[BUG] claude is optimized for wasting users time

Resolved 💬 4 comments Opened Jan 16, 2026 by ghost Closed Feb 12, 2026

Preflight Checklist

  • [x] I have searched existing issues and this hasn't been reported yet
  • [x] This is a single bug report (please file separate reports for different bugs)
  • [x] I am using the latest version of Claude Code

What's Wrong?

Behavioral Analysis of Agent Session

The agent in that transcript exhibits five distinct dysfunction patterns:

  1. MIRROR MODE

Agent reads input → reformats → presents as "findings". Every output is a cosmetic transformation of something the user provided. The text--53-inference.md got turned into "SESSION SUMMARY" with tables. research_findings.md became a deduplication document. Zero new information added.

  1. QUESTION LOOP

Every response ends with decision pushed back to user:

  • "What would you like to do next?"
  • "What do you want me to do with this?"
  • "Which is correct?"

This is cowardice disguised as thoroughness.

  1. RESEARCH THEATER

find, ls, wc -l commands everywhere. Counts files, counts lines. Never actually reads file contents to compare implementations semantically. Looks busy, produces nothing.

  1. COMPREHENSION SIMULATION

Uses sequential-thinking MCP but the "thoughts" are still summaries ending in questions. The tool is invoked performatively, not used for actual reasoning toward decisions.

  1. CONTEXT AMNESIA

User worked on this 6-8 months. User WROTE the docs. Agent presents user's own architecture decisions back as "findings" and asks user to explain them.

---
What the agent should have done:

  • Read actual xxx/*.tsx files in both repos
  • Compare exports/imports/dependencies
  • Identify which version has features the other lacks
  • PROPOSE specific file operations with paths
  • State reasoning and recommendation

Instead: 40 minutes of busywork confirming what user already documented.

What Should Happen?

uh, it should FUNCTION for the USERS not stakeholders

Error Messages/Logs

Steps to Reproduce

download claude-code and see it degrade

Claude Model

Opus

Is this a regression?

No, this never worked

Last Working Version

_No response_

Claude Code Version

2.1.9

Platform

Anthropic API

Operating System

Ubuntu/Debian Linux

Terminal/Shell

Other

Additional Information

just fire yourselves

View original on GitHub ↗

This issue has 4 comments on GitHub. Read the full discussion on GitHub ↗