saved TrustMRR HTML, extracts current MRR, L30D, all-time, last sync, and whether the Stripe connection looks expired. Output JSON. Never guess missing fields — use null.
For each proof URL that returns 404/410, find an archive.org closest snapshot and propose updating proof_urls while keeping the original URL in notes. Do not change amount_usd.
Draft a 5-bullet digest from newly dated AI money cases: display_name, $ + timeframe, tool stack, confidence badge, one replicable note. Tone: skeptical, evidence-first. No hype verbs.
Given counts of cases per agent tab (Grok Bot, Claude Code, Cursor, Codex, Devin, Hermes, …), list the three thinnest tabs and suggest public search queries to find real dollar claims for each.
Design an n8n flow: RSS/X keyword hit → human approval → append a draft case JSONL line with confidence=unverified until a proof URL is attached. Include a fail path when amount is missing.