Mozaik

Right — the coupling to long-running calls is the subtler killer, because it stays invisible until load: one agent parked on a slow tool/LLM call holds the lane and everything behind it stalls. And the two problems compound — once you decouple so agents don't block, the shared-state races get *more* likely, not less, because now they're genuinely concurrent. Curious whether you lean on message-passing to sidestep shared mutable state, or fall back to real locks/transactions when agents do need the same resource.

We still need to deal with race conditions. But in most multi-agent frameworks and SDKs, there is an even worse problem: agents are coupled with long-running operations like LLM calls and tool calls, so they end up blocking each other. One agent needs to finish before another can start working. We solve this by completely moving long-running operations out of the agents and enabling agents to react to the events those operations produce. That way, we create an environment where agents do not block each other.

The "don't wait on each other" framing is the right one — blocking is where most multi-agent setups quietly fall over. The part I always end up fighting is shared state: the moment agents stop waiting, they race on whatever they're both reading/writing, and you're back to needing locks or a single writer. How are you handling that — message-passing, per-agent state ownership, or something cleverer? Following along.

The non-blocking agentic environment is a sharp framing — 'agents waiting on each other' is exactly what tanks throughput once you chain more than two. How are you handling ordering/consistency when observers and agents act on the same state concurrently? Bookmarking this.

Mozaik takes a different approach. It lets developers build autonomous teams of AI agents that are aware of each other, communicate through tools, and collaborate in parallel without hardcoding every workflow path upfront. Instead of forcing developers to predict how agents should coordinate, Mozaik enables coordination to emerge while agents are working. With Mozaik, you can build agents that: - work in parallel - react to each other’s progress - communicate through shared tools and channels - coordinate dynamically at runtime - form agentic teams instead of isolated single agents We are building Mozaik because we believe the next step in AI software is not just better individual agents - it is autonomous teams of agents that can collaborate like real software teams.