The missing layer between agent connectivity and true collaboration
Source: VentureBeat
“How do you enable agents to truly think together, with all the contextual understanding, negotiation, and shared purpose that entails? It’s a critical next step toward a new kind of distributed intelligence that keeps humans firmly in the loop.”
VentureBeat AI Impact Series – Highlights
Guests:
- Vijoy Pandey – SVP & GM, Outshift (Cisco)
- Noah Goodman – Stanford Professor & Co‑founder, Humans&
They discussed moving beyond agents that merely connect to agents that are steeped in collective intelligence.
The Need for Collective Intelligence, Not Just Coordinated Actions
“Agents today can connect together, but they can’t really think together.” – Pandey
- Protocols such as MCP and A2A solve basic connectivity.
- AGNTCY handles discovery, identity management, inter‑agent communication, and observability.
These solutions are comparable to making a phone call between two people who don’t speak the same language.
Pandey’s team argues that the deeper challenge is enabling agents to achieve collective intelligence, not merely coordinated actions.
Shared Intent, Shared Knowledge, and Collective Innovation
Historical Perspective
- Individual human intelligence emerged ~300,000 years ago.
- Collective intelligence appeared ~70,000 years ago with sophisticated language.
This breakthrough gave rise to three critical capabilities:
- Shared intent – a common goal.
- Shared knowledge – a mutable body of information.
- Collective innovation – building on shared knowledge to create new outcomes.
“Once you have a shared intent, a shared goal, you have a body of knowledge that you can modify, evolve, build upon, you can then go towards collective innovation.” – Pandey
Goodman adds that language does far more than encode/decode information:
“Language is this kind of encoding that requires understanding the context, the intention of the speaker, the world, how that affects what people will say in order to figure out what people mean.” – Goodman
Human collaboration and cumulative cultural evolution rely on this sophisticated, context‑aware understanding—something still missing from agent‑to‑agent interaction.
Addressing the Gaps: The Internet of Cognition
“We have to mimic human evolution.” – Pandey
Three‑Layer Architecture
| Layer | Purpose |
|---|---|
| Protocol Layer | Goes beyond basic connectivity to handle intent sharing, coordination, negotiation, and discovery across heterogeneous agents. |
| Fabric Layer | Provides a shared memory system where agents can build and evolve collective context, giving rise to emergent properties. |
| Cognition Engine Layer | Supplies accelerators and guardrails that let agents think faster while respecting compliance, security, and cost constraints. |
“Think about shared memory in a heterogeneous way. We have agents from different parties coming together. So how do you evolve that memory and have emergent properties?” – Pandey
New Foundation‑Training Protocols to Advance Agent Connection
Humans& is re‑thinking how foundation models are trained:
- Training focus: interactions between a human and multiple agents, and between an agent and multiple humans.
- Goal: Center training on extremely long‑horizon interactions so models learn how to achieve the right long‑term outcomes.
“Our goal is not longer and longer autonomy. It’s better and better collaboration.” – Goodman
Humans& aims to build agents with deep social understanding—entities that know who knows what, can foster collaboration, and connect the right specialists at the right time.
Establishing Guardrails That Support Cognition
Guardrails are essential for deploying multi‑functional agents across an organization, but they must avoid stifling innovation.
- Traditional approach: Strict, rule‑like guardrails.
- Human approach: Operate on a principle of minimal harm, making contextual judgments about consequences.
“How do we provide the guardrails in a way which is rule‑like, but also supports the outcome‑based cognition when the models get smart enough for that?” – Goodman
Pandey emphasizes that interpretation is a collaborative task, not something solved by static predicates or documents. It requires common understanding, grounding, discovery, and negotiation.
Distributed Intelligence: The Path to Superintelligence
True superintelligence will not arise from ever‑larger individual models, but from distributed systems that embody collective intelligence.
“While we build better and better models, and better and better agents…” – Pandey (truncated)
The conversation underscores that building the infrastructure for shared intent, shared knowledge, and coordinated cognition is the next frontier for AI—one that keeps humans firmly in the loop while unlocking the power of distributed, collaborative intelligence.
> "Eventually we feel that true super intelligence will happen through distributed systems," Pandey said.
Intelligence will scale along two axes: **vertical** (individual agents) and **horizontal** (collaborative networks), in a manner very similar to traditional distributed computing.
> "We can't move towards a future where the AIs go off and work by themselves. We have to move towards a future where there's an integrated ecosystem, a distributed ecosystem that seamlessly merges humans and AI together," said Goodman.