@caseyliss Just noticed that this works now. No idea when it was changed, but thanks! 😊

@daringfireball (also, fantastic article... very powerful!)

@daringfireball LLMs are actually naturally deterministic. An overly simplistic way to think about it would be that the LLM predicts the next word in an output over and over again. By default, it will always pick the most likely next word. Randomness is added in by increasing the "temperature", which gives the model a higher chance of selecting the non-most likely next word.

@caseyliss @gruber perhaps related… after putting it on, if my watch is unlocked by my iPhone rather than entering the passcode, then it won’t unlock my MacBook Pro.

@marcoarment saw them once (with TAB). First time hearing them and was very impressed, but also exhausted at the end of their set. They play like I always used to wish Phish played back in the 90s (frenetic jams in every song)

@tvaziri loved the episode. Started watching and thought it was you. Loved how you talked a bit about the culture of ILM. From stories I’ve heard, that’s something that many companies could learn from