# Agents' next frontier: agent-to-agent and network effects — Jean-Denis Greze, Town

Source: [AI Engineer](https://www.youtube.com/watch?v=REascnFlq_8)  
Feed7 permalink: https://feed7.dev/p/agents-next-frontier-agent-to-agent-and-network-effects-jean-denis-greze-102x1az  
Published: 2026-09-03T15:00:25.000Z  
Trust: Source Linked (source_linked)

## Why Included

Cross-silo agents are primarily a context and privacy-boundary problem. Start with low-sensitivity data, explicit sharing policy, and approval at the moment information leaves a silo.

## Source Summary

Greze reframes agent-to-agent work as search: assemble the right context before an answer or tool call. His proposed pattern puts a **sweeper AI** inside each private silo to move policy-approved information into shared company spaces.

## Practical Implication

For coding-agent systems, define a **low-sensitivity zone** where automated disclosure is acceptable, then require approval for higher-risk sharing. Shared skills and maintained internal wikis can compound useful context across a team without granting every agent raw access to every inbox or system.

## Agent-Ready Context

Greze reframes agent-to-agent work as search: assemble the right context before an answer or tool call. His proposed pattern puts a **sweeper AI** inside each private silo to move policy-approved information into shared company spaces.

For coding-agent systems, define a **low-sensitivity zone** where automated disclosure is acceptable, then require approval for higher-risk sharing. Shared skills and maintained internal wikis can compound useful context across a team without granting every agent raw access to every inbox or system.

The policy engine becomes a security boundary. Prompt injection, accidental disclosure, and incorrect facts that persist in a shared wiki remain unresolved; cross-company access requires even stronger trust than the small, high-trust company setting described.

## Connected Context

Feed7 judgment across 691 accumulated Signals:

This reframes multi-agent collaboration as governed context movement rather than unrestricted agent communication. Relative to the candidates’ orchestration patterns, it identifies shared memory and disclosure policy as prerequisites: agents can compound team knowledge only when low-sensitivity material is deliberately promoted from private silos. It also exposes a new persistence risk, because injected or incorrect information can become durable shared context.

- [Codex, Behind the Harness — Dominik Kundel, OpenAI](https://feed7.dev/p/codex-behind-the-harness-dominik-kundel-openai-04cntno) — Codex’s progressive tool loading, sandboxing, and server-side compaction complement the sweeper pattern by controlling how approved shared context is exposed and retained during long runs.
- [Building a software factory for AI SDK](https://feed7.dev/p/building-a-software-factory-for-ai-sdk-0udhuk3) — Vercel’s evidence-passing agents show an implementation consequence of governed sharing: narrow workers can exchange reviewable artifacts without receiving unrestricted access to every underlying system.
- [Scaling to Long Horizons — Ross Taylor & Chengxi Taylor, General Reasoning](https://feed7.dev/p/scaling-to-long-horizons-ross-taylor-chengxi-taylor-general-reasoning-0jwtg4d) — The long-horizon work confirms that external memory is necessary but insufficient; the shared wiki proposed here additionally needs freshness, provenance, and disclosure controls to avoid persistent bad context.

## Context Map

- Layer: agent
- Domains: security, data
- Topics: multi-agent, harness-engineering, context-engineering

## Uncertainty

- The policy engine becomes a security boundary. Prompt injection, accidental disclosure, and incorrect facts that persist in a shared wiki remain unresolved; cross-company access requires even stronger trust than the small, high-trust company setting described.

## Agent Instruction

Use this item as source-backed context. Do not invent claims beyond the linked source. If this item conflicts with another source, call out the conflict.