By early 2025, the AI agent category had passed its initial hype peak and entered the harder phase: proving real utility. The easy money had been made on GOAT, ai16z, and Virtuals launches. What remained was the actual work of building agents that people would use daily — not just tokens that people would trade. The category needed to answer a fundamental question: are AI agents in crypto a product category or just a new flavor of memecoin?
The evidence pointed both ways. On the product side, AIXBT’s market analysis was genuinely useful. Trading bots powered by AI frameworks were increasingly competitive. Agent-managed DAOs were running small treasuries. Creative agents like Zerebro were producing output that attracted real audiences. There was a core of actual utility being built, even if it was small relative to the speculative volume.
On the memecoin side, most AI agent tokens had no functioning agents, no utility, and no plan beyond riding the narrative. The category had the same 90%+ failure rate as memecoins in general, and many projects that launched with AI branding were essentially pump-and-dumps with robot logos. The speculative layer was much larger than the utility layer, which is normal for early crypto categories but makes it hard for outside observers to take the space seriously.
What comes next will probably be a split. A small number of AI agent projects will build genuine products — autonomous traders, research agents, creative agents, governance agents — and their tokens will accrue value based on actual usage. A much larger number will fade as the narrative cools and attention moves to the next category. The ELIZA framework and similar infrastructure will survive regardless, because the developer tooling is genuinely useful even if the tokens built on top of it are speculative. The long-term winners in AI x Crypto will be the ones still shipping when the hype is gone.
Leave a Reply