← sources

Are We Ready For The Three Phase Apology Tour?

source: reddit (reddit:ClaudeCode)published 2026-09-28 17:39ingested 2026-09-29 02:42

retained text

I've been using the exact same prompt, for an identical workflow, for quite a while now.

This morning suddenly Sonnet 5 started parsing it all improperly, not following instructions, and throwing in emoticons.

So begins the Anthropic three phase apology tour.

  
Phase 1: Intentionally and silently nerf a model.

Phase 2: Get caught, explain it away as transient failures, but not publicly, never publicly, in fact, what outages? We're not even sure nerf is a word. I think you're making that up. You must be doing it wrong. Our API status page shows we've been running solid for decades! No failure here, and certainly no need for any kind of update or announcement to the plebs.

Phase 3: Double down on it not being your fault, because who could possibly be at fault if nothing happened, am I right? While simultaneously performing a mid-week token reset, purely out of the kindness and generosity of your own incredibly large hearts, that makes everyone forget you fucked them out of tens of thousands of dollars of cumulative usage by nerfing models so you can force them onto models that cost you less money, deliver inferior work, and claim it's an upgrade.

document fetched 2026-09-29 00:41 · canonical https://reddit.com/r/ClaudeCode/comments/1wsl9ol/are_we_ready_for_the_three_phase_apology_tour