Back to the Code Why CTOs Must Get Technical Again to Solve AI’s Real Pro
summary
CircleCI applies agent decomposition to replace fragile monolithic prompts and ships an autonomous product called Chunk autonomous validation agent that automatically fixes flaky tests by coordinating LLM calls with deterministic orchestration; some prompts produce a correct code refactor in 2 minutes to a perfectly good PR, exposing the need to move checks from the outer loop into the inner loop. Experiments use Claude/Claude Code and Cursor, reveal MCP risks, and a chart shows managers using genAI ship 5–6x more PRs.
AI-written summary · about
chapters (15)
- 0:00Introduction to talk
- 0:22AI and expertise in leadership
- 3:11CTOs facing AI challenges
- 5:18Writing code and abstraction layers
- 8:31Learning AI tools quickly
- 10:14Inner loop acceleration with AI
- 11:57Products built with AI uncertainty
- 14:22MCP server pitfalls and strategy
- 16:30Agent vs workflow design
- 18:16AI productivity for managers
- 20:41Leadership credibility and empathy
- 21:24Q&A on AI proofs of concept
- 22:28Failure patterns in AI scaling
- 24:12Team expectations of CTO AI knowledge
- 24:54Balancing speed and learning
related talks
30mBeyond the Hype: Using AI Coding Agents Today
25mFrom AI Chaos to Production Ready: Using AI Agents to Improve Your Code Quality
01h 01m10 tips to level up your ai-assisted coding
32mEscape the Try-Again Loop: Verifying AI-Generated Code
22mBuilding Better Software: A New Blueprint for the Agentic SDLC
26mBeyond the code: Leading engineering teams in the GenAI era
26mI’m Not Writing Code Anymore—Now What? | The 7-Step Workflow
32mBeyond the Hype: A Pragmatic Look at How Developers Actually Use AI Tools