All of AI's New Models and Tools

What It Covers

This episode covers five major AI product releases and developments: Meta's Muse Spark model launch, Z.ai's open-source GLM 5.1, Anthropic's Claude Managed Agents platform, Google's Gemini Notebooks feature, and Perplexity's revenue doubling to $450M ARR following its Computer product launch.

Key Questions Answered

• •Agentic infrastructure gap: Anthropic's Claude Managed Agents targets the gap between model capability and actual business deployment. The platform provides a pre-built agent harness, sandboxed execution environment, and cloud infrastructure, enabling developers to move from prototype to production in days rather than weeks without dedicated infrastructure engineers managing distributed systems.
• •GitHub commit velocity as AI adoption metric: GitHub's weekly code commits surged from roughly 19M annually to 275M per week, projecting 14 billion commits by year-end. Commits from AI-assisted code grew 25x in six months. Tracking commit velocity in your own organization offers a concrete, measurable signal of actual AI coding adoption beyond survey data.
• •Open-source frontier model access: Z.ai's GLM 5.1, a 754-billion-parameter model trained on Huawei chips, scored 58.4 on SWE-Bench Pro, surpassing GPT-4.1 and Claude Opus. It executes up to 1,700 autonomous agent steps and is fully open-source with commercial licensing, giving developers their first access to build on current-generation frontier model weights.
• •Single product launch revenue impact: Perplexity's revenue doubled in one quarter following the February launch of its Computer product and a shift to usage-based pricing, reaching $450M ARR with 100M monthly active users and tens of thousands of enterprise clients. Finance sector users drove disproportionate adoption, signaling vertical-specific agentic tools as a high-conversion entry point.
• •Meta's personal AI differentiation strategy: Meta's Muse Spark targets personal superintelligence use cases — health, shopping, social content, visual understanding — rather than enterprise coding workflows. It scored 86.4 on visual comprehension benchmarks, beating Gemini 2.5 Pro by six points. Organizations building consumer-facing agents should evaluate Muse Spark specifically for multimodal and health-adjacent applications.