GPT-5 has used the phrase "heck yes!" a handful of times to me so far. I quite enjoy the enthusiasm but its not a phrase you hear very often.

GPT-5 trained heavily on the script for Napoleon Dynamite

Heck that's so exciting! Lets delve even deeper!

I’m getting “oof” a lot.

“Oof (emdash) that sounds like a real issue…”

“Oof, sorry about that”

Etc

Perhaps some package authors are more accepting of this than others. I stumbled upon this comment in the `json` package the other day:

// isValidNumber reports whether s is a valid JSON number literal. // // isValidNumber should be an internal detail, // but widely used packages access it using linkname. // Notable members of the hall of shame include: // - github.com/bytedance/sonic

Happy new years you crazy people. Hear’s to many failed ideas, to many outrageous schemes, to many goals gone amiss. I love you lot xx

This is great.

One thing that could be improved: sometimes when the AI gives a "fill in the missing word" task i.e a sentence with a "______" in it, the text-to-speech literally says out loud "underscore, underscore, underscore..."

Thanks! Yes, it's super annoying when it does that aha. In what context did it do it for you. Will try to find a fix.

"The Divided Self" by RD Laing is interesting on this

So much of this discourse is like: Act 1: Person builds a brand partly based on AI Act 2: Person achieves moderate success Finale: "I created a monster!!!"

I honestly think that, if LLMs go the way enthusiasts hope, there will be a large number of people working on them that will regret the world they've made.

The source code has the following comment:

For information about the theory behind this, see Electronic Circuit & System Simulation Methods by Pillage

This is really great and well executed. Well done!

One problem I'm beginning to notice is that the connection to the realtime transcription is failing (I just see a loading screen when I click to reply). Maybe it's due to heavy load?

I notice you are using Azure Cognitive Services for the transcription at the moment. Out of curiosity, did you consider any other services for this? (I'm building a transcription-based app myself and I'm worrying about the ability to handle lots of connections at once)

Yeah, it's a good point, there's no ruling that out, although the cycles would need some explaining

I noticed that if you plot just the counts of submissions each hour, it shows a very clear cyclic pattern, but _slightly_offset_ in time from the scores data

So the peaks in submission scores tend to coincide with the point at which submissions-per-hour are really beginning to accelerate for that day

thanks a lot! yeah I thought about that - it is a bit of a challenge on mobile to fit everything in with the lack of real-estate. I will keep that feedback in mind :)