A.I. Stuff

[cashcatz]

Yeah that's a hard no on Claude good freaking Lord

Claude is incredible, easily the best AI model. I think you ment the OpenClaw malware?

 

[Fernando]

Claude is incredible, easily the best AI model. I think you ment the OpenClaw malware?

You got me interested in checking out Claude.
I'm always looking for best AI model.
Copilot and ChatGPT failed in some of my requests.
I'll see if Claude is the best. I just want accurate up to date information.

 

[Fernando]

Asking questions in planning my coming vacation, I liked how Claude gave map and options how to get there. It's better than ChatGPT and CoPilot so far.
And none of those the snide remarks CoPilot would give me.

 

[cashcatz]

You got me interested in checking out Claude.
I'm always looking for best AI model.
Copilot and ChatGPT failed in some of my requests.
I'll see if Claude is the best. I just want accurate up to date information.

Here's $20 of Claude for free. Be sure to download their desktop app, that's where the full experience is.

https://claude.AI/referral/XPAB0POYqA

Be sure to manually type the AI part to lowercase, AF messes up .AI and capitalizes it when sharing for some reason

 

[Digger88]

Claude is incredible, easily the best AI model. I think you ment the OpenClaw malware?

negative my friend like I've said somewhere before each AI has strengths and weaknesses but specifically Grok pizzed me off when guessing too (still nost fun for its uncensored responses and nothing being off limits) much and doesn't have cross chat memory. Gemini I thought would be good for SEO due to Google synergies which it did do a bang up job but also hallucinated hard setting up my vocal processor finally admitting after gaslighting me for 2 freaking hours and straight up said I'm guessing what's in the manual even though you uploaded pics. Claude guessed what my website said for a while after admitting it can't crawl live sites. Did vibe code a live html website end to end for me which looked ok said would be much better in pro. I've seen some good KIMI K2.5 stuff I'll play around with that a bit.

 

[Digger88]

Spot on lol

https://twitter.com/x/status/2026927250176881090

 

[Vermonter]

LLMs will tell you what it thinks you want to hear unless prompted otherwise. They work by statistical prediction, not truth seeking. Every chat you start should be prompted with something along the lines of:

"Avoid agreeing automatically. Only agree if the reasoning stands up to logical, factual, or empirical validation."

Or Devils Advocate:
"Act as a devil’s advocate and argue against my idea: [insert your idea]. Point out flaws, risks, and why it could fail, even if it seems solid."

Harsh Truth:
"Answer my question as if you cannot sugarcoat anything, and I only want the harsh truth: [insert your question]."

Etc. There are plenty out there on the web for specific tasks or you can try a prompt generator- https://originality.AI/blog/AI-prompt-generator

Or just use one chat to generate a prompt for a second chat with your actual question.

 

[cashcatz]

Claude guessed what my website said for a while after admitting it can't crawl live sites.

Well duh you are using it on mobile, it's severely gimped on phones. It absolute can read sites via the desktop app with the right permissions as well and use other tools and MCPs on your actual machine.

This is what you need:
- https://support.claude.com/en/articles/10065433-installing-claude-desktop
- https://support.claude.com/en/articles/13345190-get-started-with-cowork
- https://claude.com/download

 

[Digger88]

I knew Claude sucked LOL

https://twitter.com/x/status/2023829505518018837

 

[cashcatz]

I knew Claude sucked LOL

https://twitter.com/x/status/2023829505518018837

I think the real issue is that Anthropic employers are revolting about working with the Dept. of War. As they should be.