ChatGPT can now complete tasks for you using its own computer.

Hey there, and welcome to AI Toast!

Today’s menu:

• ChatGPT Agent just landed — what’s new, what’s weird, and why it matters.

• Mira Murati’s Thinking Machines scores a $2B seed round (yes, billion with a B).

• Superhero tools to power up your workflow.

• Quick tutorial: How to minimise ChatGPT’s hallucinations

• Quick AI Toasts: Latest AI News

Estimated read: 6 minutes.

ChatGPT Agent can now do work for you

Imagine asking your AI to handle your inbox, research your competitors, and whip up a slide deck — all before you finish your latte. That’s the promise of ChatGPT Agent.

In a nutshell: ChatGPT Agent can now think, act, and juggle tasks online using its own virtual computer. It’s like hiring an intern who never sleeps, never eats your snacks, and won’t ask for a raise.

Key highlights:

• It’s proactive. ChatGPT Agent doesn’t just chat — it can browse the web, fill out forms, run code, and connect with your apps (like Gmail or GitHub). You ask, it acts.

• You’re the boss. The agent always checks with you before doing anything big, like logging in or making purchases. You can pause, stop, or take over at any time.

• Handles real work. From summarizing meetings to creating editable slides and spreadsheets, it’s built to finish what it starts.

• Safety first. OpenAI baked in guardrails: explicit permission for risky moves, privacy controls, and defenses against sneaky “prompt injection” attacks (think: hackers hiding instructions in web pages).

• Still learning. The slideshow feature is in beta, and sometimes outputs look like they were made by a PowerPoint rookie. But updates are rolling out fast.

Personal take: These AI agents will take care of boring tasks, help us find information faster, and make work and learning easier. They’ll be like smart helpers in our daily lives, letting us focus on what matters most. Some jobs will change, but new ones will appear. In the future, it’ll be hard to tell where human effort ends and AI help begins.

Receive Honest News Today

Join over 4 million Americans who start their day with 1440 – your daily digest for unbiased, fact-centric news. From politics to sports, we cover it all by analyzing over 100 sources. Our concise, 5-minute read lands in your inbox each morning at no cost. Experience news without the noise; let 1440 help you make up your own mind. Sign up now and invite your friends and family to be part of the informed.

Sign up today!

Mira Murati’s Thinking Machines: The $2 Billion Bet

Mira Murati, ex-OpenAI CTO, just raised $2 billion for her new AI startup, Thinking Machines Lab. The company’s only been around since February and hasn’t shipped a product yet. Investors? Still throwing money at it like it’s the next Google.

Here’s what’s wild:

• Valued at $12 billion. For a startup with no product or revenue. That’s like buying a house based on the architect’s sketches.

• Heavy hitters on board. Andreessen Horowitz led the round, with Nvidia, AMD, Cisco, and more piling in.

• Open source vibes. Murati promises the first product will have a significant open source component, aiming to help researchers and startups build their own models.

• Talent magnet. Nearly two-thirds of the founding team are ex-OpenAI folks. It’s like the Avengers, but for AI.

• Big vision. The goal? Safer, more reliable AI that’s useful for more people — and maybe a shot at dethroning the current giants.

It’s a reminder: In AI, sometimes the pitch is worth more than the prototype.

Learn AI in 5 minutes a day

What’s the secret to staying ahead of the curve in the world of AI? Information. Luckily, you can join 1,000,000+ early adopters reading The Rundown AI — the free newsletter that makes you smarter on AI with just a 5-minute read per day.

Sign up to start learning.

Superhero Tools for Your Workflow

• Replit: Build, test, and deploy apps right in your browser — no setup, no downloads. Real-time collaboration and AI help built in.

• Notion AI: Your all-in-one workspace just got smarter. Summarize docs, brainstorm, fix grammar, and search across your notes and files with a single AI assistant.

• Postwise: Write, schedule, and grow your Twitter and LinkedIn presence with AI-crafted, viral-ready posts. No more writer’s block.

• n8n.io: Automate anything. Connect your favorite apps and build custom workflows — like Zapier, but open source and way more flexible.

How to Reduce ChatGPT from Hallucinating

AI chatbots can't tell what's true or false. They just predict what words should come next based on patterns, not facts. Adding instructions to a prompt doesn't make them suddenly understand truth - it just changes what they're likely to say. Many users think special prompts guarantee accurate answers, but they don't.

This gap between expectations and reality frustrates users when AI gives wrong information. Also, when AI uses overly formal or robotic language, it can sound fake and turn readers off.

These AI systems don't have a built-in fact-checker. They can sound very confident even when completely wrong. The prompt below isn't a trick, it's a helpful structure that encourages the AI to admit when it doesn't know something.

Try copying and pasting this prompt to see how it works.

You can get prompts for Gemini, Claude and other AI Chatbots here.

Quick AI News Bites

• Elon Musk’s xAI is hiring engineers to make AI-powered anime girls, aiming for digital companions like Ani the goth waifu and Bad Rudy the red panda. The job is all about building AI that “understands the universe” — and, apparently, your heart.

• Lovable, a Swedish AI startup, hit unicorn status just eight months after launch by raising $200 million at a $1.8 billion valuation. Their tool lets anyone build websites and apps with simple language, and it’s already got over 2.3 million users.

• Perplexity, an AI search startup, is betting big on India to catch up with OpenAI. They’ve partnered with Airtel to give 360 million users free access, and their downloads in India are soaring, even as they work to turn all that attention into real revenue.

• Anthropic’s Claude Code users are frustrated after sudden, stricter usage limits were imposed without warning. Even those paying $200 a month are hitting new caps, and the lack of clear communication has left many users confused and annoyed.

Boost revenue and gain new customers by partnering with us

Reach over 35K AI enthusiasts with your product.

Join our newsletter to connect with tech professionals, investors, engineers, managers, and business owners worldwide. DM now!

That’s a Wrap

Thanks for reading! Got a tip, a story, or a favorite AI tool? Hit reply or drop me a note. Always love hearing from fellow explorers.

Stay curious — and don’t let your AI gaslight you.

Until next time, — Poonam Soni