Claude Code user bans subagents after repeated 200K-token failures, then asks if the tradeoff is worth it
MysteriousFloor1406 · reddit · 2026-07-24
A Claude Code user asks whether subagents are worth the token cost after seeing them repeatedly fail on a large project.
- The user noticed Opus routing work to Sonnet as a subagent, which burned through roughly 200K tokens per attempt.
- Because the subagent kept failing due to context limits, the user banned subagents in claude.md and still finished the project successfully.
- The question now is whether the ban should be removed, and what practical benefits subagents actually provide.
- The post is essentially a real-world cost/benefit question about agent decomposition, context management, and workflow efficiency.
More from coding & agent
- OpenAI ships GPT-6.1 Terra; Brockman says its computer use now beats his own speed — gdb · 2026-10-27
- Kolab Benchmark Uses Pro Screen Recordings to Show AI Agents Overestimate Collaboration Quality — minnesotanlp · 2026-10-27
- Dev argues Claude burns far more tokens than GPT models for the same coding work — haider1 · 2026-10-27
- Stop building agents that can't change models: stable contracts, external state, real evals — rseroter · 2026-10-27
- Cheap models can cost more in agent harnesses than frontier ones, researcher warns — vasuman · 2026-10-27
- Kanzo: A 5k-Line Rust Kanban Binary With MCP for Claude Code and Codex — allmudi · 2026-10-27