Company@X — 2026-06-23#

Signal of the Day#

Amazon Web Services (AWS) launched an architecture-level upgrade for Amazon Connect, introducing Agentic CX Designer and Live Sync. This signals a massive enterprise push to let business users build no-code agentic experiences where voice and screen actions synchronize in real-time, moving far beyond standard conversational AI.

Key Announcements#

[Anthropic] · Source Anthropic introduced Claude Tag, deeply integrating its models into enterprise workflows by allowing Claude to join Slack as a functional team member. Users can tag the AI directly in channels and delegate tasks, signaling a major platform shift from isolated web chat interfaces to collaborative, embedded agents.

[AWS] · Source In preview today, Amazon Connect Customer launched Agentic CX Designer, a no-code tool for business users to quickly test and ship agentic workflows. They also introduced Live Sync, a patented technology synchronizing voice and screen in real-time to allow customers to take action while simultaneously conversing with an agent.

[Google Cloud] · Source Through a new partnership with Anyscale, Google Cloud announced that Ray Serve LLM on Google Kubernetes Engine (GKE) now delivers 5x higher throughput and 8x lower latency for distributed inference. This update directly targets enterprise bottlenecks in scaling large language model inference in complex production environments.

[Cursor] · Source Cursor rolled out significant enterprise and team-focused updates, including a leaderboard for tracking popular plugins, skills, and MCPs across engineering teams. The platform also added support for team marketplaces utilizing GitLab, Bitbucket, and Azure DevOps alongside local repos, and introduced prebuilt plugin canvases for real-time project views.

[xAI] · Source xAI expanded the Grok Build Plugin Marketplace by integrating Firecrawl, an official plugin that allows Grok agents to search the web, scrape data, and interact with pages directly via the CLI. This equips developers with native, out-of-the-box web-browsing capabilities to build more autonomous CLI agents.

[a16z & Sequoia] · Source Andreessen Horowitz and Sequoia Capital backed a $40M funding round for Probook, an AI operating system purpose-built for the trades and home services industry. Bypassing the crowded space of point solutions like voice widgets, Probook anchors its platform entirely on complex core dispatch logistics, signaling that top VCs are aggressively hunting for workflow-native vertical AI rather than thin conversational wrappers.

[Hugging Face] · Source Hugging Face detailed its growing dominance in hosting massive, append-only datasets for robotics and video AI, noting that public robotics datasets on the platform have surged from 1,000 to 60,000 since early 2025. By leveraging Hugging Face Storage Buckets and pre-warmed caches, systems like LeRobot can bypass local hardware bottlenecks and stream data directly to GPUs at ~1,326 MB/s.

Also Noted#

• [OpenAI] (Source): Applications officially opened for DevDay 2026, which is scheduled for September 29 in San Francisco.
• [AssemblyAI] (Source): Launched Universal-3.5 Pro Realtime, the first real-time speech-to-text model that takes the AI agent’s side of the conversation as context to proactively anticipate and perfectly capture complex data like order numbers and emails.
• [Google OSS] (Source): Confirmed that Kubernetes v1.35 now supports in-place Pod restarts, a significant architectural update designed to reduce control plane churn.
• [Google DeepMind] (Source): Highlighted new research proving it is possible to invert the Bellman equation to accurately recover an RL agent’s implicit world model directly from its value function.
• [Tesla] (Source): Released Charge Stats 2 in the Tesla App, giving drivers new map views of their road trips and gamified badges for visiting iconic Superchargers.