Reports

Tue, Jun 3 - OpenAI's Super Assistant Plan and Google's Edge Gallery Launch

Hak

03 Jun 2025 — 4 min read

It's Tuesday, June 3, 2025, and you're reading the Agentive Daily Report.

Busy People's Section

OpenAI leaked document reveals plans to position ChatGPT as a "super assistant" interface to the internet with upcoming o3-pro model.

Google quietly released AI Edge Gallery, allowing users to download and run AI models locally on Android phones without internet connection.

Kling 2.1 emerges as key competitor to Google's Veo 3 for AI video generation, offering high-quality video creation from static images.

ChatGPT discovered CVE-2025-37899, a critical zero-day vulnerability in the Linux kernel that security researchers had missed.

Anthropic hit $3 billion in annualized revenue, tripling from $1B in just five months, largely driven by enterprise AI adoption.

Samsung and Perplexity near deal to integrate AI search technology across Samsung devices, including preloaded apps and browser integration.

Today's Top Stories

OpenAI's "Super Assistant" Strategy Leaks in Court Document

A heavily redacted internal document from OpenAI, revealed during the Justice Department's antitrust case against Google, outlines the company's vision to transform ChatGPT into a comprehensive "super assistant" that serves as the primary interface to the internet. The strategy details plans to launch o3-pro soon and develop capabilities that would allow ChatGPT to manage calendars, book travel, navigate software, and even contact professionals on users' behalf.

The document describes the assistant's capabilities as "T-shaped" – broad enough for daily tasks yet deep enough for specialized work. OpenAI is building its own search index to reduce dependence on traditional search engines and is focusing on integration with hardware. Revenue models appear to be a secondary concern, with the company prioritizing usage growth in the first half of 2025 before exploring monetization strategies in the second half of the year.

Google Launches AI Edge Gallery for Local Model Execution

Google has quietly released AI Edge Gallery, an experimental Android application that allows users to download and run open-source AI models directly on their smartphones without requiring an internet connection. The app enables on-device processing for tasks like image generation, code writing, and question answering using models such as Gemma 3n.

This move represents a significant step toward device-level AI independence, giving users access to powerful AI capabilities while maintaining privacy through offline processing. The app features a configurable "Prompt Lab" for single-turn interactions, with performance dependent on device hardware and model size. Google's commitment to bringing AI capabilities closer to the hardware level parallels industry-wide efforts to balance cloud-based AI services with local processing options.

Kling 2.1 Challenges Google's Veo 3 in AI Video Generation

Kuaishou's new AI tool, Kling 2.1, has emerged as a serious competitor to Google's Veo 3 in the rapidly evolving AI video generation space. The tool allows users to create high-quality videos from static images using text prompts to guide movement and transformation. Kling 2.1 can generate 1080p videos in under a minute, with pricing based on an internal credit system (approximately $1.50 per high-detail video).

This development signals a significant acceleration in AI video tool democratization, making professional-quality video creation more accessible and affordable. Unlike earlier generations of video tools, Kling 2.1 produces results quickly and with greater precision to text prompts. The competition between Kuaishou, Google, and other players in this space is driving rapid innovation, with millions of AI-generated videos already created with Veo 3 alone in the past few days, according to DeepMind CEO Demis Hassabis.

Fast Forward

AI Security Breakthrough: ChatGPT discovered CVE-2025-37899, a previously unknown remote vulnerability in the Linux kernel's SMB implementation, demonstrating AI's potential to find security flaws that traditional researchers might miss. While this shows AI's value for cyber defense, it also raises concerns about malicious exploitation.
Anthropic Revenue Surge: Anthropic's revenue jumped from $1 billion to $3 billion in annualized revenue in just five months, primarily driven by enterprise adoption of AI coding tools powered by Claude. As demand for enterprise AI solutions accelerates, the company is expanding capabilities while considering a potential AI usage tax.
Samsung-Perplexity Partnership: Samsung is nearing a deal to integrate Perplexity's AI search throughout its ecosystem, including preloaded apps on devices and integration with the Samsung browser. The technology may also enhance Samsung's Bixby assistant, with features likely to debut on the Galaxy S26 lineup in early 2026.
Meta's AI-Powered Risk Assessment: Meta plans to automate 90% of its internal risk and data protection reviews using AI, shifting from human-led evaluations to algorithmic systems. The company will spend over $8 billion on this transition, focusing initially on low-risk decisions while maintaining human oversight for critical safety features.
Mary Meeker's AI Analysis: Legendary tech analyst Mary Meeker released a comprehensive 340-slide report documenting unprecedented AI adoption rates, with ChatGPT achieving global penetration in three years compared to the internet's 23-year timeline. The report highlights how AI chatbots are now being mistaken for humans 73% of the time, up from 50% six months ago.

New Tools Discovered

Chatterbox: An open-source voice cloning model from Resemble AI that can mimic any voice using just 5 seconds of audio, outperforming ElevenLabs in user preference tests.
Circuit Tracer: Anthropic's open-source toolkit that visualizes how AI models process information internally, offering transparency into AI decision-making.
Google AI Edge Gallery: A platform enabling developers to run AI models locally on Android devices without internet connection, prioritizing privacy and offline capabilities.
Perplexity Labs: A productivity tool that generates reports, spreadsheets, dashboards, and interactive web apps using AI, with results organized in a dedicated workspace.
TextFX: Google's free creative writing assistant offering 10 specialized tools for writers and content creators to enhance and transform text.

Discover more tools at Agentive.Directory

That's a wrap for today! Thank you for reading this report.

Have thoughts on today's edition? Hit reply and let us know what you're thinking. Or if you've discovered a cool AI tool we should feature, drop us a line.

Until tomorrow,
Hak from Agentive.Studio