Why isn't Anthropic eating their own dogfood? A Max subscriber's view

Anthropic tells every enterprise customer to use AI to transform operations and automate support. Their own support experience is a chatbot named Fin, a missing contact page, and search results that send paying users to Reddit.

I'm paying $200/month for Claude Max on my own dime, not my company's. I'm a Technology Director at a Fortune 50 company, using Claude personally to learn and then advocate for the right tools in our enterprise environment. That context matters for what follows.

Here's what's broken:

Claude Code OAuth fails repeatedly because the model Anthropic labels "Default (recommended)" — Opus 4.6 with 1M context — has a known OAuth refresh bug. Their own UI is pointing users at the broken option.

Claude Desktop throws "This isn't working right now" constantly. Power users send every other prompt twice, especially during business hours. 127 incidents in 90 days per StatusGator. No SLA, no credits, no advance notice when what's included changes. What I'm paying for seems to shift daily — that's a product management maturity problem, not a scaling problem.

The deeper issue: when errors fire, there's no diagnostic agent, no inline fix, no curated Anthropic knowledge base that Claude searches before hitting Google. No check of the real-time status board so users don't waste time. The tokens I spend troubleshooting their bugs count against my usage limit.

Anthropic is effectively profiting from its own poor support experience — customers pay for the service and then pay again to fix it. They're charging customers for the time it takes to identify the defective product they sold them.

I stopped using OpenClaw because of its instability. I didn't have time for the constant churn. Anthropic is starting to feel the same way — the difference is they're charging me $2,400 a year, which raises the bar considerably.

The people most frustrated right now are technical early adopters actively building on the platform. Anthropic is telling everyone to eat their own dogfood — to demonstrate AI's value by using it internally. They aren't doing it themselves. That's a credibility gap that compounds every outage.

If Claude is frustrating me on personal projects, I'm not going to advocate bringing it into an enterprise environment where the stakes are much higher. That's the retention problem Anthropic should actually be worried about.

Why isn't Anthropic eating their own dogfood?