Latent Space: The AI Engineer Podcast · Latent.Space

Dreamer: the Personal Agent OS — David Singleton

·1 hr 4 min·4 clips

He built a conference app in 25 minutes from LLM-readable conference data.

Okay, Dreamer is pitched as the access layer, but the interesting part is how practical the pitch gets. David Singleton describes a product where people can discover, build, use, and share AI agents and agentic apps, with consumers treated as users and possible builders. The sister example keeps it from floating away. It makes the sharper question obvious: what happens when smart, non-technical people finally get software for the problems sitting right in front of them? The conversation keeps coming back to builders. Dreamer wants people to make tools, publish them on the platform, and get paid when others use them. The Lego analogy works because it gives the marketplace a shape you can picture. Singleton says Dreamer's Builder in Residence program came straight from Lego's master builder idea. Then the host gets to payments. Since Singleton was CTO of Stripe, the episode naturally checks whether agentic commerce needs stablecoins, micropayments, or some new payment layer. His answer is grounded. Dreamer's first commerce move is helping creators sell useful agentic experiences, and Stripe Connect already handles that job. The flywheel is plain enough: better tools attract more builders, more builders make more agents, and the platform gets more useful as the loops start feeding themselves. There is also the prize hook. The best tool added by mid-April gets $10,000, more seed spark than spectacle. By the end, the conversation turns philosophical again. Machines are getting closer to what people want, but that final five percent still has teeth. For Dreamer, that gap is the whole bet: people may build around imperfect intelligence if the product loop is useful enough.

Listen to the show on

Apple Podcasts Spotify

Play clips from this episode More from Latent Space: The AI Engineer Podcast