AI & Emerging Tech

On Tuesday I hosted a roundtable with ~25 engineering leaders on 

“Redefining Good Software in the Age of Agentic Coding.”

For most of our careers, “good software” has been shaped by one key constraint: 

 Over the last six months with tools like Claude Code and Codex we’ve entered a world where producing large volumes of code is suddenly 

Here are the takeaways that most resonated with the group:

A practical way to assess AI-generated code quality: document what the model built, delete the codebase, and ask it to recreate the system from the documentation. Then measure time-to-rebuild, token usage, and what changes, breaks, or goes missing. (Hat tip to Ruslan Belkin.) 

 teams likely need different guardrails and support depending on their stage of adoption.

Before letting AI act autonomously, ask: 

what’s the cost of being wrong at this step?

Low cost (no users yet, easy rollback): move fast

High cost (payments, production): tighten controls

This lens applies across the pipeline, decision by decision.

We barely touched testing and validation. If more code becomes a black box, then evaluation becomes the core competency. We’re planning a follow-up roundtable focused on this. 

I’d love additional frameworks, metrics, or practices you’re seeing work.

Recap: Redefining Good Software in the Age of Agentic Coding Roundtable

Home

Forum

Events

Content

People

Contributors

Albums

Chapters

Chat

Channels

Sub-Communities

Tools

Help

ELC is a safe, supportive space for engineering leaders to connect, learn, and grow. We value authenticity, sharing, and thoughtfulness, with a mission to unite leaders and advance tech leadership.

Copyright

Terms of Service

Privacy Policy

Code of Conduct

Login to the ELC community to connect with others, attend community events, and more!

Login or Sign Up for even more

To unlock all parts of the community and get the best experience, complete your profile.