Fargo's latest season is quite good. Dark and funny, plus John Hamm makes a great villain.

I’ve always wondered if it’s some sort of a McGGUFfin

Timing, eh?

I know, I kept thinking it was a UI bug or something until I've googled it. Dunno what they were thinking.

I mean…Fortnite is old school? Is that a mod for Team Fortress?

Which is a good reason to discuss returning the current limits to something healthier. Life of the author plus 25 years is good with me.

You can read the research over here: https://arxiv.org/abs/2307.09009https://github.com/lchen001/LLMDrift Fin/

GitHub - lchen001/LLMDrift

Contribute to lchen001/LLMDrift development by creating an account on GitHub.

4️⃣ Conclusion: If you're depending on closed models released as APIs you'll have to test your prompts for data drift each time a new version comes out, forever and ever. This sounds like a good argument for picking open, instead of closed models. 8/

Which sounds like it would be caused by a new system prompt optimized for chatbots rather than raw output. So, presumably, if we change the system prompt the models will start outputting raw code again? Who knows. 7/

For code generation, I don't think this is as big an issue as the news makes it seem, especially after reading the research -- they mention that one of the leading causes of code not being compile-able is that in the June versions, it's surrounded by code fences. That's right, code fences. 6/