One AI agent may beat a swarm of agents
Using many together can be harder to manage than it first seems. The difficult part is not creating the agents, but keeping their work in sync. It is usually fine for several agents to read the same document, but problems start when they write to the same place.
In one real case, one agent wrote a summary while another added to the same notes file. The later write replaced the earlier work, so the disappeared without an error. looked normal, and the problem was only found days later when a did not happen.
Once agents share and change the same state, the setup starts to behave like a small . That becomes even harder when one part of the system is an LLM.
Key points
- One capable agent in a loop may be enough for many tasks.
- Many agents reading the same material is usually safer than many agents writing to the same state.
- Shared notes or files can be silently overwritten when agents write close together.
- A failure may not show an error, even when important information is lost.
- A can become a that needs careful coordination.