Anthropic's trust problem is becoming a real business risk

Two separate threads are eating Anthropic alive today. The first is a discovery that Claude Code appears to be steganographically marking API requests, apparently to fingerprint resellers and catch model distillation. The HN discussion is harsh: one commenter notes the technique mostly punishes 'normal developers doing weird but legitimate things' rather than actual bad actors. Another says flatly, 'The more I learn about Anthropic the more they disgust me.'

The second thread is about Claude Fable 5 remaining unavailable for nearly a month, with Anthropic saying almost nothing about when it returns. Builders are frustrated because they made product decisions around Fable and now have no timeline, no clear explanation, and no recourse. Anthropic keeps launching new products (Claude Science, the export control lift on Fable 5 and Mythos 5) while leaving teams hanging on the original outage.

The pattern here: Anthropic is behaving like a company that prioritizes optics and control over the developer relationships that underpin its business. The steganography story, true or not, signals that Anthropic is treating API customers as potential adversaries rather than partners. That's a dangerous posture when your revenue depends on developers trusting you enough to build on your platform.