In partnership with
Hello from The AI Night,
Today in AI:
Google Introduces Gemini Embedding 2 for Unified Multimodal AI
Anthropic’s New Institute Targets AI’s Biggest Hurdles
Google Has Acquired Cloud & AI Security Platform “WIZ”
Why is everyone launching a newsletter?
Because it’s how creators turn attention into an owned audience, and an audience into a real, compounding business.
The smartest creators aren’t chasing followers. They’re building lists. And they’re building them on beehiiv, where growth, monetization, and ownership are built in from day one.
If you’re serious about turning what you know into something you own, there’s no better place to start. Find out why the fastest-growing newsletters choose beehiiv.
And for a limited time, take advantage of 30% off your first 3 months with code GROW30.
Image Source: Google blog
Here's the deal: Google released Gemini Embedding 2 in public preview via the Gemini API and Vertex AI. The model maps text, images, video, audio, and documents into a single unified embedding space, built on the Gemini architecture.
The Breakdown:
Supports text (up to 8,192 tokens), images (up to 6 per request), video (up to 120 seconds), audio (no transcription needed), and PDFs (up to 6 pages).
Handles interleaved multimodal input, meaning you can pass image + text together in one request.
Uses Matryoshka Representation Learning (MRL) for flexible output dimensions scaling down from the default 3,072.
Covers 100+ languages for semantic understanding.
Embedding 2 outperforms leading models across text, image, and video embedding benchmarks.
Gemini Embedding 2 is available now in public preview.
The bigger picture: Builders no longer need separate embedding models per modality or custom glue code to align them. One unified space means a video, a PDF, and a text query can now be compared directly, making multimodal RAG and cross-media search significantly easier to ship and cheaper to maintain at scale.
Image Source: Anthropic blog
Here's the deal: Anthropic announced the Anthropic Institute, a new internal body focused on understanding and shaping the real-world consequences of frontier AI systems. The Institute sits inside Anthropic and works across four defined challenge areas.
The Breakdown:
Covers four pillars; AI's impact on jobs and economies, societal threats and resilience, how AI systems behave once deployed, and governance of autonomous AI research and development.
Combines technical research with policy-facing analysis, examining both upsides and risks.
Explicitly frames AI self-improvement and human oversight as a core governance question.
Operates from inside a frontier lab, giving it direct access to model behavior and deployment data.
The bigger picture: Anthropic is investing in structured, long-term research on AI's societal effects not just model capabilities. For policymakers, researchers, and builders, the Institute could become a key reference point as debates around AI labor displacement, misuse, and autonomous development intensify in the months ahead.
Image Source: Google blog
Here's the deal: Google has officially completed its acquisition of Wiz, the cloud and AI security platform. Wiz joins Google Cloud while retaining its brand and multicloud support across AWS, Azure, and Oracle Cloud.
The Breakdown:
Wiz's platform connects code, cloud, and runtime into a single shared context, enabling teams to prevent risk early and protect applications continuously
Google will combine Wiz's Cloud Security Platform with Google Security Operations, Threat Intelligence, and Mandiant Consulting into one unified offering
Wiz has expanded over the past 12 months with AI security capabilities covering visibility into AI applications, AI-native risk prevention, and runtime protection for AI workloads
The combined platform integrates Gemini for accelerated threat hunting, remediation workflows, and audit documentation
Wiz will continue supporting multicloud environments and third-party partnerships through Google Cloud Marketplace
The bigger picture: Enterprises deploy AI agents fed with business-critical data, the attack surface is shifting from traditional infrastructure to AI models themselves. This acquisition positions Google Cloud as a single-vendor option for securing cloud-native and AI workloads end-to-end across hybrid and multicloud environments.
What else you need to know:
Anthropic is opening a Sydney office, its fourth in Asia-Pacific, to serve growing enterprise demand in Australia and New Zealand while exploring local compute infrastructure expansion.
Thinking Machines Lab partnered with NVIDIA to deploy at least one gigawatt of Vera Rubin systems for frontier model training, with NVIDIA also making a significant investment in the company.
ElevenLabs launched ElevenCreative, a dedicated brand account for sharing product updates, creator tips, and showcases across its voice, music, sound effects, image, and video tools in 70+ languages.
Anthropic added a /btw slash command to Claude Code, letting users start side conversations without interrupting Claude's active task, enabling parallel context switching during coding sessions.
Google's Gemini in Sheets hit 70.48% on the SpreadsheetBench benchmark, claiming state-of-the-art performance that the company says nears human expert ability in autonomous spreadsheet manipulation.
That’s it for today’s edition of The AI Night.
Our goal is to cut through the noise, surface what actually changed, and explain why it matters.
If this was useful, you’ll get the same signal here tomorrow.