Sam Altman Stole Steve Jobs's Designer To Kill The iPhone (+18 AI Updates)
This transcript covers 18+ major AI updates from a single week, including Sam Altman's plans to build an iPhone competitor with Apple's legendary designer Jony Ive, Meta's failed attempt to acquire Chinese AI tool Manus AI, and Google's controversial $200 million military AI deal. The video also features a deep dive tutorial on Grok's new Imagine Agent Beta, an infinite canvas creative tool capable of generating images, videos, storyboards, and full ad campaigns.
Summary
The video opens with three headline stories: Sam Altman partnering with Jony Ive (the designer behind every iconic Apple product for 30 years) to build an AI-first smartphone meant to replace traditional apps with AI agents; Mark Zuckerberg's failed $17,000 crore bid to acquire Chinese AI company Manus AI, which was blocked by China; and Sam Altman giving all ChatGPT Plus and Pro users free access to OpenClaw agent without requiring separate API credits.
On the hardware and platform front, Manus AI also launched a 'Cloud Computer' product that runs bots, Python scripts, and software 24/7 without users needing to manage servers or code. OpenAI and Microsoft renegotiated their landmark partnership, allowing OpenAI to now distribute its models across all cloud platforms (AWS, Google Cloud, etc.) rather than exclusively through Microsoft Azure, with the arrangement extending to 2032.
Several creative and productivity tools received major updates. Adobe integrated 50+ Creative Cloud tools directly into Claude, enabling photo editing, Instagram story creation, and video reformatting within the chat interface. Google's Gemini app launched the ability to generate Docs, Sheets, Slides, and PDFs directly in chat. Google's NotebookLM introduced auto-labeling and categorization for sources. ElevenLabs launched both '11 Music' (a platform combining music creation, streaming, and monetization) and real-time collaboration in '11 Flows' for team-based creative pipelines.
Other notable updates include: Lovable launching a mobile app for building web apps from a phone; Snapchat introducing AI Sponsored Snaps allowing users to interact with brand AI agents; Telegram launching 'Lobster Father' for no-code AI agent creation; Google Translate adding real-time pronunciation feedback powered by AI; Google Photos introducing an AI wardrobe feature that scans past photos to build a digital closet; and OpenAI releasing an official prompting guide for GPT-4.5 recommending short, outcome-focused prompts over step-by-step instructions.
The video's main tutorial segment covers Grok's Imagine Agent Beta, an infinite canvas creative tool. The presenter demonstrates three use cases: generating a short film with storyboards, cinematic teaser, and music from a single prompt; creating consistent multi-angle product photography and a full Instagram video ad from basic t-shirt photos; and building detailed fantasy world concepts with infographics and animated scenes. The presenter argues this tool compresses what would normally require multiple professionals and days of work into minutes.
Key Insights
- Meta attempted to acquire Chinese AI company Manus AI for approximately $17,000 crores, but China blocked the deal, preventing Zuckerberg from obtaining a tool the presenter describes as 'widely considered better than anything OpenAI or Anthropic has put out so far.'
- Google signed a $200 million classified AI deal with the US Pentagon, directly contradicting its own 2018 public promise that its AI would never be used for weapons or surveillance, prompting over 600 employees to sign an open letter to CEO Sundar Pichai urging the company to reject military AI work.
- OpenAI's official prompting guide for GPT-4.5 argues that short prompts defining only the desired outcome produce better results than step-by-step instructions, because detailed instructions 'narrow the model search space and lead to stiff mechanical responses.'
- Grok's Imagine Agent Beta generated a complete Instagram video ad — including consistent model appearances across five different poses, motion, transitions, and music — from just two basic product photos of a t-shirt, a process the presenter says would otherwise cost 'a few lakh rupees' involving models, photographers, studios, and editors.
- The revised OpenAI-Microsoft partnership ends the requirement for companies using OpenAI models to exclusively use Microsoft Azure, allowing OpenAI to now distribute across all cloud platforms including AWS and Google Cloud, while Microsoft remains the primary partner through 2032 with a revenue share maintained through 2030.
Topics
Full transcript available for MurmurCast members
Sign Up to Access