- Author
- Posted by mpweiher, with replies from many commenters
- Published
- 3 August 2026
- Kind
- Hacker News discussion thread
- Original
- news.ycombinator.com/item?id=49153374
- Subject
- Review, taste and expertise
- Themes
- Skill under automation, Getting good
What it is
The user mpweiher posted the link on Hacker News. The linked post is by Ankur Sethi. It asks coding assistants to show edits in chat, and the author types them in. Commenters debate whether that prevents loss of understanding.
What it says
- Sethi's post says retyping builds a mental model and helps catch hallucinations.
- The post says the author is probably only 2x faster, not 10x, for deeper understanding.
- Commenters such as trentor say retyping magazine listings taught them a lot.
- Commenters such as WJW say thinking, not typing, drives the learning.
- Commenters such as jader201 say employers value productivity over cognitive health.
- Commenters such as simondotau split code by value. Core code is handwritten; AI writes tests.
- Some commenters liken LLMs to compilers. Anamon replies that compilers are deterministic.
What evidence it gives
The thread is anecdote and argument from many commenters. Few cite studies, though estebarb links an arXiv paper on learners and cognitive debt. The linked post reports the author's own months of use, with no measurement.
Interests and limits
Hacker News commenters are self-selected and often work in software. Replies mix hobby, work and teaching contexts. Claims about future model ability are predictions. The cited paper was not checked.
Connections
These links are this wiki's own reading, not claims the sources make.
- Too Fast to Think: The Hidden Fatigue of AI Vibe Coding. Schmidt says AI output outruns the forming of a coherent picture. Retyping slows input to the speed of hands. A commenter objects that thinking drives learning, not typing. So which part helps, the typing or just the delay?