Mumps

A coin flip can make infinite decisions on all possible input taxonomies.

Zero-shot is indeed very cool. And it's super useful/useful for the developer masses that don't know, care, or work pressures don't allow, for proper evaluation and calibration.

But it would be good that we don't over-hype these things.

roughly how much code are you outputting? I'm shocked to hear that API pricing could be lower than subscription. Do you mind sharing how much you're spending monthly on Luna API ?

This is off topic.

Do people not feel like LLM speak (Claudisms) is infecting their own diction? Saying 'a new "shape" of LLM' sits so very poorly.

Thank you

Are you on the foundation research team for Thomson? (If so, hiya from B!) Why would you expect Thomson to be particularly good at spam clf? I figured your additional corpus was all news and legal?

I get where you're going, but the argument isn't quite right.

A LLM never has and, in their current architecture, never can/could experience suffering or real change of state.

The general and expected case for humans is that capacity. A human my indeed lose capacity, e.g. being braindead and in severe cases we do indeed say that they are not conscious or able to suffer (different argument: some would of course say that is suffering in and of itself)

This is beautiful and I'm now itching to make a workbench (finally! that initial "how many sheets and what for what size" problem has been the mental blocker)

+1.0mm vote to adding metric please!

What's the problem with "eliminate deictic language" ?

It is clear, specific, and terse. It keeps both llm tokens down and human prose-reading to a minimum.

Thank you!

Christ in pijamas. TLAs should be a capitol offence. Even worse so, somehow, when undefined.

I absolutely LOVE Tailscale. but uhh. I think they shoulder exactly the same risk, right?

I like this. I thought to iterate on it a bit, for the folk who respond better to higher-tact phrasing:

"Thanks. I have access to ChatGPT as well. But I ask people for help when it fails. Your thoughts are smarter than GPT's, please provide those, next time."

Though, I'd like to be more succinct/terse.

I feel like you really need to mention BabyLM. For example you have:

> Directions we think are wide open ... Curriculum learning

BabyLM and offshoot published a pretty convincing body of work on exactly that (which suggests it's not particularly relevant to LM training).

As I read your page, I really felt like the brevity-thoroughness tradeoff went the wrong way.

Isn't the converse then equally appropriate?

Move to DST and if you want the ability to start your day later and end later, [...].

Yes, if OP did a full vocabulary comparison and took just those sub-threshold, it would be hacking. I'm not sure that's the case here, though? Given that (the post) OP started with em-dash, and probably didn't do repeated sampling, then it should be a pretty fair hypothesis that em-dash usage is a marker.

Your comment about p<0.05, feels out of place to me. The p-values here are << 0.05. Like waaaaay lower.

Perhaps Fisher's exact is more appropriate, on the per-word basis?

Thanks for sharing!

Like OP, I've been similarly struggling to get as much value from CC (grok et c) as "everyone" else seems to be.

I'm quite curious about the workflow around the spec you link. To me, it looks like quite an extensive amount of work/writing. Comparable or greater than the coding work, by amount, even. Basically trading writing code files for writing .md files. 150 chat sessions is also nothing to sneeze at.

Would you say that the spec work was significantly faster (pure time) than coding up the project would have been? Or perhaps a less taxing cognitive input?