RubyLLM 1.16 shipped today. The useful bit is concurrent tool execution: when a model asks for multiple tool calls, RubyLLM can run them in threads or fibers instead of serially. Also adds Rails-style instrumentation and api_base support across native providers.