Google DeepMind Partners With Live-Action Filmmakers to Craft Veo Short Feature

⚡️ Headlines

🤖 AI

Kimi‑Dev 72B open‑source coding LLM released: Kimi‑Dev 72B, a 72‑billion‑parameter model optimized via large-scale reinforcement learning using real-world code, achieves a new open‑source high of 60.4% on SWE‑bench Verified (Hugging Face).

Google to split from Scale AI after Meta deal: Google, Scale AI's largest customer, plans to end its partnership after Meta acquired a 49% stake, amid rising data‑security concerns among major AI companies (Reuters).

Thousands of UK university students caught cheating via AI: Nearly 7,000 AI‑assisted cheating cases were confirmed in UK universities during 2023–24—roughly 5.1 per 1,000 students—highlighting a sharp surge in academic misconduct (The Guardian).

Google expands Gemini 2.5 model family: Gemini 2.5 Flash and Pro are now generally available, and a new cost-efficient Flash‑Lite preview is introduced as part of Gemini’s high-speed, hybrid‑reasoning model lineup (Google Blog).

Adobe launches Firefly AI mobile app on iOS/Android: Adobe’s Firefly mobile app lets users generate and edit images/videos on the go—syncing to Creative Cloud and integrating third-party generative models, with premium features using Firefly credits (The Verge).

Anysphere unveils $200/mo “Ultra” plan for Cursor AI: Anysphere introduced an "Ultra" tier at $200/month, offering 20× the usage of its AI coding tool Cursor across multiple model providers plus priority feature access (TechCrunch).

Sam Altman: Meta’s $100 M offers failed to poach top OpenAI talent: OpenAI CEO Sam Altman said Meta dangled nine‑figure signing bonuses up to $100 M to recruit his top researchers, but none of them left—citing loyalty to OpenAI’s mission and culture (TechCrunch).

🤳 Social Media

TikTok Symphony AI tools for advertisers: TikTok’s Symphony Creative Studio now offers generative AI tools that convert images, text, and product links into TikTok‑style videos within minutes—ideal for marketers looking to streamline content creation (TikTok).

Reddit Community Intelligence unveiled live from Cannes Lions 2025: Reddit introduced Community Intelligence, an AI‑powered platform offering real‑time trend tracking and ad‑integration tools aimed at helping brands harness authentic community insights (Reddit).

Threads adds dedicated fediverse feed and search: Meta’s Threads added a separate feed showcasing posts from federated networks like Mastodon and introduced user search across the fediverse to enhance cross-platform discovery (The Verge).

WhatsApp ads debut in Status tab: For the first time, WhatsApp is rolling out ads in its Status section (analogous to Stories), keeping them separate from personal chat threads (NYT).

Meta AI Discover feed exposes private queries, company issues fix: Meta’s public “Discover” AI feed unintentionally surfaced private user queries; Meta has since applied a fix and is re-evaluating visibility policies (Business Insider).

Facebook converting all uploaded videos into Reels: Meta confirmed every new Facebook video will now be categorized and displayed as a Reel, with the Video tab being renamed accordingly (TechCrunch).

🔬 Research

Seizing the agentic AI advantage: This report outlines how enterprises can overcome the "gen‑AI paradox"—high adoption but low bottom‑line impact—by leveraging autonomous AI agents with scalable infrastructures and rethinking workflows (McKinsey).

⚖ Legal

TikTok ban delayed 90 days by Trump administration: Former President Trump has postponed enforcement of the U.S. TikTok ban by 90 days, citing ongoing negotiations and legal challenges (CNN).

New York State updates WARN notices to flag AI-related layoffs: New York now requires companies to label layoffs caused by AI adoption in WARN notices, adding transparency to workforce automation trends (Bloomberg).

Senate passes GENIUS Act on stablecoins: The U.S. Senate passed the GENIUS Act, establishing federal guidelines for stablecoins while excluding restrictions on presidential family holdings; the bill now heads to the House (CNBC).

Musk’s X sues New York over hate speech disclosure law: X (formerly Twitter) is suing New York State over a new law requiring platforms to report how they monitor hate speech, arguing it violates free speech protections (Bloomberg).

🔌 Plug-Into-This

“Behind “ANCESTRA:” combining Veo with live‑action filmmaking”

Google DeepMind partnered with Darren Aronofsky’s new Primordial Soup studio and director Eliza McNitt to create ANCESTRA, a short film that seamlessly blends live‑action cinematography and AI‑generated video using the Veo model. The project highlights Veo’s new capabilities for personalized, emotion-driven shots and precise motion matching, marking a milestone in generative video within a professional filmmaking context.

• The film, premiering at Tribeca on June 13, 2025, pairs human performance with AI‑generated sequences to tell a deeply personal story of a mother and child with congenital heart issues.

• Google DeepMind assembled a multidisciplinary team—over 200 people including filmmakers, VFX artists, technologists—to co-create the film’s AI-enhanced visuals.

• The generative pipeline used Gemini for prompt design, Imagen for image generation, and Veo for sequence generation aligned with mood, color, and tone.

• New Veo features support personalized video, precision motion matching, and seamless integration of AI-generated and live-action footage.

• This represents part one of a planned trio of shorts between DeepMind and Primordial Soup, signaling a broader push to embed generative AI into mainstream film production.

🎥 This pilot presents AI as a creative collaborator—not a replacement—helping filmmakers visualize metaphysical themes and highly emotive moments. It’s part of a generalized response to fears in Hollywood around replacement of creatives.

OpenAI and Microsoft tensions are reaching a boiling point

OpenAI and Microsoft’s partnership, founded in 2019, has soured amid disagreements over OpenAI’s shift toward a for-profit structure, ownership stakes, compute access, and control of acquired IP, notably Windsurf. OpenAI is reportedly exploring antitrust action as a means to reshape the relationship.

• OpenAI requires Microsoft’s consent to convert into a public-benefit company and raise external capital; delays threaten a potential $20 billion funding shortfall.

• Central contention involves OpenAI’s $3 billion acquisition of Windsurf—a direct competitor to Microsoft-backed GitHub Copilot—with Microsoft demanding access to Windsurf’s IP.

• Microsoft insists on retaining a larger stake and continued access to future OpenAI models post-“AGI”—a stance OpenAI views as overreach.

• OpenAI is pursuing diversification of its cloud partnerships—potentially with Google Cloud—reducing reliance on Microsoft Azure.

• Despite escalation, both sides maintain optimism publicly, characterizing ongoing discussions as constructive.

🥊 OpenAI and Microsoft are now at loggerheads over control and strategy—OpenAI wants freedom, Microsoft wants security, and the end result could dramatically reshape the AI industry. This power shift reflects deeper structural tensions in AI growth models: vertical alignment versus open, multi-cloud competition, with long-term implications for regulatory scrutiny and strategic autonomy.

OpenAI wins $200 million U.S. defense contract

OpenAI has secured a one-year, $200 million contract from the U.S. Department of Defense to prototype frontier AI capabilities for warfighting and enterprise applications through July 2026, under its new OpenAI for Government initiative.

• Announced June 16, the contract marks OpenAI’s first direct DoD engagement, following policy shifts allowing military applications.

• The award supports AI systems aimed at healthcare, acquisition analytics, and proactive cyber defense for the military.

• OpenAI launched “OpenAI for Government” to streamline public sector deployments across federal agencies.

• This entry into defense contrasts sharply with earlier policy that banned military use of AI tools, a stance reversed in January 2024.

• Analysts suggest this dovetails with OpenAI’s work with Anduril and participation in the broader “Stargate” infrastructure push.

🔧 OpenAI is now in the defense business—helping the military use its AI like it uses ChatGPT, but for serious things like cyber defense and healthcare. With this, OpenAI becomes a key AI tech provider in national security, likely triggering a broader debate over military AI ethics, governance, and oversight…though it’s been quiet so far.

 🆕 Updates

📽️ Daily Demo

🗣️ Discourse