Meta Muse's agent control plane times out under 120 simultaneous subagent spawns
I stress-tested Meta Muse until its agent control plane started timing out
I stress-tested Meta Muse, Meta's personal AI agent, by spawning up to 120 subagents at once. At 120 attempts, 87 failed with database lock timeouts, while 33 completed. A staggered run with a coordinator agent succeeded all 80 times, but it changed both cadence and topology, so it's not a clean control. The investigation reconstructs the gap between parent and child agent states from PostgreSQL traces, revealing a sharp failure cliff between 80 and 120 concurrent spawns.
The runtime was not asked for 33 workers; it was asked for 120.
- bradfa
This writing style, which reads like an LLM wrote it, is draining to read.
I put up with it when I'm actively using an LLM tool, because I just accept that this is how these tools write, but when reading a blog I expect to read words written in one of the vast varieties of normal human-written blog formats.
- nzoschke
I dont fully understand how these black box agent computers will actually work.
One of the huge value adds of the agent computer paradigm is your agent can also fix and customize its own software.
I’m working on a similar agent computer and having the full source code plus a plugin system is the thing that really made it start working.
Something tells me Meta and Grok bots won’t give you this level of control.
More thoughts at this blog: