Embed Notice

Btw so no one can claim I don't know what the fuck I'm talking about—I've been using last weekend to run a bunch of the latest local LLMs, even doing a cluster to see what they can and can't do. (Biggest model I've tested so far is Qwen 3.5 27B.)

It told me that the original Macintosh shipped with Apple II compatibility (it didn't).

Also that tomatoes are condiments.

And eggs are dairy.

Also tried it to convert some older single file open source JavaScript libraries I wrote to modern syntax, with mixed success (like it did 80% a good job but messed up a lot of it, too). With Ruby is was kind of a disaster.

Now—I get that these models are much smaller than the commercial ones that run in data centers and have much more training data; on the other hand at no point did the at-home models warn me that they can't do something or can't do it well.