Company@X — 2026-06-12#

Signal of the Day#

SpaceX officially began trading publicly on the Nasdaq today under the ticker $SPCX. This highly anticipated public market debut marks a historic milestone for the aerospace industry, prompting widespread congratulations across the tech ecosystem from key investors like Andreessen Horowitz and Sequoia Capital.

Key Announcements#

[MiniMax] · Source MiniMax launched M3, an open-weight model engineered for agentic workflows and coding that features ~428B total parameters with ~23B activated during inference. The release natively scales context up to 1 million tokens using a custom Sparse Attention kernel, which the company has also published on Hugging Face to deliver massive speedups for long-sequence tasks.

[OpenAI] · Source OpenAI is rolling out the ability for Codex users to save and bank their rate limit resets for later use, giving Go, Plus, Pro, and Business tier users one free banked reset to start. To drive adoption, they are pairing this flexibility with a referral program that grants Plus and Pro users additional banked resets when invited friends send their first Codex message.

[Google Labs] · Source Google expanded global access to its interactive world-building AI, Project Genie, making it available to subscribers of a new “Google AI Ultra 5X” tier. This indicates an aggressive strategy to lock high-compute generative capabilities behind premium, top-tier subscription walls.

[OpenClaw] · Source The open-source agent framework OpenClaw shipped version 2026.6.6, adding critical integration for Claude Fable 5’s adaptive thinking and Gemma 4 reasoning replays. The update focuses heavily on security infrastructure for autonomous agents, implementing stricter sandbox bindings, host environment inheritance, and tighter execution approvals.

[Google DeepMind] · Source DeepMind launched a European Robotics Accelerator, selecting 15 startups focused on the future of physical AI. The three-month program provides these early-stage companies with direct access to Google’s specialized Gemini Robotics models and hands-on infrastructure support.

[Browser Use] · Source Browser Use unveiled version 0.13.0 beta, featuring a complete Rust rewrite designed to move away from legacy GPT-4 architecture toward state-of-the-art models. The framework introduces custom LLM and browser harnesses specifically optimized for deploying and managing thousands of long-running, autonomous web agents.

Also Noted#

• [Claude] (Source): Claude Managed Agents are now available on platforms like Blaxel and Google Cloud, allowing developers to execute agent loops on managed infrastructure with secure, persistent sandboxing.
• [Google Gemma] (Source): Google demonstrated real-time social robotics by running its Gemma 4 model locally on a Reachy Mini robot alongside cloud-based Gemini Live interactions.
• [Extend AI] (Source): Extend open-sourced its CLI, which includes a built-in agent skill that allows tools like Claude Code, Codex, and Cursor to natively parse and extract entire document corpuses directly from the terminal.
• [Starcloud] (Source): Showcasing the durability of space-based compute infrastructure, Starcloud-1 successfully ran DOOM on an H100 GPU in orbit.
• [Andreessen Horowitz] (Source): A new a16z market report flags AI summaries as the primary driver behind both the decline of traditional search as a traffic source and the recent spike in zero-click queries.