本文由 AI 分析生成
建立時間： 2026-04-13 來源： https://x.com/intuitiveml/status/2043545596699750791
Summary
The CTO of CREAO (an agent platform, 25 employees) details a full organizational redesign around AI agents — from architecture to CI/CD to team roles. The central insight is that AI-assisted (adding AI tools to existing workflows) and AI-first (rebuilding processes assuming AI is the primary builder) produce multiplicatively different outcomes. They went from a monthly release cadence to 3-8 deployments per day in two weeks.
CREAO 的 CTO 詳述了圍繞 AI agents 的完整組織重新設計——從架構到 CI/CD 到團隊角色。核心洞察是 AI 輔助（在現有流程中加入 AI 工具）與 AI 優先（以 AI 為主要建構者重設流程）在結果上有乘法級差異。他們在兩週內從月度發布節奏提升至每天 3-8 次部署。
Key Points
- Bottlenecks shift when build time collapses: PM planning (weeks) and QA (days) become the constraints when agents implement features in 2 hours
- Monorepo conversion is a harness engineering necessity — fragmented codebases are opaque to agents
- Self-healing loop: CloudWatch + Claude triage engine auto-creates Linear tickets with severity scores, sample logs, and suggested investigation paths; resolves and auto-closes on verification
- Three parallel AI review passes on every PR (code quality, security, dependency scan) — review gates, not suggestions
- Junior engineers adapted faster than senior engineers; accumulated coding skill is devaluing faster than critical evaluation skill
- Two future engineer roles: Architect (1-2 people, designs the harness/SOPs) and Operator (everyone else, executes within the system)
Insights
The most structurally significant observation is the collapse of management overhead: the CTO dropped from 60% time managing people to under 10%, enabling a return to direct building. This is a concrete data point for how AI-first organizations concentrate architect-level judgment at the top and route all other work through automated systems. The risk is organizational fragility — if the one or two architects leave, the harness loses its designer.
Connections
Raw Excerpt
“AI-first means you redesign your process, your architecture, and your organization around the assumption that AI is the primary builder. You stop asking ‘how can AI help our engineers?’ and start asking ‘how do we restructure everything so AI does the building, and engineers provide direction and judgment?’”