I got rejected by Anthropic today.

Yesterday I had a technical leadership screen.

Scheduled from 17:00 to 17:25. The recruiter arrived at 17:07, and I had to send an email to find him.

He was kind. But he was also running a machine-paced script: six questions delivered at very high speed. So fast I had to lean in to keep up. He was done at 17:20. 14 mins in call. 5 minutes early on a 25-minute call.

An AI interview intelligence platform (BrightHire) was active in the meeting. Transcribing my answers in real-time. Prompting the next question. The recruiter had 10+ years in SaaS recruiting, not AI architecture. Claude was the only evaluator on the call technically capable of assessing my answers. And it was scanning for keyword matches, not trade-offs.

When he (they?) asked about reducing hallucinations in enterprise deployments of CapabiliSense, I answered at the architectural level. We didn't tune model parameters. We built a deterministic rules engine behind the LLM, semantic tagging by the model, conflict resolution by the engine. Built for compliance boards that need verifiable logic, not some tunable probabilistic outputs.

My theory: the AI scored it as a NO-GO after the 6th question, closed the prompt and handed control back to the recruiter at 17:20 with a standard wrap-up script :

“I’m going to take it back to the hiring manager and come back to you on Wednesday.”

The recruiter didn't invite my questions. I inserted myself anyway.

I asked what success looks like at 90 days. His answer:

"drive impact, getting in front of customers at an early stage, adapt to speed and pace, and the access to tools internally versus externally."

No milestone. No metric. No structure.

I asked how the technical deployment lead interacts with the sales team day-to-day. His answer:

"there's a very close alignment to the sales team... they're pretty much living in each other's pockets."

No cadence. No OKR. No named mechanic. Pure filler.

The man hired to screen candidates for a technical deployment lead role could not define success for that role or how it functions operationally. His Claude assistant had run its script. What followed was improvisation.

I'm not angry, I'm recalibrating. The only path for operator-grade hires at frontier AI companies is bypassing the transcription layer and going direct to the P&L owner.

Or skipping this bizarre process altogether.