I asked four different AI agent setups the same question about two startups: how did Granola (granola.ai) actually grow, and how did Lovable (lovable.dev) actually grow? Each had to build a dated 24-month marketing history from free public data, show when growth stepped up, and pick three plays to copy and one to skip. Their reports are attached: granola-1 to granola-4 and lovable-1 to lovable-4. I'm not telling you which model wrote which. Be the judge. For each company, tell me which report a founder should actually act on and why, then rank all four. Say what each one found that the others missed, and where each is thin, unsupported or wrong. Check the evidence: are the numbers sourced, is the chart built from real data, are the dated moves real? Spot-check claims against public sources where you can. Some of the reports followed Corey Haines' reverse-engineer-marketing skill; say whether that made them more useful or just longer. Deliver one self-contained HTML page: a two-line verdict per company at the top, then the rankings and your notes. Keep it short enough to read in two minutes. Don't stop to ask me questions; make sensible assumptions and say what they are.