AI Intelligence Trend Snapshot - 2026-10-28
Trend #1: Chinese labs advance open-weight agentic coding models
- Status: strengthening
- Confidence: Medium
- Run review: Qwen Code changed at the runtime layer rather than model weights. No new cross-organization evidence changes the assessment.
Trend #2: MCP enters enterprise security and identity governance
- Status: emerging
- Confidence: Medium
- Run review: GitHub workload identity and Gemini CLI private-network MCP OAuth add identity-boundary signals, but cross-product deployment evidence remains limited.
Trend #3: Coding agents converge on multi-agent runtimes
- Status: established
- Confidence: High
- Run review: Codex, Claude Code, Gemini CLI, Qwen Code and OpenHands all strengthened session recovery, background work and runtime lifecycle handling, reinforcing the established assessment.
Trend #4: Frontier-lab safety disclosure moves toward third-party review
- Status: strengthening
- Confidence: Medium
- Run review: No new independent safety review or institutionalized disclosure evidence appeared.
Trend #5: Enterprise self-hosted execution planes emerge as agent infrastructure
- Status: candidate
- Confidence: Low
- Run review: Run:ai memory enforcement improves quota control, but one vendor update is not enough to raise the trend.
Trend #6: AI-generated frontier mathematics receives Lean verification
- Status: emerging
- Confidence: Medium
- Run review: No new formal-verification or independent mathematics evidence appeared.
Trend #7: Agent execution capabilities separate from policy controls
- Status: strengthening
- Confidence: Medium
- Run review: Copilot workload identity, Gemini private MCP OAuth, Qwen approval enforcement and Run:ai quota controls strengthen the signal, but independent production evidence is still missing.
This cycle did not produce enough evidence for a new trend.