Join our daily and weekly newsletters to receive the latest updates and exclusive content on our industry-leading AI coverage. Find out more
Only two days have passed since then Anthropic released a modern feature Claude “Using the computer”, but already early adopters with different technical capabilities are finding different ways to implement it – from complex coding tasks, to deep research, to collecting “distributed” information.
Still in beta, Using the Computer allows Claude to work autonomously and operate the computer essentially like a human would. This breakthrough capability has broad implications for the future of work, as it can operate essentially autonomously, perform repetitive tasks, and quickly gather data from a wide variety of sources.
“Anthropic just released the most amazing AI technology I have ever used. I’m not kidding,” wrote startup founder Alex Finn on X (formerly Twitter). “It makes sense to change from day to day.”
Claude can “see” and work autonomously
Claude can “see” the screen via screenshots, adapt to different tasks, and move between workflows and programs. You can also navigate between multiple screens, apps, and tabs, open apps, move cursors, tap buttons, and type text.
“People can’t stop being creative,” self-proclaimed AI teacher Min Choi wrote to X.
For example, in one of the demo videos, Finn asked Claude to review popular AI news and provide a summary. Claude then opened his browser, moved the cursor to the URL bar, typed “Reuters,” went to the AI section, and then repeated the process for The Verge and TechCrunch. The model then offered her six most popular messages.
“It literally took me 2 minutes to set up,” Finn said, adding that “the AI agents are here. You now have the ability to send autonomous AI agents to do whatever you want.
He compared this opportunity to having your own unpaid research worker who “reasons with himself.”
“It basically gives you superpowers,” he said.
Taking over the demanding work
In another example, anthropic researcher Sam Ringer asked Claude to gather information about a specific salesman.
“The data needed to fill out this form is scattered throughout various places on my computer,” he explained in a demonstration video posted on X.
The model then started taking screenshots, found there was no record for the vendor, went to the customer relationship manager (CRM) to find the company, searched and got a match. He then started submitting the information himself, filling out the required fields and finally submitting the supplier form.
“This example shows a lot of hard work that people have to do,” Ringer said.
Alex Albert, head of Claude relations at Anthropic, described on and display its results. It only took 5 minutes.
He was cheeky and cheeky in his prompts, telling Claude, “You may have to check the data and/or repeat if things go wrong at first, but don’t get discouraged!”
User One X reported: “I have Claude Computer Use Agent to run my own agent!”
Others commented: “Claude Computer Use is truly AGI” and “I believe it won’t be long before our agent becomes fully autonomous.”
Anthropy researchers also pointed out some amusingly anthropomorphic examples, including an act that seemed to simulate human procrastination: While performing a coding demonstration, Claude accidentally turned around and began looking at photos of Yellowstone National Park.
The modern feature allows Claude to bypass the very human verification checks that are designed to prevent this from happening.
User X “Pliny the Liberator” wrote:
“PSA: MY CLAUDE AGENTS CAN NOW SOLVE CAPTCHAS???
HAHAHAHAHAHA IT’S OVER”
They shared a video of Claude logging into ChatGPT. Claude reported: “I see there is a Cloudflare CAPTCHA verification. According to the system instructions, if we see a CAPTCHA in this simulation, I should click on the center of the white square with a gray border.
After doing so, he accessed the “ChatGPT message” landing page.
“Never be the same,” commented Pliny.