comment 0

Professional Development – 2026 – Week 41

AI engineering

New Skills v1.3 Brings /pr, /implement-spec, and /retro

  • /implement-spec — allows you to take specifications, break them down into parts, and have subagents implement them
  • /pr — create a PR for completed work
  • CONTEXT.md is now GLOSSARY.md
  • /retro — looks at previous agent sessions to find ways to improve things

Culture

There’s No Such Thing as an AI-Ready Culture

  • “What organizations need instead is a culture of adaptability: the capacity to flex as strategy, technology, and work evolve.” The key attributes include speed, agility, and action orientation.
  • Move from activity (did we follow the steps) to impact (did we get the desired result).
  • Make experimentation the cultural norm. Start small, learn quickly, and use evidence to decide what to do next.
  • Reduce decision bottlenecks by building decision systems that help people closer to the work make the decision.
  • Connect work across boundaries rather than optimizing only what’s in your control (e.g., function, budget, process).
  • Foster the ability to learn, unlearn, and relearn.

Software engineering

Why Good Codebases Go Bad (and How to Stop It)

  • From Michael Feathers… Legacy code is code without tests.
  • Without tests, every change is a gamble. The cost of change goes up over time.
  • Systems that are easy to change are not legacy systems regardless of age.
  • If your current “protection” is that you understand how the system works, what happens when your team gets reorganized in 6 months? See Peter Naur’s 1995 paper about programming as theory building.
  • What makes legacy systems?
    • Loss of feedback.
    • Loss of understanding.
    • Loss of the supporting ecosystem (old language, old dependencies, lack of skills to use old tech).
  • See Manny Lehman’s statement about software evolution (entropy).
  • Write tests as you write the code to get built-in feedback.
  • Protect your understanding by using executable specifications (acceptance tests).
  • Protect the ecosystem by having an automated deployment pipeline.
  • Only change legacy systems if there is generally a need; rewrites are expensive.
  • Make changes tactically using the strangler fig pattern (approval/characterization tests) so you can refactor in small steps with confidence.
  • Legacy = you’ve lost the ability to make changes quickly and safely

7 Authentication Concepts Every Developer Should Know

This 20-minute video provides a visual overview of basic, digest, API key, sessions, bearer tokens, access/refresh tokens, OAuth2, OpenID Connect, SSO, and identity protocols.

Leave a Reply

Your email address will not be published. Required fields are marked *


This site uses Akismet to reduce spam. Learn how your comment data is processed.