Adobe Premiere Pro Finally Lets It’s Video Extend Tool Out Into The Wild

⚡️ Headlines

🤖 AI

Amazon's new AI agent is designed to do your shopping - Amazon introduces Nova Act, an AI model capable of performing web searches, making purchases, and scheduling tasks directly within a browser. [The Verge]

ChatGPT users have generated over 700m images since last week, OpenAI says - OpenAI reports that ChatGPT users have created over 700 million images in the past week, highlighting the platform's growing engagement. [TechCrunch]

AI video startup Runway valued at $3 billion in funding round - Runway, an AI video creation startup, reaches a $3 billion valuation following its latest funding round. [Bloomberg]

Spotify Steps Up Advertising Push With New Tools, Partnerships - Spotify unveils new advertising tools and partnerships aimed at boosting its ad revenue and platform capabilities. [Investor's Business Daily]

OpenAI and Anthropic are fighting over college students with free AI - OpenAI and Anthropic launch competing educational initiatives to provide free AI tools to college students, aiming to become the default AI resource in academia. [The Verge]

Amazon's internal forecast suggests a $700 million financial gain from its AI shopping assistant Rufus - Amazon projects that its AI shopping assistant, Rufus, will indirectly generate over $700 million in operating profit by 2025. [Business Insider]

Adaptive Security raises $43 million - Adaptive Security secures $43 million in funding to enhance its cybersecurity solutions. [Adaptive Security]

Google in advanced talks to rent Nvidia AI servers from CoreWeave - Google is reportedly in advanced discussions to lease Nvidia AI servers from CoreWeave to bolster its AI infrastructure. [The Information]

AI lie detector: How Halloumi's open-source approach to hallucination could unlock enterprise AI adoption - Halloumi's open-source method for detecting AI hallucinations aims to enhance trust and facilitate enterprise AI adoption. [VentureBeat]

🦾 Emerging Tech

Riot Platforms hits post-halving Bitcoin production high as it expands AI capacity - Riot Platforms achieves its highest Bitcoin production since the halving and plans to expand its AI and high-performance computing capabilities. [CoinDesk]

🤳 Social Media

X (formerly Twitter) releases luxury fashion trends report - X, previously known as Twitter, publishes a report highlighting current trends in the luxury fashion industry. [Social Media Today]

TikTok for Artists gets tested as Trump mulls ‘TikTok America’ plan - TikTok begins testing a new platform for artists amid discussions of a potential U.S. ownership plan. [Music Ally]

🔬 Research

Reasoning Models Don't Always Say What They Think - Anthropic's research paper examines the faithfulness of chain-of-thought reasoning in AI models, revealing that models often do not verbalize their use of reasoning hints. [Anthropic]

🎱 Random

Intel, TSMC Recently Discussed Agreement to Form Chipmaking Joint Venture - Intel and TSMC have reportedly engaged in discussions to form a joint venture focused on chip manufacturing. [The Information]

Microsoft turns 50 - Microsoft celebrates its 50th anniversary, marking five decades since its founding on April 4th, 1975. [The Verge]

Superman’s latest trailer introduces his team of robot assistants - The new Superman trailer showcases the superhero's team of robot assistants, adding a futuristic twist to the classic narrative. [The Verge]

Are new Google E2EE emails really end-to-end encrypted? Kinda, but not really - An analysis reveals that Google's new end-to-end encrypted emails may not offer complete encryption as initially perceived. [Ars Technica]

🔌 Plug-Into-This

Adobe Introduces AI-Powered Features in Premiere Pro

Adobe has launched a suite of generative AI features for Premiere Pro, including video extension, advanced search, and automated caption translation. Powered by its Firefly Video Model, the release underscores Adobe’s strategy to embed generative tools directly into creative workflows.

• Generative Extend allows users to add up to two seconds of video and ten seconds of ambient audio, enabling smoother transitions or bridging footage gaps.

• AI-driven Media Intelligence supports semantic search across large video libraries using natural language prompts.

• New caption translation capabilities support 27 languages, streamlining international distribution and localization.

• Updated color management automatically converts log and raw video into HDR or SDR formats, simplifying the color grading process.

• The entire toolset benefits from enhanced performance across both Apple silicon and Windows machines.

Think of this as Adobe giving editors a smart assistant that can fill in missing footage, translate dialogue, and instantly find the right clip—all without leaving Premiere.

🎥 Adobe’s focus on integrated AI editing hints at a broader shift: instead of externalizing generative tools, creative suites are becoming intelligent environments where generative capabilities are ambient and invisible—baked into every step of production.

Google DeepMind Says AGI Might Outthink Humans by 2030—and It's Planning for the Risks

DeepMind has publicly outlined its timeline and risk-mitigation plan for Artificial General Intelligence (AGI), projecting systems that could surpass human cognition in most domains by 2030. The company is advancing both technical and governance mechanisms to preempt misuse or misalignment.

• AGI is defined by DeepMind as systems capable of outperforming humans at nearly all economically valuable cognitive tasks.

• Key concerns include model misuse, catastrophic accidents, value misalignment, and structural power imbalances.

• The company is developing evaluation benchmarks and access restrictions to minimize risk before deployment.

• DeepMind acknowledges uncertainty in timelines, but is orienting safety planning around the possibility of AGI within five years.

• Emphasis is placed on aligning AGI behavior with human intent through interpretability, robustness, and continual oversight.

🧠 While the 2030 prediction is bold, the real signal here is institutional: DeepMind is shifting from AGI speculation to AGI scenario planning—treating advanced general systems as a near-term engineering and governance challenge, not a distant future abstraction.

Midjourney Releases Its First New AI Image Model in Nearly a Year

Midjourney has unveiled V7, a substantial upgrade to its image-generation platform emphasizing personalization, fidelity, and efficiency. The release marks the company’s return to rapid iteration, with major changes to both model behavior and user interaction.

• V7 introduces a new personalization layer where users rate ~200 images to fine-tune generation to their aesthetic preferences.

• Text prompt comprehension has improved, leading to more accurate and nuanced image outputs.

• Significant gains in visual coherence, particularly in rendering hands, faces, and complex spatial arrangements.

• Draft mode offers fast, low-cost previews that can be upscaled later, optimizing both speed and pricing.

• Turbo and Relax generation modes remain, giving users flexibility between performance tiers.

To put it simply: Midjourney now learns your visual taste—and creates faster, more accurate images that actually reflect it.

🖼️ Personalization via preference training signals a broader shift in generative UI: from prompt engineering to profile shaping. As models improve, the bottleneck is no longer model capability—it’s user alignment and expressive control

 🆕 Updates

📽️ Daily Demo

🗣️ Discourse