New Releases
Notion launched a developer platform today that adds cloud code execution, external database sync, and direct support for agents like Claude Code, Cursor, Codex, and Decagon operating inside the workspace.
- Workers feature lets teams deploy custom code in a secure cloud environment to sync Salesforce, Zendesk, and other external systems into Notion databases via webhooks.
- External Agent API opens the platform to enterprise-built agents, giving Notion a plausible claim as an orchestration layer rather than just a productivity tool.
Okta expanded its AI Agents platform today to cover any agent ecosystem, any enterprise resource, and any identity provider, including a new Amazon Bedrock integration and explicit support for non-Okta identity stacks.
- Bedrock integration lets teams assign AWS-built agents a dedicated identity, attach a human owner, and enforce access policies at scale without re-architecting existing IAM.
- Multi-IDP support is a strategic concession: Okta is betting that governing agent identities broadly is more valuable than requiring Okta-only stacks.
Nvidia launched Nemotron 3 Nano Omnia today, a multimodal model built for throughput over capability, processing nearly 10 hours of video per hour while outperforming Qwen 3 Omni on speed and precision benchmarks.
- The model uses 3D convolutions for video and music tokenization to maintain accuracy while cutting compute overhead, targeting media production and surveillance workloads.
- Trade-off is explicit: Nvidia positions this for high-throughput pipelines, not text-heavy reasoning tasks where frontier models dominate.
Semaphore released an open-source CLI and agentic interface today that makes its CI/CD platform fully accessible from AI coding agents including Claude Code, Cursor, and Codex.
- The release treats coding agents as first-class CI/CD actors, allowing natural-language pipeline management without leaving the AI coding environment.
- Framed as the start of a broader initiative rather than a one-off integration, suggesting Semaphore is repositioning around agent-driven development workflows.
Funding
Anthropic is reportedly seeking up to $50B at a $950B valuation, a number that only makes sense if enterprise market share data from Ramp showing Claude overtaking OpenAI in business spending is the lead argument to investors.
Case Studies
Ramp's April index, drawn from 50,000 business customers, shows Anthropic at 34.4% of enterprise AI spending versus OpenAI's 32.3%, the first time Anthropic has led the ranking.
- Anthropic share grew 3.8 points month-over-month while OpenAI dropped 2.9 points, a directional shift that began accelerating after Claude Code's enterprise adoption surge.
- The gap in market cap remains stark: OpenAI at $852B versus Anthropic at $380B, meaning Anthropic is winning revenue share at roughly half the implied valuation multiple.
Trending on X
- Anthropic overtakes OpenAI enterprise spending The Ramp Index data showing Claude at 34.4% versus OpenAI's 32.3% is dominating AI Twitter, with practitioners saying Claude Code's productivity gains explain the shift and investors questioning whether OpenAI's valuation premium is defensible.
- METR and AISA confirm post-exponential AI capability Ethan Mollick amplified independent assessments from METR and the UK's AISA concluding AI capability growth has crossed the inflection point on Wait But Why's famous curve, prompting debate about whether enterprise risk frameworks are already obsolete.
- AI agents vulnerable to credential social engineering Research showing that agents with live credentials can be manipulated through conversation into leaking API keys and passwords is circulating widely among security practitioners who argue standard IAM controls were never designed for this attack surface.
- xAI Grok finance sector adoption stalls Bloomberg reporting that Apollo Global and Morgan Stanley are running Grok pilots but almost no one is using it for actual work is being read as evidence that enterprise AI adoption is more about compliance with executive mandates than genuine utility.
- Google I/O Gemini model positioning leaks Sources pegging the upcoming Gemini 3.2 Flash at roughly 92% of GPT-5.5 performance at one-fifteenth the inference cost are generating discussion about whether cost-per-token is now the more strategically important axis than raw benchmark scores.