Task: Research
Introduction to ChatGPT agent
OpenAI's July 17, 2025 livestream introducing ChatGPT agent mode. In the demo, the agent plans for a friend's wedding: finding an outfit that fits the dress code, hotels, and a gift, using its own browser.
Introducing ChatGPT Atlas
OpenAI's October 21, 2025 video introducing ChatGPT Atlas, a web browser with ChatGPT built in, including a preview of agent mode.
Claude computer use plans a sunrise hike
Anthropic's October 2024 demo of computer use. Claude searches for a sunrise spot, checks the distance in Maps, looks up sunrise time, and adds a calendar invite.
Claude for Chrome handles three browser workflows
Anthropic's video showing Claude for Chrome pulling dashboard data into one analysis doc, addressing slide comments, and testing code in Chrome.
Project Mariner shops for paints in Chrome
Google DeepMind's demo of Project Mariner, a Chrome extension prototype. The agent finds a famous painter, a colorful painting on Google Arts and Culture, then adds a paint set to an Etsy cart and asks before checkout.
Introducing Comet
Perplexity's launch video for Comet, its AI browser, released with the July 2025 launch. In the clip, the assistant is asked to pull up a specific YouTube video.
Comet for Android
Perplexity's video for Comet on Android. The assistant adds a scarf under $30 to a cart, pulls a takeaway from an article, and quizzes the user on a textbook.
Opera Neon agentic browser intro
Opera's May 2025 video introducing Neon, a browser with built-in agents that can take actions for you. It shows an AI Replay feature that lets you trace and replay each step the agent took, and openly says the agent can still misread or drift off task.
Manus Browser Operator intro
Manus introduces Browser Operator, a browser extension that lets the Manus agent work inside your own browser, using sites you're already logged into, after you grant it access.
Manus general AI agent launch
Manus AI's launch video for its general-purpose agent, showing it take a request and work through a multi-step task on its own computer.
Gemini in Chrome auto browse preview
Google team members talk through Gemini in Chrome and preview auto browse, where Gemini browses for you to handle tedious web chores, and discuss how they handle things like CAPTCHAs and safety checks.
Amazon Nova Act finds apartments
Amazon's intro to Nova Act, a toolkit for building browser agents. The demo searches for two-bedroom apartments in Redwood City, pulls the listings into a table, and checks biking distance to the nearest Caltrain station for each one.
Devin benchmarks AI providers and builds a site
Cognition's March 2024 intro to Devin. Asked to benchmark Llama across several API providers, it makes a plan, reads API docs in its browser, hits an error, adds a debug print to fix it, then builds and deploys a website to show the results.
Genspark Super Agent plans a trip and calls restaurants
Genspark demos its Super Agent planning a trip, then using a Call For Me tool where the AI phones a restaurant in a human-sounding voice to book a table, mentioning a shellfish allergy and a vegetarian guest. It also makes short videos from a recipe.
Claude for Chrome hands-on: saves a file, struggles with a form
An independent YouTuber tests Claude for Chrome. It refuses to type a Social Security number, leaves the final save click to the person, and on a long tax form it gets stuck in a loop and wrongly says it finished until the tester points it out.
Coming soon: the $1 list
A short weekly email with the best real agent runs. It will cost $1 when it launches. For now, just join the waitlist. No payment is taken here, and we only collect your email.
See our privacy page. Unsubscribe anytime through our contact form.