OpinionSecure developmentMay 20265 min read
A sustainable case for hand-typed code
Why hand-typed code still earns its place in modern, AI-assisted engineering.
Open the report, 4 pagesDownload the PDFAI writes a lot of our code now, and that’s mostly a good thing. But the research keeps returning the same finding: code nobody typed is code nobody fully understands. Here is a short case for hand-typing the parts of a system that guard the door.
45%
of AI-generated code samples failed security tests across 100+ models.
1 in 5
package suggestions pointed to libraries that don’t exist.
More sure
AI-assisted developers wrote less secure code and trusted it more.
8×
more duplicated code blocks in 2024, as copy-paste overtook refactoring.
The premise
This isn’t an argument against AI. We use it every day for scaffolding, tests and the chores that used to eat an afternoon. Our point is narrower. Typing code by hand forces you to make every decision in it. That friction used to be unavoidable, so nobody counted it as a security practice. Now that it’s optional, we think it’s worth keeping on purpose for the code where one wrong decision becomes a breach. The data is cited; the conclusions are ours.
The detail
The evidence, and what we do about it
The full piece is in the PDF above. These are the parts worth pulling up on their own.