How developers use AI
The tools developers now code with, how far they run on their own, what the automated reviewer catches, and how AI-shaped work reads in a codebase.
Lessons
- How do developers use AI to write software? The one guide to AI-assisted engineering: the tools, how much they do on their own, what automated review catches, and what an AI-heavy workflow leaves behind in the code.
- What are Cursor, Copilot, and Claude Code? The AI coding tool landscape: autocomplete in the editor, chat in the editor, and agents in the terminal. All of them tools a developer adopts in a week.
- What is a coding agent, and how much does it do on its own? The autonomy ladder: autocomplete that suggests a line, a supervised agent that works while a developer watches, and a background agent that takes a ticket and returns a finished change.
- What do developers mean by context, skills, and MCP servers? Why an agent is only as good as what it can see and touch: context as everything the model is shown, skills as packaged instructions, and MCP servers as the sockets to real systems.
- Can AI review code, and should it be the only reviewer? What automated review reliably catches, what it misses, and why teams that automate the first pass keep a person on architecture, intent, and product judgment.
- Does AI-written code look different, and can you tell? What heavy AI assistance leaves behind in a codebase, why AI-assisted says nothing by itself, and the honest state of the argument about whether the tools erode skill.
- Will AI replace software developers? Which tasks compressed, why junior-shaped work compressed first, what a developer still owns, and why lab claims and shipping teams report different things.