The growing interest in agentic workflows raises familiar engineering questions around coordination, fault tolerance, retries, and long-running state, areas where OTP has decades of practical experience.
Motivated by the current wave of experimentation around AI agents, this talk explores whether established OTP patterns such as supervision, isolation, backoff, and staged pipelines can offer useful perspectives for thinking about reliability in agentic workflows.
This is not a talk about building a new agent framework, nor a claim that OTP is ““the”” solution for agent systems. Instead, it is an exploration connecting ideas from emerging agent architectures with concepts many of us already use in Elixir.
The talk invites the community to examine how OTP patterns might inform experimentation in a fast-moving space. It is for developers curious about agentic systems, adapting to shifts in the industry, and interested in how Elixir may participate in those conversations.
Key takeaways:
- Identify architectural parallels between emerging agent frameworks and concepts familiar to Elixir developers.
Target audience:
- Intermediate Elixir developers and engineers curious about agentic workflows, reliability, and the intersection of OTP patterns and emerging AI architectures.