Posts tagged: ai

PHP2Shell sent our team into an around-the-clock effort to secure client websites. It also made something clear: AI is making it easier to find and exploit old bugs, and we need to get just as good at defending against them.

As coding agents take on more responsibility, AGENTS.md and CLAUDE.md are becoming part of how we manage software quality. The right instructions aren't necessarily the shortest. They are the ones that prevent mistakes, make checks clear, and match the authority we give our agents.

One experienced engineer with a group of AI agents may soon be able to write an entire application alone. But building a good software product still needs disagreement, experimentation, and perspectives beyond our own. If AI takes over coding, the human side of software may matter even more.

When code becomes cheap to produce, the hard part of software development shifts to deciding what should exist, proving it works, and deciding who gets to trust it. My predictions for teams, code review, open source, computing costs, and the software we use.

AI can write a feature in minutes. That does not mean it can own the result. The difference between vibe coding and agentic engineering is not which model you use, but how you make decisions, verify the work, and take responsibility for what ships.