CMU study found AI coding tools yield only a 3-month productivity spike before quality debt erases gains
“there was, in fact, a temporary spike in productivity... but it lasted about 3 months and then it went back down”
code-quality
9 tracked signals on code-quality.
AI agents are driving a 2450% explosion in code volume that humans cannot review fast enough
“You have an explosive growth of 2450% in the number of lines of code written.”
Replace repeated agent prompting with codified rules via tests, typing, and linters
“you should stop prompting, you should start codifying the actual rules that the agent should follow”
The AI bottleneck is codified judgment around models, not model capability itself
“Replace the model with a smarter one. You will get a slightly better answer. Now take away the tests, inspections, reviews, and it all falls apart. So the model was never really a bottleneck. The bottleneck is the judgment around the model.”
Cognition launches FrontierCode, a benchmark grading code quality and maintainability over passing-test 'slop'.
“Many SWE-bench-Passing PRs Would Not Be Merged into Main”
AI agents are creating commit volumes that break traditional CI pipelines, demanding new merge-queue architectures
“At Uber, we had AI agent-scale commit volumes before the agents themselves existed.”
Boundary ships a compiler without code reviews using AI-readable architecture docs as the invariant standard
“Code can be slop, writing cannot.”
Visual Studio remains the premier tool for professional C#/C++ developers in the agentic AI era.
“it is the tool for professional C# and C++ developers on Windows.”
Three identifier-naming rules keep code readable: short local scopes, no redundant context, no type prefixes.
“Your code shouldn't just compile, it should communicate.”