On Tuesday afternoon, Anthropic initiated a unique experiment by launching Claude Plays Pokémon on Twitch. This livestream features Anthropic’s newest AI model, Claude 3.7 Sonnet, as it plays a game of Pokémon Red. The stream has become a captivating demonstration of the capabilities of current AI technology and the public’s reactions to it.

Researchers in the field of AI have utilized various video games, including Street Fighter and Pictionary, to test new models, often for entertainment purposes rather than practical applications. However, Anthropic found Pokémon to be a valuable benchmark for Claude 3.7 Sonnet, which can effectively “think” through the types of puzzles present in the game.

Similar to OpenAI’s o3-mini and DeepSeek’s R1, Claude 3.7 Sonnet can “reason” its way through challenging tasks, such as playing a video game designed for children. While its predecessor, Claude 3.5 Sonnet, failed to exit the player’s home in Pallet Town, Claude 3.7 Sonnet successfully obtained three gym leader badges.

Despite its advancements, Claude 3.7 Sonnet still encounters difficulties. During the Twitch stream, the model was hindered by a rock wall, which it could not walk through despite its best efforts.

A Twitch user aptly summarized the situation, saying, “who would win, a computer AI with thousands of hours put into programming it, or 1 rock wall?”

Eventually, Claude discovered that it could navigate around the wall.

Watching Claude traverse Pokémon Red can be frustrating due to its slow pace, as the model meticulously thinks through each step. Nevertheless, it is also oddly captivating. The stream displays Claude’s “thought process” on the left side, while the right side shows real-time gameplay.

At one point, Claude attempted to locate Professor Oak inside his laboratory but became confused due to the presence of other NPCs in the scene.

“I notice a new character has appeared below me — a character with black hair and what appears to be a white coat at coordinates (2, 10),” Claude wrote. “This might be Professor Oak! Let me go down and talk to him.”

However, Claude proceeded to mistakenly talk to an NPC other than the Professor, an NPC it had spoken with several times before. Some viewers in the Twitch chat started to get impatient, while others who had been watching for a while were less concerned.

“Guys chill,” one person wrote in the chat. “Before we exited and entered Oak’s lab like 10 times before understanding how to move on.”

For experienced Twitch users, the format of Anthropic’s stream may evoke a sense of nostalgia. Over a decade ago, millions of people participated in a groundbreaking online social experiment called Twitch Plays Pokémon, where users collectively controlled the player character via Twitch chat, resulting in chaotic gameplay.

Some AI researchers have cited Twitch Plays Pokémon as an inspiration for their work. In October 2023, Seattle-based software engineer Peter Whidden published a YouTube video detailing how he trained a reinforcement learning algorithm to play Pokémon. The AI spent over 50,000 hours playing the game before it learned to successfully navigate it. One challenge was that the AI preferred to admire the pixelated scenery instead of actually playing the game.

AI-powered reenactments of Twitch Plays Pokémon, such as Whidden’s and Anthropic’s, are entertaining but also somewhat bittersweet. The original stream was a pivotal moment in Twitch history because it brought people together in an unexpected way, with everyone working as a team to progress through the game.

In 2025, it seems that we are no longer teammates but spectators, watching an AI model attempt to play a game that many of us mastered at a young age. This serves as a microcosm of a larger trend: our online experiences are shifting from shared, communal activities to more solitary ones.