ChatGPT Just Got a Computer and It’s Ready to Work for You

FEATURED INSIGHT💡

ChatGPT’s Agent Just Got Real and It Has Its Own Virtual Computer

OpenAI’s newest release makes one thing clear: ChatGPT isn’t just here to chat anymore. It now acts.

The new ChatGPT agent can browse the web, write code, run it in a terminal, pull calendar events, summarize emails, plan dinner parties, and even generate full presentations, all autonomously on its own virtual machine. It's a true AI coworker that operates across browsers, APIs, and files, juggling tools fluidly to complete tasks from beginning to end.

Need a market analysis delivered as a slide deck? A financial model built from scratch? A dinner party planned and ingredients ordered? Just ask.

This rollout merges the best of previous tools like Operator (which could click and type like a human) and Deep Research (which excelled at analysis), layering in new capabilities like API connectors, browser takeovers, and terminal access, all with live narration and human override.

Still in early stages, the agent already outperforms humans in certain real-world knowledge tasks and even beats state-of-the-art models on high-stakes benchmarks like SpreadsheetBench and DSBench. It’s also trained with built-in safeguards, from prompt injection detection to secure browser sessions and hands-on oversight for sensitive tasks.

The future looks more capable, more automated, and more collaborative, with AI not just answering questions but getting real work done alongside you.

ON THE HORIZON 🌅

$2 Billion for… Something? Mira Murati’s Lab Raises Record-Breaking Round

Mira Murati, former OpenAI CTO, just closed a staggering $2B seed round for her new venture Thinking Machines Lab, valuing the stealth startup at $12 billion before even launching a product.

Backed by Andreessen Horowitz, Nvidia, AMD, Accel, and more, the lab has already recruited top AI talent from OpenAI, including John Schulman and Barret Zoph. Murati hinted their first offering will be open source and aimed at helping researchers and startups build custom AI models.

No product yet. No demo. But with a war chest like that and industry rumors swirling about Meta’s interest in acquiring them, it’s clear something big is brewing.

LATEST IMPORTANT NEWS 📰

Goldman Sachs’ Newest Hire Isn’t Human

Wall Street just got a full-stack AI developer. Goldman Sachs is now testing Devin, the autonomous coding agent from startup Cognition Labs. Devin can independently build apps, refactor code, and complete multi-step software tasks, and is slated to join Goldman’s 12,000-person developer team.

Tech chief Marco Argenti says this “new hire” could triple productivity and may soon be scaled to thousands of Devins. It’s a major signal that AI agents are moving from experiment to enterprise reality, and yes, they’ll be supervised (for now).

Europe’s Fastest AI Unicorn? It’s Lovable

Swedish startup Lovable just raised a $200M Series A, hitting a $1.8B valuation only eight months after launch. The platform lets anyone build web and mobile apps via natural language, with 2.3 million users already and $75 million in ARR.

Its secret is simplicity. Non-technical founders use Lovable to spin up prototypes and apps quickly without a dev team. That’s resonating, especially with enterprise clients like Klarna and HubSpot.

Meta Rejects EU AI Rules

Meta is refusing to sign the EU’s new AI code of practice, calling it “overreach” and “legally uncertain.” The voluntary code is designed to prepare companies for the AI Act’s looming enforcement, requiring transparency, documentation, and dataset compliance.

Meta’s pushback signals growing tension between U.S. tech firms and EU regulation, especially for general-purpose AI models. Enforcement begins August 2.

FOR THE TECHNICALLY INCLINED 🛠️

How to Post-Train a Language Model (Free Course)

Want to turn a base model into something useful? DeepLearning.ai’s latest short course gives a clean, actionable overview of post-training methods like:

• Supervised Fine-Tuning (SFT)

• Direct Preference Optimization (DPO)

• Online Reinforcement Learning

You’ll get hands-on with pre-trained models, learn when to use each method, and even build pipelines to train a model that can, say, solve math problems or adopt a specific assistant persona.

Perfect for engineers, researchers, or AI startups trying to customize models without starting from scratch.

AI TOOL OF THE DAY 🚀

Need to build a mobile app fast, but don’t code? AppStruct is an AI-powered no-code platform that lets you design, build, and deploy professional apps for any device, with zero dev experience needed.

Use natural language to generate layouts, connect APIs, and launch full products. Great for founders, marketers, or product teams testing new ideas without burning engineering time.

AI Won’t Replace You… But Someone Using It Will.

Master prompts. Get smarter results. Work faster. Grab The AI Prompt Playbook — your guide to unlocking the full power of generative AI tools like ChatGPT, Gemini, Claude, Midjourney, and more.

Now on sale (50% off!) for a limited time only. Don't get left behind.