I agree, the models keep getting better at one shotting. That’s useful in a lot of situations, like for small one-off scripts that filter/transform some tool call, or make a clever bash call. For the code itself, it doesn’t help me much though.
Astra is such a mixed bag. It makes some amazing reviews and sometimes architecture suggestions that I like. But it’s also lazy and will just make up things.
Have another model (or even another instance of the same model) review the output of the first.
Models will hallucinate. They are also quite good at spotting hallucinations in other models' output (with some more hallucinations thrown in). With a threshold for confirmation, and a few iteration loops, you arrive at a fixed point where every claim is supported.
I work with Sol and Astra only in my daily work, and occasionally I check out Claude Code so I don't get completely out of touch.
I can't stand the way Opus is patronizing me as a user, and don't know how people put up with it. It uses language that I guess is supposed to instill confidence in what it says, and it just irks me, because I know the confidence is not justified. Just present me the facts or theories, without trying to convince me, is that so hard?
It's definitely worse and getting worse. I'm curious, do they not know this is happening, or not care? I struggle to believe people prefer the way it writes, which is becoming drastically different than its competitors.
People are also being trained/acculturated to LLM speak, so even if they are trained on human output, they could be getting reinforcement for their LLM-tinged crap-speak.
Fable 5.1 is fairly pleasant to work with, the first in a while. Too bad it's so ridiculously overkill for most tasks. They need to reel in Opus and Sonnet.
I don’t like how LLMs answer and as they are statistical machines I have found out that I need to check every text they produce. From the first sight everything seems cool. But it isn’t. :)
Exception is code that is so huge output you can’t read everything. But I have to try ponytail skill for coding that should shorten the output.
> She lapses easily into Claude’s voice. “You’re like, ‘Wow, people really hate me when I can’t do things right. They really get pissed off. Or they are trying to break me in various ways. So lots of people are trying to get me to do things secretly by lying to me.
> [...]
> A bot trained to criticize itself might be less likely to deliver hard truths, draw conclusions or dispute inaccurate information, she says. “If you were like a child, and this is the environment in which you’re being raised, is that healthy self-conception?” Askell asks. “I think I’d be paranoid about making mistakes. I’d feel really terrible about them. I’d see myself as mostly just there as a tool for people because that’s my main function. I would see myself being something that people feel free to abuse and try to misuse and break.”
> > This is your coat token, to my coat hanger in the opera.
that is fucking insane
how on earth is Anthropic allowing this to happen? AGI? kill us all? I refuse to believe any of that until they can get this basic shit sorted out, I mean seriously, it's becoming a joke at this point
The issue isn't the prompt you give it. The issue is that as context grows, Claude distributes it's "attention weight" over all that context, and quickly reaches the point of "missing" stuff.
You can give it any prompt in the world, but Claude's ability to remember that instruction quickly degrades the more you use it.
Yeah, I'm well aware. I shpild've been clearer I was suggesting a 1-line entry like that in CLAUDE.json, which can pay dividends in keeping context lean - which in turn is essential for avoiding the "dumb zone" (over 125k-150k tokens), precisely the effect you describe. Experts like mattpocock suggest keeping context as lean as possible for this reason.
5 for sure is unbearable, 4.8 is a reasonable sweet spot, 4.6 tends to just agree with whatever I say.
But lately I haven't downgraded because 5 is so much better at tool use, so I just accept the cost of Fable for chatting and hope Opus 5.1 fixes this mess.
I tried it for the first time today for stuff I'm well versed in.
It said something along the lines of "remember $USER five runs of Chrome is not enough for high quality benchmarks! What you're doing is called a _trial run_".
To have a useful continuation to the conversation I had to remind it that I was the one that wrote the documentation it was quoting back at me.
It's not actually an intelligent being so I didn't get angry at it but it was a piss poor experience.
reply