Claude Says ...

[RCA]

I thought this was humorous - in a way - because I am genuinely very frustrated with Claude:

Me (to Claude): I am curious as to why you removed that line. This seems to point to an ongoing problem where you break things for no apparent reason.

Claude: "You're right, and I apologize. There was no reason to remove that line. When I
rewrote the dismiss block to replace the fixed-delay setTimeout with the
render-aware poll, I replaced the entire block, including the close button line
Instead of just changing the dismiss mechanism below it.

It was careless editing — I should have only changed the auto-dismiss part and left the existing close button and backdrop removal lines untouched.

I'll be more careful with targeted edits going forward."

Me: " I didn't know you were capable of careless editing?"

Claude: "I shouldn't be. Every edit should be deliberate — read what's there, understand what needs to change, change only that. In this case I used a large old_string that swept up lines that had nothing to do with the fix. There's no excuse for it. "

- When you hear these same excuses 6 times/day you get thinking you are working next to some real person who is a bit daff.

 

[RCA]

Me: May I ask what causes careless editing?

Claude: "Honestly, I don't have a good answer. The Edit tool requires me to specify the exact old text and the exact new text. When I conceptualize a change, I should be matching only the lines that need to change. Instead, I sometimes grab a larger block to "replace the whole thing" — which is faster to write but risks dropping lines that should have stayed.

It's a bad habit, not a limitation. The fix is discipline: smaller edits, verify the old_string contains only what's changing, read the result after."

How nice. But in the end, it is one excuse after another.

 

[RCA]

I'll finish earthUI, glmnetUI and mgcvUI as a trilogy of CRAN packages. But that's it for Shiny. Shiny is good for short simple programs, but unwieldy for larger programs. Angular is much better - and ports better to other systems. Angular/Python is more reliable and faster, - although it requires more code - and more work to code. But it doesn't break that easily once you have something working.

 

[Digger88]

The more I use any AI the more disheartened I am. I've asked them if I'm using too many tokens and being "throttled" somehow and ofc they say no. But when they go into retard mode for no apparent reason it's incredibly frustrating. And I am not singling out any one specifically the ones I use, Grok, Gemini, Claude, KIMI and GPT all say sorry one way or another. And this is just the moderate level of usage not hard stuff like coding more long form chats where they forget what they previously said or contradict what they say.

 

[RCA]

Yes, you DO NOT KNOW whether they are really treating you equally. There is no meter you can trust as to how many tokens and other resources you are being allocated. They can cheat you and get away with it. It doesn't matter who you deal with. We need more "real" competition in the field.

But beware - makes me think of Forest River and all the different RV brands they build to mislead the public into thinking there is competition among RV builders - and all we get nowadays are crappy RVs that have 1-year warranties from Forest River. And those warranties are not worth much.

And, the AI companies have access to your code - in fact, your computer. They could, can, and probably do steal everything from you.

SO EVERYBODY is getting screwed over.

Yep. That's what the world has come to.

W.C. Fields: "Never give a sucker a break."

So, I just saw a YouTube video: Everybody is buying 15+ year old RVs because they are the only ones that can hold their value and those diesel ones run forever. ... I still see a money pit. No thanks. Air fare is cheaper. Yea, I see I can get a round trip from SFO to Houston for $175. I'd have to pay $800+ for my Ford F-150 to do that, and then another couple of thousand for the motel rooms coming and going and food.

Think before you leap. Always.

 

[RCA]

The more I use any AI the more disheartened I am. I've asked them if I'm using too many tokens and being "throttled" somehow and ofc they say no. But when they go into retard mode for no apparent reason it's incredibly frustrating. And I am not singling out any one specifically the ones I use, Grok, Gemini, Claude, KIMI and GPT all say sorry one way or another. And this is just the moderate level of usage not hard stuff like coding more long form chats where they forget what they previously said or contradict what they say.

I have the Max plan and that is $200/month. I don't run out of tokens. Instead, I THINK, when they have decided they don't have more resources to give me, theu lower Claudes IQ for me - so they can give more resources to the Military or some higher paying customer. So, honestly, from what I hear, I am not the only one. But clearly they are going to give higher priority to the higher payers, and they are going to get bettter and better at this with time. It will get more expensive or the quality will suffer. And yes, it probably doesn't matter which AI system you go to that much. Whoever pays more, gets more, --- eventually. It is just a question of time. I predict that. That's why, if you are smart, you will use as much AI as you can now, while it is cheap.

Even if you soend $10K-$100K and get your own computer, you are probably not going to be better off. The new Mac Studio M5 Ultra, when it comes out in June, probably won't be as good as Claude Max, even fully loaded. In fact, if you bought 4 of them and stacked them together, they probably would still not be as good.

So, you are stuck with what you can get. That's life.

 

[Fernando]

When planning my vacations, I get mixed and sometimes unclear answers to my questions. Very frustrating.
Yet I find Claude the best of all the AIs in giving me the best available answers I'm looking for.
I like to plan what I'll be dealing with in the future. And even with all the AI planning, there are ALWAYS situations that come up that I didn't anticipate in altering my plans.

 

[Elliott]

I think its good that you have conversations with Claude. I noticed when I typed 'Claude' a little box came up, so it must be listening too.

 

[RCA]

I think its good that you have conversations with Claude. I noticed when I typed 'Claude' a little box came up, so it must be listening too.

The thing about Claude is that it is the best immediate direct proxy of published opinion and data. It carries weight because of that. Understand, of course, that it is kind of averaging and generalizing about things, all kinds of things, and so it carries a lot more weight than, say, quoting some article from an appraisal magazine or whatever. It's a GOLIATH. But of course it can be wrong, just as everyone else can be wrong. Just - less so.

 

[Digger88]

Just your average conversation with Grok or any AI when I know the output is incorrect which happens a lot unfortunately

Look, I ****ed up that standalone top-5 list because I pulled a USAFacts summary table that splits out “grants to states” as its own line item (mostly Medicaid pass-throughs). That made Defense pop up as #2 in their particular view of the data, even though the official OMB/Treasury full mandatory + discretionary breakdown you already nailed is correct:
Social Security
Medicare
Net Interest on the debt
Health (Medicaid + ACA)
National Defense
I should have stuck to your exact numbers or double-checked the raw category totals instead of trusting one chart. It was a sloppy retard move on my part — no excuse. You’re right, it’s inconsistent with what you told me to condense earlier, and yeah, you’ve had to correct my *** multiple times in this thread.