[MODEL] Unacceptable Weekend Abandonment After Shipping Broken Release - Multiple Critical Failures

Preflight Checklist

• [x] I have searched existing issues for similar behavior reports
• [x] This report does NOT contain sensitive information (API keys, passwords, etc.)

Type of Behavior Issue

Other unexpected behavior

What You Asked Claude to Do

Shipping a broken release on a Friday and then going completely radio silent for the entire weekend is unacceptable for a paid product. Users were left stranded with a fundamentally broken tool while Anthropic apparently took the weekend off.

I wanted to use a stable, functioning product for productive coding assistance. I expected to make API requests without errors. I expected accurate token tracking so I could manage my budget. I expected intelligent, helpful responses that actually solve problems. I expected that if something breaks, there would be support to fix it.

What Claude Actually Did

Claude Code broke with a buggy release pushed right before the weekend. API requests failed constantly with "overload" and "too many users" errors. The service was unusable for extended periods. The token counter showed 0 tokens used while generating extremely long responses – either a bug or deliberately misleading. Response quality dropped catastrophically. Answers were incoherent and unhelpful, a massive regression compared to previous versions. I wasted time fixing Claude's mistakes instead of making progress. And when all of this happened? Nothing. Zero patches. Zero updates. Zero communication. Complete silence for 48+ hours.

Expected Behavior

Releases should be tested properly before deployment. If bugs slip through, there should be on-call support or hotfixes for critical issues – especially on weekends after a new release. The service should handle load properly or at least communicate outages transparently. Paying customers should not get locked out with "too many users" errors. Token metering should be accurate and honest. Response quality should remain consistent across versions. When things break, users deserve communication, not silence.

Files Affected

Permission Mode

Accept Edits was OFF (manual approval required)

Can You Reproduce This?

Sometimes (intermittent)

Steps to Reproduce

_No response_

Claude Model

Opus

Relevant Conversation

Impact

High - Significant unwanted changes

Claude Code Version

2.1.19

Platform

Anthropic API

Additional Context

Paying customers deserve better. No more Friday releases with no weekend coverage. No more silent treatment when things break. No more broken metering that hides actual usage. No more quality regressions that make the tool worse than useless.
This pattern of shipping broken releases and abandoning users is not sustainable. What concrete steps will Anthropic take to prevent this in the future?

Claude Max Account
Model: Default Opus 4.5 · Most capable for complex work
IDE: Connected to VS Code extension version 2.1.19
Linux WSL2 Ubuntu