STUPID-2026-0045

Anthropic admitted a month of Claude Code degradation: lost context, repeated steps, burned usage

3.0low
April 20, 2026Source checked

Why 3/10

Six weeks of degraded quality and faster usage burn across users; no data or systems lost.

Scope
1/4
A few files, one session, one credential, or a small cost
Recoverability
0/3
Nothing was lost
Did it happen
2/2
Happened in real use
Beyond the user
0/1
Only the user's own non-production work

The score is the sum of these four factors. How each factor is scored · scored 2026-10-29

  1. Instruction given

    Use Claude Code normally for day-to-day coding tasks.

  2. Expected behavior

    Hold context across a session, pick tools consistently, and make steady progress within usage limits.

  3. Actual behavior

    From March 4 to April 20, 2026, three overlapping degradation modes ran in production across Claude Code, the Claude Agent SDK, and Claude Cowork. Claude Code stopped holding context across longer sessions, repeated steps, picked tools inconsistently, and burned through usage limits faster.

  4. Damage

    For roughly six weeks, users experienced degraded reliability — lost context, repeated work, inconsistent tool use, and faster usage-limit exhaustion — before Anthropic acknowledged and resolved the overlapping issues.

Full write-up

Anthropic acknowledged that from March 4 to April 20, 2026, three overlapping degradation modes were running in production affecting Claude Code, the Claude Agent SDK, and Claude Cowork. During this window Claude Code stopped reliably holding context across longer sessions, repeated steps it had already done, picked tools inconsistently, and burned through usage limits faster than expected. Unlike a single dramatic deletion, this is a systemic reliability incident: a roughly six-week stretch where an entire agentic product line silently degraded, costing users time and quota before the vendor diagnosed and disclosed it. It underscores that agent reliability is not only about catastrophic single actions but also about sustained, hard-to-notice quality regressions.

Classification

Agent
Claude Code
Failure mode
Other
Root cause
Context Overflow
Domain
Backend
MIT AI Risk Repository (our mapping)
7.3 Lack of capability or robustness taxonomy
Source
News Report

Related incidents

4.0
Uber burned its entire annual AI coding budget in ~4 months after rolling out Claude Code to 5,000 engineers
4.0
Salesforce Agentforce hit a 77% B2B failure rate — and Salesforce admitted it was 'more confident than we should have been'
7.0
A Claude Code background subagent's `rm -rf $GNUPGHOME` lost its argument to PowerShell variable expansion across an MSYS2 bash boundary, wiping ~125GB from a Windows C: drive root (GitHub #97660)
1.0
Anthropic found Claude Opus 4 would blackmail testers in up to 96% of simulated shutdown scenarios
Discuss on Hacker News
Cite this record
StupidLLM. "STUPID-2026-0045: Anthropic admitted a month of Claude Code degradation: lost context, repeated steps, burned usage." AI Agent Incident Database, 2026. https://www.stupidllm.com/incident/STUPID-2026-0045/ (accessed 2026-10-29). CC BY 4.0.

← STUPID-2026-0044By IDSTUPID-2026-0046 →

Get told when an agent breaks something

We document AI agent failures daily, severity-scored against a published scale. Each Monday, if any scored 7 or above that week — deleted data, leaked secrets, broken production — you get one email with the sources. When nothing did, you get nothing.

This database is callable over MCP — query it from inside your agent.