OpenAI is shutting down Sora and folding its science team, with CPO Kevin Weil and Sora lead Bill Peebles departing as the company doubles down on enterprise AI. This represents a deliberate strategic contraction away from consumer moonshots and generative media. The talent exodus and product kills signal OpenAI is ruthlessly prioritizing revenue-generating enterprise bets.
Cursor is in talks to raise $2B+ at a $50B valuation, led by returning investors a16z and Thrive, driven by surging enterprise adoption of AI coding tools. A $50B valuation for a coding assistant validates that AI-native developer tooling is a durable, large market — not a feature. This is one of the fastest value creation trajectories in enterprise software history.
Salesforce is launching 'Headless 360,' opening its entire platform to AI agents via APIs and treating the browser-based UI as obsolete infrastructure. Benioff is explicitly betting that agentic AI will route around human-facing interfaces entirely, with APIs becoming the primary interaction layer. This is a strategic repositioning of Salesforce from a SaaS UI company to an agent-accessible data and workflow platform.
OpenAI is expanding its Trusted Access for Cyber program and introducing GPT-5.4-Cyber, a specialized model for vetted cybersecurity defenders. This is a deliberate move to capture the cybersecurity vertical with a purpose-built model, following the same domain-specific playbook as GPT-Rosalind in life sciences. The 'trusted access' framing signals OpenAI is building compliance and procurement muscle for regulated, high-stakes verticals.
Google is launching 'Skills' in Chrome, allowing users to save AI prompts as one-click toolbar tools that execute instantly on any web page. With 307 HN score, this has strong resonance — it's essentially user-programmable AI automation baked into the world's dominant browser. This directly competes with Chrome extension ecosystems, prompt management tools, and lightweight no-code automation products.
OpenAI's updated Codex app for macOS and Windows now includes computer use, in-app browsing, image generation, memory, and plugin support — effectively making it a general-purpose AI agent for developer workflows. With a 1530 HN score, this is the highest-signal story of the batch, indicating strong builder resonance. Codex is no longer a code completion tool; it's positioning as a full autonomous developer agent platform.
OpenAI's updated Agents SDK adds native sandbox execution and a model-native harness, enabling developers to build secure, long-running agents that operate across files and tools without custom scaffolding. Sandbox execution is the critical addition — it closes a major security gap that previously made production agent deployment risky. This moves agent infrastructure from DIY to opinionated platform, which will accelerate adoption but also increase OpenAI platform dependency.
OpenAI has launched GPT-Rosalind, a frontier reasoning model purpose-built for drug discovery, genomics, protein reasoning, and scientific research workflows. This is OpenAI's first publicly named domain-specific model, signaling a shift from general-purpose models to vertically tuned reasoning systems. Named after Rosalind Franklin, the model appears targeted at displacing specialized bio-AI tools like AlphaFold adjacents and incumbent pharma AI platforms.
Google has released Gemini 3.1 Flash TTS, a next-generation expressive text-to-speech model in the Gemini family. Flash-tier positioning suggests this is optimized for cost and latency rather than maximum quality, targeting high-volume production use cases. This intensifies competition with ElevenLabs, Cartesia, and OpenAI's TTS offerings in the developer API market.
That's today's briefing.
Get it in your inbox every morning — free.
Help us improve AI in News
Got a suggestion, bug report, or question?