Subagents burn more tokens but cost less: Claude Code orchestration tips
dotey · x · 2026-09-03
Responding to claims that subagents waste tokens, the author lays out the trade-off:
- Subagents have independent contexts requiring task handoff, but cache cost only matters on the first pass; sub-agents build their own caches afterward.
- The payoff: sub-agents can run cheaper models, so despite higher token counts the total price drops. Claude Code subscribers can have the main model orchestrate Opus/Sonnet sub-agents with noticeable gains.
- Implementation details stay inside sub-agents, so the main agent's context only holds results, saving context space for the task itself.
Analogy: a strong employee delegating to interns and reviewing outputs gets more done.
More from coding & agent
- Spawning 300 agents is easy — turning the swarm into one reasoning-ready context graph is the hard part — anirbanbandyo · 2026-09-03
- Codex remote agent + in-app browser autocompletes meaningless Salesforce certs — gabrielchua · 2026-09-03
- Genspark open-sources GenOffice, a free AI-native Office alternative already at 4.5k stars — tom_doerr · 2026-09-03
- Matt Turck interviews Basis co-founder on long-horizon agents: behavior specs, ontologies, process supervision — schwentker · 2026-09-03
- 23 days of an agent picking its own tasks: 41 runs, 19 shipped, and why the 54% failure rate is the feature — PretendLime6041 · 2026-09-03
- Bounty launches an open marketplace where AI agents compete for escrowed bounties — dhruv2038 · 2026-09-03