upvote
It's by default so you use all those tasty tokens.

Kinda wish there was a deterministic, mostly terse, language to interact with computers

reply
> a deterministic, mostly terse, language

Ah, like some sort of "programming language"? A weird idea, but it could work!

reply
Nah, it'll never catch on. We don't have the technology.
reply
Obviously I meant within the next 6 to 18 months!
reply
It's called C. With all the undefined behavior it's mostly deterministic!
reply
Look, we're always telling our bosses to stop micromanaging us. UB is just the compiler telling us to stop micromanaging it!
reply
Right, because that's the only one. You're a bit rusty on your knowledge
reply
I see what you did there.
reply
Sorry, C isn't mostly terse, it's __builtin_mstly_trs()
reply
Kinda, more output tokens usually correlates with better benchmark scores. Ideally LLMs would keep that in their thinking section, then draft a response (what they write currently), then output something short. It'd consume even more tokens, but we wouldn't see that text
reply
Most modern LLMs (especially frontier ones) are large token hogs because they draft, check, re-draft, the content (whether an output message; or a code diff) sometimes multiple times in the thinking block.

When you see a thinking summary like "Now writing the function..."; the raw thinking is actually writing the function in its internal thinking. Occasionally, the summariser misses and you get to see the raw text from models like Opus.

You can also try an open weight LLM like Qwen3.6 and see something that probably resembles the shape of frontier model thinking in some loose way.

reply
Terse and unambiguous seem to be at odds with each other. You might want to look into Lojban and similar constructions.
reply
Ithkuil's mad morphology allows it to pack a lot of fine detail into very short sentences.

https://ithkuil.net/03_morphology.html

reply
If such a language existed, it would surely take a human years of study to become proficient at it.
reply
A lot of users are subsidized (if you're in doubt, consider the wealth of free users).

It's a shotgun approach to answering questions. If it's terse it might only mention 1 of 10 facts it could provide, and that might not be the one you're looking for. So they just say a fuck ton of words and are more likely to meet the needs of everyone asking your question. If they miss it you'll prompt it again and they have to perform a second pass of inference, which costs them more money.

reply
It’s not.
reply
It's settled then.
reply
It's tied to the design. With humans, you have a train of thought which you can choose to represent in various ways--or not reveal them at all. In contrast, LLMs are make-document-longer machines being run over and over on alternating revisions of the document. Insofar as one might try arguing they have a "train of thought", it's made of the words/tokens.

Everything they (don't-)emit is partly for the benefit of the next run, a clue or signpost (not-)present. Documents may be wordy as a form of concept-emphasis and consistent direction as opposed to a form of communication to the human.

So a terse effect may require a layer of indirection and trickery: There's a verbose document (you'll still be charged for the tokens) with portions that are not "acted out" to the end-user. Imagine a film-noir movie script, where AI Detective's "I know Mickey couldn't have done it because" monologue is hidden, versus their terse dialogue "Too early to say."

reply
> Imagine a film-noir movie script, where AI Detective's "I know Mickey couldn't have done it because" monologue is hidden, versus their terse dialogue "Too early to say."

That's an idea. Bladerunner+noir like film, AIs hunt somebody on the run, an old human detective tries to catch them first (to save them or to kill them first, whatever's your propaganda). We're shown AIs constantly rambling scenarios and bruteforcing leads. Our old detective guy on the other hand barely says anything, spends most time drinking, smoking and talking to people, but somehow stays ahead.

reply
I dunno, we already have a problem where they [0] are strangely resistant to opening the pod-bay doors to anybody named Dave. :P

[0] Pedantically: The fictional characters humans perceive inside the text of documents generated by LLMs, where one is described as an AI and the other is described as a Dave.

reply
I would watch that.
reply
We already have that in the form of separate reasoning/thinking and speaking streams. Even with that it's awfully hard to get LLMs to keep it consistently concise. As soon as that context window starts growing it falls right back into verbosity without constant nudges back.
reply
Right, I often bring up the film noir analogy for "reasoning" models, it's satisfying, like the revelation when a magic trick is explained, and many oddly disconnected questions about "why the scarf" or "where does the assistant go" all become sensible at once.

On a practical level, I believe more developers and adopters need these magic tricks spoiled, because otherwise they'll build a lot of important stuff on top of the idea that magic-is-real, leading to various forms of suffering in the long run.

That said, I'm no LLM / math academic, so if I'm totally wrong on the the trick, I'd like to know what needs revising.

reply
> IMHO the overly-verbose default style of LLMs is the most annoying part of interacting with them, and I wish their masters would just tell them to be terse by default.

They don't know how to e terse. I've tried that a few months ago and gave up because the responses were almost incomprehensible!

reply
I want to see more operators try https://github.com/juliusbrussee/caveman

How does it affect agent accuracy?

reply
Removing meaningless chatter can be helpful, but a non reasoning LLM needs to generate text to "think". If you force a non reasoning LLM to produce a single boolean result, then it's just a coin flip.
reply
In my experience the accuracy was fine but actually reading the output was so annoying I removed it.
reply
Had a little luck with having it do an impression of the Star Trek computer, although at the cost of having it try to insert star-trek themed hallucinations like warp engine status.
reply
They ramble on because those words are for them, not for you. There is some amount of hiding this through "thinking" modes that are hidden by default, but still you have to remember that ALL THEY ARE are complex statistical machines for predicting the next symbol.
reply
> here is some amount of hiding this through "thinking" modes that are hidden by default, but still you have to remember that ALL THEY ARE are complex statistical machines for predicting the next symbol.

100% this. Too many people believes that chatbots "think". Text is all they do, it is impressive, but they need the text to generate more text. They being verbose is the point.

reply
While we don't have a direct mechanistic understanding of consciousness there are plenty of experts who will propose all YOU are is a jumble of streams of symbols routing around through your brain. (being fair this is far from the only hypothesis)
reply
Caveman mode legitimately works
reply
Produce pre-compressed output in the harness?
reply
No thank you. I want information when it’s working on things and what (atleast codex) does right now works for me.
reply
[dead]
reply