The broken windows theory of coding agents

The broken windows theory of coding agents
9 minutes by Anton Zaides

AI coding agents let teams ship code much faster, but quality suffers when no one reviews what gets built. A single bad pattern spreads quickly as other agents copy it, turning small problems into serious bugs. Some teams now review only a small share of PRs and focus more on reviewing plans before any code is written.

The case for a shift-left approach: Saving time and money by understanding incidents and technical debt
sponsored by Jetbrains

Security issues become more expensive the later they’re discovered. Join the upcoming webinar hosted by JetBrains to learn how integrating security checks closer to code development can help teams reduce incidents and technical debt without slowing delivery. See how IDEs like IntelliJ IDEA help developers identify and remediate security issues as they code and bring speed and governance into their development workflow.

Same people
9 minutes by Mike Fisher

Mike shares the story of how GM's worst factory became one of its best after Toyota took over, without replacing the workforce. The same workers, many with poor track records, achieved near-top quality within a year simply because the system around them changed. Research across hospitals and software teams confirms the same pattern. The conditions people work in explain more of their performance than most managers want to admit.

Maximizing return-on-feedback
5 minutes by Aviv Ben-Yosef

Feedback is rare and easy to waste. When you get it, ask yourself whether it's true, whether it came too late, whether it was meant for you, and what it says even if it seems wrong. How you respond shapes whether people will speak up again. Handle it well and you make feedback safer for everyone around you.

Steve Yegge’s Google platforms rant
24 minutes by Steve Yegge

Amazon forced all teams to expose their data through service interfaces in the early 2000s, treating every internal system as a potential external product. This was painful but taught Amazon how to build real platforms. Google, despite doing many things better, never absorbed this lesson and kept building standalone products instead. Without a platform approach, products stay closed and limited, while competitors who open up to third party developers can scale far beyond what any single team could build.

Defensive career driving
9 minutes by Stay SaaSy

Getting recognized at work doesn't always happen on its own. You need to publicly claim credit for your work, correct anyone who misrepresents your contributions, and maintain enough leverage to negotiate for fair pay and titles. Watch out for colleagues who resent your success, as they may actively work against you. If you lead a team, build clear systems to track and reward performance so people never feel they have to fight for basic recognition.

And the most popular article from the last issue was:

• Measuring AI's impact on delivery speed