A builder-heavy week, plus a notable voice-AI distribution win.
Cursor ships Composer 2.5
On May 18, Cursor released Composer 2.5, its in-house coding model tuned for long-running agentic tasks (reported ~79.8% on SWE-Bench Multilingual) at roughly $0.50/$2.50 per million tokens. Owning the model under the agent is becoming a competitive moat for the app-layer leaders.
Anthropic's 'Code with Claude' agent features
At its Code with Claude event (May 19–21), Anthropic showed off new Claude Code agent capabilities — automated Code Review, Remote Agents (drive your laptop from your phone), and cloud Managed Agents for multi-agent orchestration. (The model powering them, Opus 4.8, landed a week later — supporting cast, not the headline.)
ElevenLabs lands inside Spotify
On May 21, Spotify launched an ElevenLabs-powered audiobook creation tool in Spotify for Authors — a major distribution win for a voice-AI company, embedding its tech directly into a consumer platform.
Model watch: Gemini 3.5 Flash
At Google I/O (May 19–20), Google launched Gemini 3.5 Flash and the any-input Gemini Omni video model — fast, cheaper tiers that power a lot of app-layer agents.
What it means if you're picking tools
Coding agents and voice tools both leveled up. Compare coding and voice & audio.
Sources: Cursor Composer 2.5 · Code with Claude (MIT Tech Review) · Spotify × ElevenLabs (TechCrunch)