Ugh I’d wish the author would not use AI in his writing. Some of the paragraphs are hard to read because of the AI style employed. It also does the annoying thing of making some grand statement early on and then never properly following up on that statement.
To quote an interview, “Why would I bother reading something no-one could be bothered to write?”
Caveat: Im going by the articles description of these papers, I haven’t read them myself yet.
Number of PR’s is a stupid fucking metric without corresponding comparative quality analysis throughout the full development cycle, preferably over many cycles.
By “quality analysis” I don’t mean “is this step faster?”.
It’s “lines of code” levels of stupid.
I will die on this hill.
The same with “more releases”.
LLM’s are not “reducing” technical debt, they are just repackaging/moving it.
Despite that, the “final thoughts” is surprisingly well rounded.
This is probably the most sensible and well rounded blog post Ive seen on AI development in the past 2 months.
“Least bad” isn’t necessarily “good”
