a retraction: earlier I was snarky about how so-called "AI" so-called "tools" have been out for 3 years and there's nothing to show for it—no groundbreaking movies or novels written by chatbots, no visible explosion of new software, no medical or math breakthroughts, no improvements to software stability or security with so-called "AI" so-called "assistance", et cetera.

turns out I was wrong; there's been an explosion in new software after all, this year, in mobile apps at least. Look at the iOS stats and tremble at the power of chatbots!

(Source: "Writing code vs. shipping code: Productivity effects across generations of [so-called] 'AI' coding tools" (Demirer et al., 2026).Graph: John Burn-Murdoch.)

@elilla more like explosive diarrhea
@elilla what technological progress! allowing mobile slop apps to be created with even less effort!

@elilla

There are in fact quite interesting things going on in mathematics with the more advanced models, but the LLMs are coming in on top of a lot of other very specialized things in that direction. This may survive the crash, and might also be manageable on local models.

(This has approximately zero relevance to the current financial bubble but it has its own particular interest for those in the field - which as noted did not suddenly arise with LLMs.)

@elilla more rats fighting over the same churro