SOCIAL MEDIA
Company@X
Sources AI at Meta / @AIatMeta Amazon Web Services / @awscloud Anthropic / @AnthropicAI Cursor / @cursor_ai Google / @Google Google Cloud Tech / @GoogleCloudTech Google …
Sources
- AI at Meta / @AIatMeta
- Amazon Web Services / @awscloud
- Anthropic / @AnthropicAI
- Cursor / @cursor_ai
- Google / @Google
- Google Cloud Tech / @GoogleCloudTech
- Google DeepMind / @GoogleDeepMind
- Grok / @grok
- Hugging Face / @huggingface
- Microsoft / @Microsoft
- OpenAI / @OpenAI
- OpenClaw🦞 / @openclaw
- Sequoia Capital / @sequoia
- Tesla / @Tesla
- Twitter / @a16z
- Waymo / @Waymo
- xAI / @xai
- Y Combinator / @ycombinator
Signal of the Day
Meta has released Muse Spark 1.2, a major multimodal upgrade that translates visual observations directly into executable code and physical robotics coordination, alongside WildArtifactBench to establish a preference-based, real-world utility standard for AI agents.
Key Announcements
Meta · Source Meta previewed Muse Spark 1.2, introducing advanced audio-visual understanding for video-heavy enterprise workflows and spatial intelligence to orchestrate physical, bimanual robotics tasks like desk tidying. Alongside the model, Meta debuted WildArtifactBench, an evaluation framework of 10 open tasks that assesses agents on complex deliverables using win rates and Elo scores from human and agentic judges. This dual release signals a strategic push toward deploying and measuring long-horizon agents capable of executing multi-turn real-world workflows.
Apollo Atomics · Source Apollo Atomics secured $31M in seed financing—including $26M in equity led by FCVC and $5M in equipment financing from Oak Ridge Financial—to build high-power output nuclear reactors (>80 MW) that fit on a truck. Currently running a low-power reactor demonstrator at MIT, the startup is developing a full-scale 1 MW prototype and collaborating with the U.S. NRC for commercial deployment in 2028. This design prioritizes mass manufacturing and extreme cost-effectiveness to solve the energy grid’s power density challenges.
Waymo · Source Waymo disclosed its engineering architecture for autonomous driving compute, detailing key partnerships with semiconductor giants AMD, NVIDIA, TSMC, Samsung, and Micron. Simultaneously, the Alphabet-backed company officially opened its ride-hailing service to everyone in Space City (Houston), marking another massive metropolitan expansion. The announcements underscore Waymo’s dual-track strategy of solidifying its physical supply chain while aggressively scaling its commercial market footprint.
Y Combinator · Source Y Combinator launched “Send a SAFE,” a free, automated tool enabling founders to generate, sign, and distribute SAFE agreements in under two minutes. Crucially, the platform is optimized to run with AI agents, allowing automated workflows to handle early-stage transaction mechanics. By updating the fundraising standard that has facilitated over $15B in funding since 2013, YC is actively building the legal rail system for agentic commerce.
Google · Source Google’s Gemma family of open models has officially surpassed one billion downloads, with the developer ecosystem producing over 100,000 fine-tuned variants. To capitalize on this, Google launched the “Awesome Gemma” GitHub repository and highlighted key domain-specific breakthroughs, including C2S-Scale (a 27B parameter model that discovered a novel cancer therapy pathway) and DolphinGemma for interspecies vocalization research. The massive distribution footprint solidifies Google’s strategy of leveraging open-weights ecosystems to maintain developer mindshare.
Pocket · Source Hardware startup Pocket reached $100M in Annualized Run Rate (ARR) just five months after graduating from Y Combinator’s Winter 2026 batch. Scaling from $1M to $100M ARR in just seven months with over 270,000 active devices deployed globally, the company marks the fastest growth trajectory of any hardware company in YC history. The milestone signals a substantial acceleration in the commercial adoption rate and venture-scale viability of consumer hardware.
Also Noted
- Idler AI (Source): Introduced a frontier data research lab with a $9M seed round led by Paradigm and released “ShelfLife,” a digital twin dataset of a profitable e-commerce business to train and benchmark models under real commercial conditions.
- Thunder Compute (Source): Raised a $13M Series A to build GPU virtualization technology designed to unlock the estimated 80% of global GPU capacity that sits idle.
- Hugging Face (Source): Released HiPHI, a public dataset featuring 617.5 hours of high-precision human motion and object interaction to train physical AI policies on real humanoid hardware.
- Grok (Source): Rolled out single-prompt app creation, hosting, and custom domain publishing for all SuperGrok and X Premium plans, alongside robust team sharing permissions.
- PostHog (Source): Disclosed a strategic transition into a “doing company,” using AI agents that can autonomously scan codebases, resolve issues, and open pull requests.
💡 What to do next: I can package these daily updates and underlying strategic trends into a polished, executive-ready slide deck outlining today’s key advancements in AI engineering and physical hardware. Let me know if you would like me to draft that presentation for you!