Post

I've had a take since Sonnet 3.5, ~Aug 2024, that editor autocomplete has done more to mislead dev perception of AI capability than any other application of the tech. 3.5 could produce style-faithful, correct-enough, maintainable, niche code. Later models learned Unix, & agents optimized the loop.

26/6/26, 2:06 pm0 reposts2 likes

Ryan Atkinson

@ryanatkn.com

26 June

When Claude Code arrived I was quick to turn off editor autocomplete. I find it disruptive to my thinking to have keystrokes predicted, and wasteful - it's so expensive to do well that I bet most results still look confusingly poor compared to agents. www.youtube.com/watch?v=dFUl...

youtube.comIt's a Unix systemYouTube video by helgi27

Ryan Atkinson

@ryanatkn.com

26 June

I will also admit, right up until trying Claude Code, the hype around "2025 is the year of the agent" had me rolling my eyes, and I was wrong. I think skepticism is a vital perspective to maintain with these tools, but a closable loop has been visible for a long time. The speed of advancement tho..

Ryan Atkinson

@ryanatkn.com

11 July

*POSIX