▲ New ChatGPT Update: AI Steps Up for Complex Tasks

Welcome future tech & AI enthusiasts — this week’s highlights:

ChatGPT Agent: AI Steps Up for Complex Tasks

Hume’s Voice Model: Personality Cloning Breakthrough

Lovable: Sweden’s “Vibe Coding” Unicorn Rockets to $1.8B

Thinking Machine Labs: $2B Seed Round and Product Tease

Runway’s Act-Two: AI Motion Capture Moves to a New Level

Let’s get into it.

In today’s issue:

• Learn: How to create a product ad with Freepik

• Learn: AI image prompt for Product photography

▲ AI AGENTS

ChatGPT Agent: AI Steps Up for Complex Tasks

OpenAI unveils an agentic version of ChatGPT that blends research, reasoning, and direct action.

What you need to know

• Launch: On July 17, OpenAI introduced “ChatGPT Agent,” an upgrade that turns ChatGPT into a proactive helper that can autonomously execute multi-step tasks by combining web navigation, analysis, and productivity tools.

• Capabilities: Users can have ChatGPT book appointments, buy groceries, create slide decks, analyse emails, and interact with various platforms — all managed through a secure virtual computer. The agent leverages skills from OpenAI’s Operator and Deep Research functionalities, switching seamlessly between browsing, data entry, and summarising findings.

• Availability: The agent is accessible to Pro, Plus, and Team subscribers, with further enhancements planned in coming months

Key Ideas

▲ ChatGPT is now moving from conversational AI to an assistant, capable of handling real-world workflows and bridging the gap between AI conversations to tangible outcomes.

▲ AI VOICE

Hume’s Voice Model: Personality Cloning Breakthrough

New EVI 3 model promises hyper-realistic voice and personality mimicry. Watch product video demo here.

What you need to know

• Launch: Hume debuted its latest Empathic Voice Interface (EVI 3) with an advanced voice cloning feature that can replicate not only your voice but also aspects of your personality, speech rhythms, and emotional style.

• How it Works: With a short 30–90 second audio sample, users can create a digital double that mimics their vocal cadence, emphasis, intonation, and conversational quirks.

• Limitations & Impressions: Early testers note striking realism in voice synthesis, including pauses and inflections, but the personality mimicry sometimes exaggerates traits or fails to capture nuanced behaviours. The system refuses to “fake” accents but is described as unnervingly lifelike.

• Use Cases: Potential for deeply personalised virtual assistants, content creation, and accessibility tools. Trained on millions of hours of speech and trillions of text tokens, it’s a leap forward for expressive AI

Key Idea

▲ EVI 3 blurs the line between real and virtual presence, bringing highly individualised voice interfaces and ethical questions to the mainstream.

▲ AI STARTUPS

Lovable: Sweden’s “Vibe Coding” Unicorn Rockets to $1.8B

AI-driven no-code platform lands $200M Series A led by Accel.

What you need to know

• Funding Milestone: Swedish startup Lovable, founded in 2023, has secured a $200M Series A at a $1.8B valuation — one of Europe’s fastest unicorn ascents of 2025.

• Product: Lovable’s platform lets users design full-featured web apps and websites through natural language conversations, democratising software development with AI-powered “vibe coding.” Non-coders can collaborate, import Figma designs, and deploy complex projects easily.

• Growth Stats: The platform boasts over 2.3M users, 10M+ created projects, $75M in annual recurring revenue, and enterprise clients like Klarna and HubSpot.

Key Idea

▲ Lovable’s meteoric rise shows how AI can make software creation accessible to everyone, powering a new generation of “citizen developers.”

▲ AI FUNDING

Thinking Machine Labs: $2B Seed Round and Product Tease

Mira Murati’s new lab raises record funding as first open-source AI product nears debut.

What you need to know

• Mega-Round: Thinking Machine Labs, founded by ex-OpenAI CTO Mira Murati, raised $2B in seed funding, valuing the startup at $12B — the largest-ever seed round in AI history.

• Team: The powerhouse group includes several key OpenAI alumni specialising in multimodal AI, safety, and foundational research.

• Product Preview: The company is finalising its first product, described as multimodal AI for conversational and visual interaction, with a significant open-source component targeted at researchers and startups.

• Mission: The lab wants to advance safer and more transparent AI, sharing research to benefit the wider community. The product launch is expected “in the next couple months”

Key Idea

▲ Massive investor bets signal continued confidence in top-tier AI teams and the value of open innovation, even before product details are public.

▲ AI VIDEO

Runway’s Act-Two: AI Motion Capture Moves to a New Level

No sensors, no suits. Just smarter video animation for everyone. Watch the fun demo video here.

What you need to know

• Launch: Runway has released “Act-Two,” a next-generation AI motion capture tool for video, games, and animation. No need for traditional mocap suits or hardware.

• Improvements: The update offers more accurate tracking of head, face, hands, and full body, allowing users to animate digital characters from a single video input. The system adapts to diverse styles and environments, delivering expressive, lifelike motion.

• Workflow: Act-Two streamlines the creative process with a user-friendly interface and supports both novices and pros. Enhanced narrative tools, effects, and music integration make high-quality video production more accessible.

Key Idea

▲ Act-Two brings advanced animation by turning the camera into a studio, reducing barriers for storytellers and digital creators.

▲ EVEN MORE NEWS

Other big ideas from the past week.

• OpenAI’s Model Smashes Math Olympiad: OpenAI’s latest experimental reasoning model achieved a “gold medal” score at the 2025 International Math Olympiad, solving 5 of 6 problems and hinting at the imminent launch of GPT-5.

• Google’s Big Sleep Battles Cybercrime: Google launched “Big Sleep,” an AI tool to detect and disable dormant web domains vulnerable to cyberattacks, aiming to curb phishing and malware abuse.

• Meta’s $14.8B AI Infrastructure Investment: Meta poured $14.8B into Scale AI to help build out its AI infrastructure.

The Lab

▲ AI EDUCATION

How to create a product ad with Freepik

What began as an experimental journey to test the full range of Freepik's AI capabilities resulted in a streamlined workflow for creating a high-quality product video ad. This tutorial documents the entire process, from a single image to a complete, shareable video. That you can use for client projects or showcasing your own products.

Note: You will need a paid Freepik subscription for this tutorial.

🔬 Interested in learning about AI for business? Let us know what you want to learn & join The Lab waitlist.

▲ AI IMAGE SHOWCASE

Create this type of image with Midjourney

Midjourney Prompt: Set of minimalistic tea packages, paper boxes, contemporary design, real photo

Image Usage Suggestion: Product photography

▲ SHARE YOUR THOUGHTS

Have specific feedback or want to get in touch? Reply to this email and I’ll get back to you.

Know someone who’d love this newsletter? Forward it to a friend and have them sign up here.

Thanks for reading — until next time.

Stay curious,
Matt Lok, Editor

Brought to you by metalabs. Helping boost your brand’s visibility and attract more leads through branding & digital marketing.