i am disconcerted by how many techies who saw through crypto *immediately* have fallen for AI krokodil
@davidgerard
With "AI", there is the perception that it sort-of works at the moment and they just need to fix a few bugs and everything will be perfect. Trying to challenge that argument is what I find most difficult.
@katrinatransfem my usual answer is "you've been saying 'next model bro' for three years"

@davidgerard @katrinatransfem there is something fishy about the most recent influencer campaign. They even got Knuth to mention a specific product by name.

It's like they did psychological profiling on individuals, and then give them exactly what they need to praise them, and not their competitor (and of course push the concept as a whole bit only secondarily).

@loke @davidgerard @katrinatransfem it's odd how it feels like it all happened at once. I don't buy into it being a grand conspiracy, but I also wish there was an explanation
@Mae @davidgerard @katrinatransfem not a conspiracy, but a well executed marketing campaign, I think.
@loke @davidgerard @katrinatransfem I think an undisclosed guerrilla marketing campaign of this size qualifies as a conspiracy
@Mae @davidgerard @katrinatransfem has there been any youtubers as part pf this campaign? Youtube requires sponsored videos to be marked as being an ad.

@loke @Mae @davidgerard @katrinatransfem My working hypothesis is that with the 4.6 model they made Claude better... at manipulating people

And we're seeing the results of that

@sitcom_nemesis @loke @Mae @katrinatransfem Writing a post/podcast on that is in my list for this week. Trouble is how to nail it down, rather than just my vibes on the topic.

given the tremendous number of social media posts from formerly competent developers and the paucity of non-shit results, i'm pretty sure this is the answer

but proving or even demonstrating it is rather more work and i need the material

@davidgerard @loke @Mae @katrinatransfem I also notice that a lot of devs use Claude for really basic shit, like variable renames and version bumps (see: the chardet guy: https://github.com/psf/requests/commit/b2a1d33f571518ca9a6148e7da787cc5827f897a). And I wonder what's the point of using a whole LLM to change 2 lines of easily findable code
Increase chardet upper limit to 8 · psf/requests@b2a1d33

Co-Authored-By: Claude Opus 4.6 <[email protected]>

GitHub

@sitcom_nemesis @davidgerard @loke @Mae @katrinatransfem

every time you use an LLM for *anything*, it makes your brain a tiny bit smoother

@pikesley @sitcom_nemesis @davidgerard @loke @Mae @katrinatransfem maralago brain: unnaturally plumped