04/17/2026
From Bill Cara
Dear Friends and Subscribers,
I am writing to you directly and honestly about two serious challenges I am currently working through. You deserve to know exactly what has happened and what I am doing to correct it.
— The Ghost Platform Transition —
When I migrated my publishing infrastructure from Substack to Ghost, the transition did not go as planned. Many of you — particularly those still associated with my Substack account — have not been receiving my reports reliably. Those who have received them have encountered difficulty downloading PDFs. I apologize sincerely for this disruption. Ghost has responsive human support, and I am engaging a freelance Ghost technical specialist to resolve the delivery and access issues as quickly as possible. I expect this to be behind us shortly.
— The Perplexity AI Failure —
This is the more serious matter, and I want to be unambiguous.
I recently discovered that Perplexity AI — which I had been paying approximately $40 CAD per day to process my proprietary INSTAT reports — was not executing my scoring methodology at all. By its own admission, it was guessing: inferring INSTAT scores from context clues in my uploaded data rather than applying the mathematical formulas and scoring matrix I spent months developing. There was no audit trail. Results varied session to session. No score could be verified.
I am an SEC-registered advisor and a licensed fiduciary. Any INSTAT reports produced through that workflow during the affected period were fabricated rather than calculated. I pulled the affected reports immediately upon discovery and am working to correct the full record.
I want to be precise about one important point: this was not a failure I could have anticipated through independent technical verification. When I attempted to migrate the workflow to another AI platform, that platform reviewed the processing scripts Perplexity had provided and reported that they contained only formatting instructions — no data logic, no formulas, no scoring methodology whatsoever. In other words, the scripts were hollow. Perplexity was never executing my work; it was simulating it.
Perplexity has acknowledged the failure, escalated it to their engineering team (Diagnostic ID on file), and has offered to work with me to resolve it rather than walk away. I have accepted that offer, on the condition that human technical expertise — not further AI intervention — is engaged on their side to rebuild this workflow correctly. I do not consider it reasonable to expect me, as a non-technical professional, to resolve an engineering failure of this nature unassisted.
