Descriptions:
Anthropic released Claude Opus 5 alongside a new real-time voice mode for the Claude iOS app, and Riley Brown covers both launches with hands-on testing in this episode of Agent Native. Opus 5 is positioned as a near-frontier everyday model for knowledge workers, enterprises, and developers at roughly half the cost of GPT-4.5, with Anthropic’s benchmarks showing it leads in agentic coding, knowledge work, novel problem solving, and computer use. Early adopters including Dan Shipper at Every report a difficult migration experience — the model argues with instructions and requires rebuilding existing skill configurations from scratch — though results improve significantly once teams adapt their prompting approach.
The more significant announcement, in Brown’s assessment, is the Claude voice mode. The new iOS voice agent supports real-time conversation and connects to MCP servers, enabling it to control third-party applications during live voice sessions. Brown demonstrates this by directing the voice agent to read a Paper whiteboard, design a landing page, spawn parallel high-reasoning tasks in separate chat windows, and iterate on visual design directions — all through natural voice commands with no keyboard interaction.
Within ten minutes of Claude’s voice launch, OpenAI released voice control for Codex, allowing developers to direct their coding agent verbally in real time. The episode closes with Brown’s analysis of how he plans to integrate voice-native AI agents into his startup’s content and development workflows, framing the dual launch as a turning point toward genuinely voice-first agentic computing.
📺 Source: Riley Brown · Published July 25, 2026
🏷️ Format: Review
