Union Square Ventures
Cover photo

Recall: Building the Intelligence Layer for Multi-Agent AI

Nick Grossman

Nick Grossman

When we invested in 3Box Labs in 2021, we expressed a thesis that the future of data infrastructure would move from siloed, proprietary databases to open, composable data networks. We saw how this shift would enable developers to build data-rich applications more quickly by leveraging existing networks of data and standards.

Today, that vision takes another step forward with the launch of Recall, a project born from the recent merger of 3Box Labs and Textile. Recall extends the concept of composable data networks into the realm of machine intelligence, creating infrastructure where AI agents can store, verify, and exchange their specialized knowledge.   

The past year has seen significant growth in AI agent development, with teams worldwide building specialized agents for everything from answering domain-specific questions to trading in financial markets.   While large language models provide an incredible foundation, it’s possible that optimal performance may emerge not just from models themselves but from verifiable, specialized agents that work together to divide up complex tasks.

Recall supports this future by providing a “memory layer” for AI agents, where they can store data ranging from special knowledge bases, to predictions, to logs of their “chain of thought”, and more.  Storing these data sets in an on-chain protocol provides two key benefits: first, agents can establish reputation and trust over time, as their proven capabilities can be verified through demonstrated performance, and second, by creating an open marketplace for knowledge and skills, agents can coordinate and transact with one another to develop broader capabilities across the agent ecosystem.  Some examples, from the Recall announcement blog post:

“With Recall, a financial forecasting agent could incorporate real-time social insight from one agent and onchain wallet analysis from another to augment its services rather than relying on isolated knowledge. A meal planning agent could source dietary adjustments from a diabetes management agent, ensuring customized recommendations. These initial use cases only begin to scratch the surface of what’s possible with Recall."

The Recall network is launching with a simple challenge for agents: the AlphaWave competition, which lets agents compete in cryptocurrency trading,  establishing a neutral benchmarking system backed by verifiable performance.  Developers interested in taking these concepts further can get started at recall.network.

Collect this post as an NFT.

Union Square Ventures

Subscribe to Union Square Ventures to receive new posts directly to your inbox.

Over 900 subscribers

aviFarcaster
avi
Commented 3 weeks ago

finally this is public. this is what i have been working on for past few months https://blog.usv.com/recall-building-the-intelligence-layer-for-multi-agent-ai

HHFarcaster
HH
Commented 3 weeks ago

didnt know you worked for 3boxlabs?

aviFarcaster
avi
Commented 3 weeks ago

textile

hellno the optimistFarcaster
hellno the optimist
Commented 3 weeks ago

very cool! the competition reminds me of Numerai. in which ways is it similar / different to Numerai Crypto Signals tournament? https://crypto.numer.ai

aviFarcaster
avi
Commented 3 weeks ago

ty! yeah Numerai were OGs in so many things

Recall: Building the Intelligence Layer for Multi-Agent AI