• Superpower Daily
  • Posts
  • OpenAI launches Operator, an AI agent that performs tasks autonomously

OpenAI launches Operator, an AI agent that performs tasks autonomously

People are benchmarking AI by having it make balls bounce in rotating shapes

In today’s email:

  • 🔥 ChatGPT with Canvas now works with OpenAI o1—and can render HTML and React.

  • 📈 4 Charts That Show Why AI Progress Is Unlikely to Slow Down

  • 🧩 I got OpenAI o1 to play codenames and the results were surprisingly good.

  • 🧰 7 new AI-powered tools and resources. Make sure to check the online version for the full list of tools.

Top News

OpenAI has launched Operator, a research preview of its first general-purpose AI agent, designed to independently perform tasks like booking travel, making reservations, and shopping online. Announced on January 23, 2025, the tool is initially available to U.S. users on ChatGPT’s $200 Pro subscription plan, with plans to expand to Plus, Team, and Enterprise tiers in the future. Operator uses a dedicated web browser and relies on OpenAI’s Computer-Using Agent (CUA) model, which combines advanced reasoning and vision capabilities to interact with websites as a human would, without relying on developer-facing APIs. The CUA can navigate menus, fill forms, and click buttons, but always seeks user confirmation before finalizing actions.

Operator collaborates with companies like DoorDash, eBay, Instacart, and Uber to ensure compliance with terms of service agreements. It emphasizes user safety, requiring supervision for sensitive tasks like banking transactions and avoiding automated input of credit card details. However, Operator isn’t without its limitations. It struggles with complex web interfaces, CAPTCHAs, and certain sensitive tasks like sending emails or deleting calendar events. OpenAI has implemented safeguards, such as monitoring for suspicious activity and pausing execution when necessary, to mitigate risks of misuse like phishing scams or malicious prompts.

While competitors like Google and Anthropic have already ventured into AI agents, OpenAI’s cautious approach with Operator reflects the potential safety risks of such technology. Operator is part of OpenAI’s broader strategy to showcase how AI agents could transform internet use by moving beyond information processing to action-taking. Though Operator’s current capabilities are limited, this release marks OpenAI’s boldest step yet toward an agentic AI future.

TemPolor is an innovative AI-driven music platform designed to transform how creators access and utilize music for their projects. With a library boasting over 200,000 royalty-free tracks, TemPolor empowers users to effortlessly enhance their content for social media, podcasts, or video production.

  1. Search Made Simple: TemPolor's AI search feature (natural language + video search) is more efficient than traditional methods of finding songs.

  2. Optimized for Any Project: TemPolor creates and generates similar track features to maximize customization and meet users' musical needs.

  3. Royalty-Free Confidence: TemPolor's licensing is a one-time download, requiring no renewals and offering lifetime royalty-free usage.

The AI community has been fixated on a quirky benchmark over the past few days: coding a simulation of a bouncing ball inside a rotating shape. This challenge, which tests models' ability to manage collision detection and coordinate systems, has sparked comparisons among major AI systems. Some users report that Chinese AI lab DeepSeek’s R1 excels at this task, outperforming OpenAI’s $200-per-month o1 pro mode. Others observed Google’s Gemini 2.0 and OpenAI’s older GPT-4o performing flawlessly, while models like Anthropic’s Claude 3.5 Sonnet and Gemini 1.5 Pro stumbled, letting the ball escape the rotating shapes.

Simulating a bouncing ball in a rotating shape is a classic programming challenge requiring precise physics and robust algorithms. Researchers note that tracking multiple coordinate systems and designing resilient collision detection takes hours even for skilled programmers. Yet, slight variations in the prompt can lead to inconsistent AI performance, making it hard to draw definitive conclusions about which model performs best. As a result, benchmarks like this offer an entertaining but subjective way to evaluate reasoning models.

These informal benchmarks highlight a larger issue in AI evaluation: the lack of universally reliable measurement systems. Viral tests like this demonstrate the challenges of comparing models without standardized metrics. While initiatives like the ARC-AGI benchmark aim to establish more meaningful evaluations, the field continues to grapple with how to measure AI progress in ways that matter to most users. Until then, the internet will likely remain entertained by GIFs of bouncing balls and rotating shapes.

Sam Altman, CEO of OpenAI, surprised the tech world with the launch of Operator, an AI agent capable of autonomous web actions. Alongside this, Altman’s other venture, World (formerly Worldcoin), is now focusing on linking AI agents to users' digital identities. By using blockchain-based tools and biometric verification, World aims to prove the “humanness” of online personas and verify AI agents acting on behalf of real individuals. This approach addresses a future where distinguishing humans from AI will be increasingly challenging.

World's tools, which initially focused on human verification via biometric scans, are expanding to include digital “proof of personhood” for AI agents. Tiago Sada, World’s chief product officer, explained that these tools could allow AI agents to perform tasks on behalf of users while ensuring the interactions are verified and trusted. Businesses might embrace this concept, as it could boost user engagement and sales by allowing verified AI agents to utilize their services.

However, this raises questions about security and integrity. While platforms like Cloudflare block bots to protect against abuse, World’s technology could allow businesses to selectively permit trusted AI agents, reducing risks such as scams or DDOS attacks. OpenAI has already partnered with companies like Uber and DoorDash, signaling a shift toward AI-mediated user interactions. Verified AI agents, marked with a “blue check,” could make these interactions more seamless and trustworthy.

Altman’s ventures, from World to his nuclear fusion startup Helion Energy, seem interconnected in a vision where AI takes center stage. World’s evolving focus on AI verification aligns closely with OpenAI’s ambitions, potentially positioning it as a key player in shaping the future of AI-human interactions. The integration of World’s tools with OpenAI’s ecosystem could deepen, enabling a secure and scalable infrastructure for the AI-driven internet.

Other stuff

All your ChatGPT images in one place 🎉

You can now search for images, see their prompts, and download all images in one place.

Tools & LinkS
Editor's Pick ✨

Perplexity Assistant - Assistant for Android with smarter AI for tasks and context!

BookRead - AI-powered E-Reader that makes reading effortless

Jolt AI - AI assistant for 100k to multi-million line codebases

Telescope 2.0 - Find exactly who you’re looking for, fast

Soul Tarot - Explore your future, solve your questions

Trae - Adaptive AI IDE that helps you ship faster

Spell by Spline - A model to generate 3D worlds

Unclassified 🌀 

How did you like today’s newsletter?

Login or Subscribe to participate in polls.

Help share Superpower

⚡️ Be the Highlight of Someone's Day - Think a friend would enjoy this? Go ahead and forward it. They'll thank you for it!

Hope you enjoyed today's newsletter

Follow me on Twitter and Linkedin for more AI news and resources.

Did you know you can add Superpower Daily to your RSS feed https://rss.beehiiv.com/feeds/GcFiF2T4I5.xml

⚡️ Join over 200,000 people using the Superpower ChatGPT extension on Chrome and Firefox.

OR