Ask HN: Why Python for Agents?
For most LLM “agent” frameworks, the core work seems to be API orchestration, state management, retries, tool calling, and I/O—not numerical computing or data science. Given that, I’m struggling to understand why Python is the default choice over languages like Java, C#, or Go, which generally offer:
-stronger typing and compile-time guarantees -better concurrency and async models -higher throughput and lower latency for I/O-heavy services -more mature tooling for large, long-running backend systems
If the heavy ML work is happening remotely (via OpenAI, Anthropic, etc.), what advantages does Python provide here beyond ecosystem momentum? Is the choice mostly historical (ML roots), or are there concrete technical reasons Python is better suited for agent orchestration?