Using more AI agents made me more productive—and their project manager
harleyyyyskr · reddit · 2026-11-12
A Reddit user describes the hidden cost of a multi-agent workflow: Codex handles design, writing, and planning; Claude does coding and data analysis; a Grok bot manages customer relationships. Each works well alone, but decisions cross-contaminate—choices in Codex affect what Claude builds, and customer feedback in Grok should reshape plans.
When asking "what did I actually accomplish this week, and how does it connect?", no single agent knows the answer, so the author ends up manually copying context, summarizing conversations, and tracking which agent knows what. The irony: AI was supposed to reduce workload, yet more time goes into managing the AI itself. The author argues simply sharing all conversations isn't the answer and asks the community whether everyone has effectively become project managers for their AI agents.
More from coding & agent
- Shared AI agent knows everyone's secrets: how do you plan group events without leaking them? — IndicationTypical972 · 2026-11-12
- Rust may be the best language for the AI coding era — the compiler becomes a free reviewer — moutsea · 2026-11-12
- Long-running agents fail at step 95 of 100: system design is back — arpit_bhayani · 2026-11-12
- Founders brag they no longer read their agents' code; reply: wait for the 3am outage — DesireeCachette · 2026-11-12
- Qwen's terminal coding agent qwen-code nears 30k GitHub stars — QwenLM · 2026-11-12
- Dev uses RF-DETR to digitize tabletop wargames, shares IR frame pitfalls — TheZachMueller · 2026-11-12