I'd love to know what the hell Antrhopic has done to make Claude's writing so, so bad.

Opus/Fable output these days though is... not enjoyable. It's just really bad. The code quality is fine, but i want information from claude and it's just awful to read.

My biggest problem honestly is that i can't move my day job.. we're using enterprise claude and i'm not sure how much effort it would be to get access to another provider. I should inquire though, claude is really frustrating these days.

Semi related, but i would hate to read that PDF but i also hate reading what LLMs write lol.

LLMs are pretty terrible at being concise. Using an LLM these days means putting up with bizarre and often confusing phrasing, wordy explanations, etc. It's kinda crazy to me how good they are but how bad their writing style is for me personally. Even though i use an LLM constantly i can't stand reading its responses.

Where i've felt lacking for the last ~5 years though is output. Specifically blocked by having the energy to create all these damn ideas.

I find LLMs neat, albeit their own type of exhausting, as they open up a lot of possibilities for things i want. Ie software i've wanted but have never cared to find the time for or didn't have the time for.

LLMs feel like a software equivalent to a 3D Printer. I don't have to carve it out of wood now. But also like a 3D printer there are some tasks that LLMs are just horrible for. So it's a niche, a skill even, to find software you want that also are a good for for unsupervised development. It's also of course massively more risky than a 3D printed doohickey depending on its internet access/etc.

With that said i still don't really take pride in something vibe coded. I just take enjoyment out of using a thing that i've wanted but never had time for.

It seems a huge "the sky is falling" to think all LLM use is bad for fear of some "addiction" to me. Even many skeptics (eg Hashimoto) have come around to the idea of using LLMs.

One thing is clear in my mind. VCs are burning cash, and so if you have an LLM flow you find useful, take some free cash. The sky is not falling in this respect imo.

Which is not a defense of AI, to be clear. AI may very well be a big societal problem, but in this context i don't see Hashimoto/etc becoming heroin addicts like ya'll are so concerned about. It's just a fancy autocomplete. The fear seems a bit over hyped.

My expectation is that this cash handout is going to stop soon, so take while the giving is good right?

I should note that for my workflow /loop is eating the credits, so it’s basically no skin off my back. I just queue up some more work and let it go.

Generally i felt fine. I'll keep it in mind, thanks

Aside from the headaches what addictive effects are you referencing?

It's a wonderful project. Even small amounts help. Thanks!

It is surprisingly offensive.

The only friction for me is the general expensiveness of trying out top tier models, eg OpenAI's Fable equivalent (Sol?) to run for a trial period. I'd like to see a like-like comparison, eg buy x20 on OpenAI and see how much Sol i can use, how well it works, etc.

edit: Though surprisingly Claude seems to think Codex doesn't have hooks? That'll be tough, i use Claude hooks quite a bit.

Faster iteration means i mentally checkout less and am more involved with the code being created.

My hope is that in the far far future, we can get LLMs so fast that i can work in my IDE like normal and the LLM will just be an extension of autocomplete. I can state a goal, rough out functions, code, etc, and it'll just work around me like a very fast pair programmer / autocomplete.

The chat interface is an intermediate step that frankly i hate. The faster it is the less i wait.

Now for vibe-slop i'm making on the side, yea i don't care about speed. But that's not something i'm employed to do or anything i truly care about. It's a different workflow entirely.

I’m not offended by that because I’m already maximally offended lol.

My point was that i feel like humans have these two aspects, the ability to have a fact, and the ability to have a signal to the facts strength. I propose that as an explanation why we can internally analyze our understanding and come to a conclusion that yes, we do "know" it.

We also at times can't figure out how we "know" it, either because we've made up a detail (filling in blanks, assumptions, etc), or because we forgot where we embedded this detail. The lower the signal strength of this validation the more we can be unsure about a "fact".

I feel like LLMs are one half, but not the other. They effectively need a RAG for all of the knowledge they have, and if the counter for a given fact/idea/etc is low enough, then it's an low signal.

The question i have is how to do this efficiently. Of course i'm just speculating too, i don't know any of this.

Now is that because 4.8 is nerfed compared to 4.6 and thus more token efficient? No idea. I just know on x20 with a pretty plain workflow i struggle to use my tokens every week.

A worse product could win right now if it simply does as its asked.

It’s feeling quite similar to why I distributed out from google all those years ago. I didn’t want a hugely important centralized google account to be banned and cause friction to various aspects of my life.

The ability for LLMs to more easily catalogue user behavior and intent is going to get more interesting. Weird days. Feels like anyone could become a Facebook level metadata hoarder.

In that silly example/thought, the LLM would effectively need the ability to query the strength of a fact/spatial region/etc.

Right now i believe the LLM is more just the output of those weights. It has no way to inspect the strength of the signal. Eg it doesn't know if blue in "The sky is " is a strong or weak signal, it just predicted that next token.

If we could somehow encode strengths along with every token, eg "The sky is " or something we could perhaps give it a sense of [un]certainty. Though i imagine it would look differently than that since we'd want to encode this information in some sort of multi-dimensional space, rather than purely by token - since tokens aren't that valuable. Eg the "knowledge" in an LLM goes beyond tokens, and so too should signal strength.

I'm of course speculating on all of this and i have no clue on anything.

It obviously breaks down with humans too, given we so easily hallucinate and confuse things we "know". However i still suspect we're more reliable at probing information we've experienced vs not. Even if the case of poisoned knowledge, eg a crime scene accidentally implying information to a witness that the witness doesn't actually know, we still "know" that poisoned information via incorrect inference. Ie we "experienced" it.

Wonder what architecture would allow for this style of information/weight probing for an LLM.