Use a browser agent to play Wordle

A computer-using agent tried the NYT Wordle puzzle through a normal browser and solved only a small share of runs.

Done with Operator· Intermediate

What they did
Researchers sent OpenAI's Computer-Using Agent into the New York Times Wordle game and let it interact with the page like a person would. They tested it across a week of daily puzzles and watched closely to see where it stumbled.
How it went
The agent finished just 5.36% of roughly 200 attempts. Its biggest weakness was telling tile colors apart reliably, and its readings changed depending on the surrounding context.
Tips
This kind of task needs very careful visual checking, especially when the result depends on color. If you use a browser agent on a game or form, ask it to explain what it sees before it acts on it.

Try it yourself with Operator

Open [Wordle or another browser word game] in a browser and play today's puzzle. After each guess, tell me the color of every tile and explain what each color means before you choose the next move. If any tile color is unclear, pause and ask for my approval before making another guess. Stop when the puzzle is solved or the guesses are used up, then list the guesses and any colors that were uncertain.

Fill in the [brackets]. Agents can make mistakes, so review anything before it is sent, paid or deleted.

Read the original on arxiv.org ↗

Get 3 agent uses like this every morning

With the day's top AI news. Free, and one click to unsubscribe.