Company@X — 2026-05-28#

Signal of the Day#

Anthropic announced a massive $65 billion Series H funding round, driving its post-money valuation to $965 billion, while also reporting an astounding $47 billion in run-rate revenue earlier this month. This capital injection coincides with the release of Claude Opus 4.8, signaling that the enterprise AI market has reached an unprecedented scale and Anthropic is cementing its position as a dominant, highly capitalized frontrunner.

Key Announcements#

Anthropic · Source Alongside its monumental $965 billion valuation and $47 billion revenue milestones, Anthropic shipped Claude Opus 4.8. The new model builds on Opus 4.7 by offering sharper judgment, greater transparency about its progress, and the capability to execute independent workflows for significantly longer durations. Opus 4.8 is already available in Cursor and directly via the Gemini Enterprise Agent Platform.

Google Cloud · Source Google made its top-tier image generation models, Nano Banana 2 and Nano Banana Pro, generally available through Google AI Studio and the Gemini Enterprise Agent Platform API. Nano Banana 2 now features deep video understanding, allowing it to process video files as input prompts to analyze specific subjects, visual context, and audio to generate infographics and thumbnails. Additionally, Google expanded its Universal Commerce Protocol (UCP) to hotel and food ordering verticals.

Glean · Source Enterprise AI platform Glean has surpassed $300 million in ARR, growing roughly 3x over the past 15 months and nearly doubling its Fortune 500 customer count year-over-year. Glean is positioning itself as the critical “context layer” for enterprise AI systems, asserting that deploying models securely connected to internal knowledge and permissions is the true moat in the B2B space.

Hugging Face · Source The Hugging Face science team fundamentally optimized disaggregated Reinforcement Learning (RL) by shipping an async RL weight sync method in TRL that slashes bandwidth usage by nearly 100x. By only transmitting changed elements between RL steps via a sparse safetensors file dropped into an S3-like Hub Bucket, developers can now run multi-replica inference and disaggregated RL without requiring a shared cluster, RDMA, or VPN.

Replit & Visa · Source Replit announced an R&D partnership and strategic investment from Visa to build out infrastructure for agentic payments. This collaboration aims to seamlessly integrate programmable value and payment movements directly into the modern AI stack, enabling autonomous developer agents to handle transactions natively.

Liquid AI · Source Liquid AI launched LFM2.5-8B-A1B, an optimized 8B parameter Mixture-of-Experts (MoE) model designed specifically for edge devices, including phones, laptops, PCs, and robots. Trained on 38 trillion tokens with a 128K expanded context window, the model features fast tool calling and operates highly efficiently on a single GPU.

OpenClaw · Source The OpenClaw agentic framework released version 2026.5.27, introducing tighter security runtime boundaries and achieving up to 2.9x faster agent turns. The update also adds OpenAI-compatible embeddings and deepens provider integrations, making PixVerse a native, credential-aware video generation plugin.

AWS · Source AWS launched the next generation of Amazon OpenSearch Serverless, completely rebuilt for agentic AI workloads with the ability to autoscale 20x faster than previous iterations. Simultaneously, AWS debuted Resilient Network Graphs for data centers, delivering 33% better throughput and a 40% reduction in power consumption by eliminating legacy hierarchical network bottlenecks.

Also Noted#

• Microsoft (Source): Redesigned Copilot for a simpler, faster interface and launched “Command Line,” a new technical blog detailing developer tool architecture ahead of Microsoft Build.
• Waymo (Source): Showcased its sixth-generation Waymo Driver in the new Ojai fleet, with ride expansions rolling out soon in San Francisco, Los Angeles, and Phoenix.
• Cursor (Source): Released the Cursor Developer Habits Report, highlighting a growing power user gap and revealing that input tokens have become the majority of price-equivalent AI costs as agents utilize deeper context.
• Bagel/Paris 2.0 (Source): Released Paris 2.0 weights to Hugging Face, introducing what it claims is the world’s first decentralized trained video generation model with benchmark-topping FVD scores.
• x.ai (Source): Rolled out Grok Build 0.2.7 featuring shared terminals across subagents and improved image understanding parameters.