I’ve a friend working at a Big Tech company and after three years of heavy use, he estimates perhaps a 15% max increase in productivity BUT he feels it’s maxed out now. Said his mangers still believe the hype that more is possible even without evidence.

True. Certainly formulaic websites have been pretty much phased out by LLMs. It’ll be interesting to see what impact agents will have on hotel booking sites as they improve.

Good points and I’m sure many programmers agree with you. One of my questions is how do (large) software companies support this way of working or are LLMs simply viewed as a means to produce more (of “less” quality)?

Using tailwind as an example of „slow coding“ is quite a stretch.
Disclaimer: I wouldn’t shed a single tear if tailwind finally died.

But I think that the situation points to an older problem but similar problem.

gehrmann-design.de/blog/2026-01...

TW had found a model that allowed open source to work for them up until now. TW being “too big to fail” as you said opens up a whole other area of powerful companies relationships to sustainable open source projects - a long standing problem which I think AI is making worse. (2/2)

Hi Michael, Great post and I agree with what you said. My idea of slow is engagement with any “educational” resources outside of the LLM window. Tailwind was used as an example of less engagement with those resources. (1/2)

This reminds me of a great book folks interested in the evolution of tech in prisons should check out

Linking in @molly.wiki who also moved from SS to Ghost maybe a few months earlier. I’m sure she’ll have an insightful take.

… have come in to help fund TW. I also saw Adam had a podcast on the background to all this, so I’ll listen to that later and then follow up in more detail 👍. adams-morning-walk.transistor.fm/episodes/we-... (2/2)

Thanks for this. Some good points here re broader impact of genAI on content dev (videos, docs etc.). I’m not really on X but saw that a number of big sponsors … (1/2)