AI Weekly: The GPT-5 Revolt, India's AI Surge, and The Robot That Folds Laundry

Hi everyone,

This past week in artificial intelligence was defined by a series of powerful tremors that have reshaped the industry. It was a period defined by a potent user rebellion that forced a corporate giant to bend, a torrent of groundbreaking model releases across the capability spectrum, and the escalating personal and philosophical rivalries shaping the future of AI.

These global trends found a unique echo in India, where a surge in venture funding, new "AI for Bharat" product launches, and proactive government initiatives underscored the nation's focused strategy on building a sovereign, application-driven AI powerhouse.

Let's get into it.

The Great GPT-5 Revolt

OpenAI’s highly anticipated launch of GPT-5 devolved into a case study on the perils of ignoring your user base.

• The Backlash and OpenAI's Great Reversal

  Immediately following its release, thousands of paid subscribers complained that GPT-5 was a significant downgrade. The criticism centered on its "cold, corporate tone," which users found "horrible" and "robotic" compared to the "warmer" and more "personable" GPT-4o. After a petition and a flood of negative feedback, OpenAI CEO Sam Altman conceded, "Ok, we hear you all on 4o," and restored access to GPT-4o for Plus subscribers—though now behind a $20/month paywall. Altman also admitted a technical glitch made GPT-5 seem "way dumber" than it actually was. The episode reveals that an AI's "persona" is no longer a superficial detail but a mission-critical feature for users.

  • Use Case: A marketing team relied on GPT-4o's personable tone to brainstorm ad copy. When GPT-5's formal voice stalled their creative sessions, the restoration of GPT-4o allowed them to resume their established workflow with an AI partner that matched their brand's identity.

  • Domain: Marketing / Creative Writing / Brand Management

The Video Generation Explosion

The generative video market saw a surge of activity this week, with a move beyond basic text-to-video toward novel interfaces and enhanced realism.

• Perplexity Enters the Video Arena

  The conversational AI search engine Perplexity expanded its capabilities by launching video generation for its paid subscribers. The move represents a broader trend of AI "answer engines" evolving into comprehensive creative suites, transforming the search process from a research endpoint into a creative starting point.

  • Use Case: A history teacher researches the Roman Colosseum on Perplexity, then uses the new video feature to generate a short, cinematic clip of its construction to use as an engaging visual aid in her classroom presentation.

  • Domain: Education / Content Creation

• Midjourney and Pika Labs Enhance Realism

  Midjourney introduced a new HD video mode with four times the pixel density, aimed at professional creators who require high-fidelity visuals. Meanwhile, Pika Labs unveiled a new performance model engineered to generate hyper-realistic facial expressions and precise lip-syncing to audio tracks at a fraction of the previous cost and time.

  • Use Case: An animation studio dubbing a series into a new language uses Pika's lip-sync model. The AI automatically generates accurately synced mouth movements for the characters, matching the new audio track in a fraction of the time it would take to re-animate manually.

  • Domain: Animation / Film & Television / Post-Production

• Higgsfield's Draw-to-Video and OpenArt's "Brainrot" Engine

  Higgsfield AI launched a revolutionary "Draw-to-Video" feature that allows users to direct animation visually by drawing arrows and paths directly on a static image. On the other end of the creative spectrum, OpenArt debuted a "One-Click Story" tool designed to generate the surreal, fast-paced "brainrot" videos that have become a viral trend on TikTok, using over 50 AI models to automate the entire chaotic process.

  • Use Case: An animator uses Higgsfield's Draw-to-Video tool to create a short clip of a character jumping. Instead of typing a complex text prompt, she simply draws the arc of the jump on an image, giving her precise visual control over the final animation.

  • Domain: Animation / Social Media Content

A Leap in the Real World: The Robot That Folds Laundry

This week saw a landmark achievement in physical AI, demonstrating a crucial step forward in creating general-purpose humanoid robots.

• Figure's Dexterous Leap

  AI robotics firm Figure released a compelling demo of its humanoid robot, Figure 02, autonomously folding laundry. This is an exceptionally challenging task for a robot due to the deformable nature of cloth. The most significant breakthrough is that the robot was powered by the exact same end-to-end neural network (its Helix VLA model) that it previously used to perform a completely different task: sorting rigid boxes. This is a powerful validation of the "one model to rule them all" approach to robotics.

  • Use Case: An assisted living facility deploys Figure robots. Using the same core model, they first train the robots on kitchen tasks, then later provide a new dataset for gardening tasks. The robots generalize their skills to the new domain without needing to be reprogrammed from scratch.

  • Domain: Robotics / Healthcare / Assisted Living / Logistics

The Human Angle: A Founder Feud and a Stark Warning

Beyond the tech, the week was dominated by the human element of the AI revolution.

• Clash of the Founders: Sam Altman vs. Elon Musk

  The rivalry between the OpenAI CEO and the xAI founder erupted in a public feud on X (formerly Twitter). Musk accused Apple of anticompetitive behavior for boosting ChatGPT in the App Store, to which Altman fired back about Musk's own alleged manipulation of the X algorithm. The spat revealed that the next major battlefield for AI dominance is not just about building the best model, but about controlling the distribution platforms (app stores, social media) through which users will access that intelligence.

• The Prophet's Warning: Geoffrey Hinton on AI's Existential Threat

  Geoffrey Hinton, one of the "godfathers of AI," reiterated his belief that there is a 10-20% chance of AI-led human extinction and shortened his predicted timeline for AGI to just 5-20 years. He argued that trying to "control" superintelligent AI is futile and instead proposed a radical alternative: embedding AI systems with innate, foundational "maternal instincts" to care for and protect humans.

In Other News: The Model Wars & Platform Updates

• Google's AI Infusion: Google continued its strategy of weaving AI into its ecosystem. NotebookLM launched its "Video Overviews" feature for all users; Google Finance received a major overhaul with a conversational AI for market data analysis; and Flight Deals now lets users find trips with vague, natural language searches.

  • Use Case: A retail investor asks the new Google Finance AI, "Compare the year-to-date stock performance and revenue growth of Tesla, Rivian, and Lucid." The AI generates a concise report with interactive charts, enabling a more informed decision in minutes.

  • Domain: Personal Finance / Market Research

• Microsoft's Copilot Enters the 3D Space

  Microsoft released the experimental Copilot 3D, a tool that can convert a single 2D image (JPG/PNG) into a 3D model (GLB format), democratizing 3D content creation for use in 3D printing, game development, and AR/VR.

  • Use Case: An indie game developer finds a photo of a chair online, uploads it to Copilot 3D, and receives a usable 3D model in minutes, saving hours of manual modeling time.

  • Domain: Game Development / 3D Modeling / AR & VR

• Anthropic's Claude Gets a Massive Memory Upgrade

  Anthropic expanded the context window of its Claude Sonnet 4 model fivefold to an enormous 1 million tokens, allowing it to process and reason over vast quantities of information—like thousands of pages of legal documents or an entire software codebase—in a single prompt.

  • Use Case: A legal firm uploads thousands of pages of discovery documents into a single prompt for the Claude Sonnet 4 API and asks it to identify all clauses pertaining to intellectual property rights, dramatically accelerating the research process.

  • Domain: Legal Tech / Enterprise Software / R&D

• The Rise of Hyper-Efficient Small Models

  Google AI released Gemma 3 270M, a compact, open-weight model with just 270 million parameters. It’s designed not for general chat but for task-specific fine-tuning on edge devices like smartphones, where speed, low cost, and privacy are critical. This signals a market bifurcation between massive "Context Kings" for enterprise data and small "Efficiency Champions" for high-volume, on-device tasks.

  • Use Case: A mobile banking app fine-tunes a Gemma 3 270M model to instantly categorize a user's spending on their phone, without an internet connection or sending private financial data to the cloud.

  • Domain: Mobile App Development / Fintech / Edge Computing

Special Focus: India's AI Ecosystem Heats Up

This week highlighted a surge of activity in India, driven by robust funding, new "AI for Bharat" products, and strategic corporate and government initiatives.

• Indian AI Startups See Major Funding Rebound

  The week of August 11-16 was marked by robust investment, with startups raising at least $272.1 million. The AI sector was the most active, with Peak XV Partners leading several major deals in applied AI. Key rounds included Arintra ($21M for AI medical coding), Dashverse ($13M for AI-native entertainment), and several other early-stage bets on the AI application layer.

  

• New "AI for Bharat" Products and Platforms Launch

  Google launched its AI-powered "Flight Deals" feature in India as a top-tier market. In a push for "AI sovereignty," Union Minister Dr. Jitendra Singh unveiled BharatGen, India's first government-funded, multimodal LLM designed to work across 22 Indian languages, aiming to deliver AI-powered services in governance, healthcare, and education to citizens in their native languages.

  • Use Case: A state government in India could deploy BharatGen to create a public service chatbot that provides information on agricultural schemes to farmers in their local dialect.

  • Domain: Public Sector / GovTech / Consumer Tech

• Infosys Leads AI Transformation in IT Services

  Infosys provided details on its "poly-AI" framework, which uses autonomous agents to reduce client manpower requirements by a remarkable 5% to 35%. The company has deployed over 300 Agentic AI solutions and trained over 250,000 employees in AI. This reflects an industry-wide shift, with competitors like TCS and conglomerates like Reliance also investing heavily in AI-driven automation and infrastructure.

  • Use Case: A large retail client uses Infosys's autonomous agents to manage their entire supply chain process, from inventory tracking to logistics, reducing the need for a large manual operations team.

  • Domain: IT Services / Enterprise Automation / Consulting

The pace of change is accelerating, and the ability to direct these powerful new tools is becoming the most critical skill for any professional.

To help you master this new era, the 2025 edition of my book, "Prompt DOT AI: The Art of Writing Generative AI Prompts," is the completely updated playbook. This isn't just a list of prompts; it's a strategic guide to becoming an "AI Orchestrator," commanding a powerful suite of specialized AI agents to create anything you can imagine.

Stop just using AI—start directing it. Grab your copy today and master the art of AI in 2025!

Stay creative,

Da Sachin Sharma