Time: 3:05 PM
Speaker Bio: Executive Distinguished Engineer for Developer Experience at Capital One. Author of “Code Simplicity.” Previously Technical Lead for Code Health at Google, on DevEx at LinkedIn.
Speaker Profile: Full Speaker Profile
Company: Capital One uses OPA (Open Policy Agent) to bake compliance into deployment pipelines. Real-world enterprise DevEx challenges.
Focus: How to create DX that supports agents while maintaining compliance, security, and developer autonomy. 20+ years in the field.
Notes
- Since we are adopting the new shiny really fast now, different than we used to do, the future is really hard to predict
- Use tools in the same way that the outside world does development so that you can use agent tools
- Even more important to have
- Better tooling, better testing, better messages
- All that is old is new again, good software practices matter
- Designing coding with testing in mind — much more effective
- A lot of talk about documentation, the value of it is debated, but not the agents need written words
- If the code structure is good, maybe we don’t need docs on that
- It did not attend your verbal meeting that has no transcript
- There are orgs right now that use that tribal knowledge
- Telling the agent the context around it
- This sort of documentation is even more important
- We spend more time reading code than writing it, and even more so now
- Every software developer becomes a code reviewer
- We need to speed up code review velocity
- Make each individual response fast
- Figuring out how to assign work, whose turn it is to take action (similar to the McKinsey thing)
- Software quality: “not perfection, good enough, and better than it was before”
- The people the best at code reviews aren’t doing code reviews
- Standardize environment
- Dev environments
- CLIs and API needed at dev time
- Improve deterministic validation
- Refactor for testability and ability to reason about code base
- Write down external context and intentions
- What’s good for humans is good for AI applause
- When we invest in this we will help the humans no matter what
Slides
Slide: 2025-11-20-15-16
Key Point: The slide emphasizes the importance of maintaining high standards in code review processes, including encouraging quality reviews, maintaining high bars, and using reviews as a teaching tool for junior engineers.
Literal Content:
- Title: “Code Review Quality”
- Three bullet points:
- “You have to encourage great reviews”
- “That hold a high bar and reject code that doesn’t meet it”
- “Where junior engineers learn about coding and reviewing by getting reviews from seniors.”
Slide: 2025-11-20-15-17
Key Point: The slide warns about a negative feedback loop where AI agents trained on bad code produce low-quality reviews, which further degrade the codebase, creating a vicious cycle of declining code quality.
Literal Content:
- Dark background
- Title: “The Vicious Cycle”
- Flow diagram showing:
- “Bad Codebase” (code file icon) → “AI Agent” (robot icon) → “Many Low-Quality Reviews” (stamp icons) → “Worse Codebase” (code file icon)
- Arrow loops back from “Worse Codebase” to “AI Agent”
Slide: 2025-11-20-15-18
Key Point: The slide presents the positive alternative where good codebases and great tools enable AI agents to provide strong reviews, which leads to increased development velocity—a virtuous cycle.
Literal Content:
- Dark background
- Flow diagram showing:
- “Good Codebase” (code file icon) and “Great Tools” (wrench icon) → “AI Agent” (robot icon) → “Strong Reviews” (magnifying glass icon) → “Velocity” (fast-forward and lightning bolt icons)