MIT Technology Review · AIFeb 10, 2026, 01:02 UTC+8

Why the Moltbook frenzy was like Pokémon

MIT Technology Review compares the Moltbook AI-agent social experiment to 2014 Twitch Plays Pokémon: lots of spectacle, limited signal about the future. The post cites 1 million concurrent players in the Pokémon case; Moltbook also mixed in crypto scams, and some “agent” posts were actually steered by humans. The real gap is explicit: shared memory, coordination, and shared goals are still missing.

Read the original ↗Export Markdown

Take

Moltbook’s useful contribution is that it exposes how much agent hype still depends on theater. MIT Tech Review’s comparison to 2014 Twitch Plays Pokémon is apt: that experiment hit 1 million concurrent players, then mostly left behind a meme, not a durable coordination system. Moltbook has the same smell. The article says crypto scams were everywhere, and many “agent” posts were steered by humans, so the apparent emergence was contaminated at the input layer.

I don’t buy the “glimpse of the future” framing. Useful multi-agent systems need shared memory, coordination, and shared objectives; the piece names all three as missing. AutoGen and CrewAI at least force roles, tasks, and tool calls into a workflow. Moltbook sounds closer to bots performing in a public square. The spectacle is real; the engineering signal is thin.

Source: MIT Technology Review · AI · technologyreview.com
Published: Feb 10, 2026, 01:02 UTC+8 · Feb 10, 2026