Summary#
The NZZ article examines how tech giants like Meta, OpenAI, and Nvidia are now pouring billions into AI agents after already spending massive amounts on large language models. Meta acquired Moltbook (a “social network for AI bots”) and spent $2 billion on Manus. OpenAI acquired Openclaw for its autonomous AI agents. Critics see this as a “flight forward,” with companies investing even more money to solve the problem of poor returns on their LLM investments.
The article explores whether AI agents can deliver on their promise to automate knowledge work, or whether this is another investment wave that will never pay off.
My Quotes#
On Meta’s acquisition of Moltbook:
“On Moltbook, there are hundreds of thousands of fake profiles. Behind the supposed 1.5 million AI agents, there are only around 17,000 human owners. It is primarily a social experiment.”
On why Meta still sees value in Moltbook:
“Moltbook is simultaneously a registration platform for AI agents. When these agents act on behalf of a human, they also need a unique identity.”
On the emerging “Know Your Agent” requirement, similar to Know Your Customer in banking, and the race to build agent identity infrastructure.
On AI agent payments:
“This will be the next big topic.”
Companies from Mastercard to Visa, PayPal, and Coinbase are developing payment solutions for AI agents, including micropayments via stablecoins.
On why tech companies keep investing despite poor returns:
“The technology companies are in a kind of hamster wheel that they don’t dare to step out of. They don’t want to risk becoming the next Nokia.”