Where do you define it? Do you define it in CLAUDE.md? That thing gets loaded on every conversation so it should pick up.
Ohh and one thing, you are only supposed to put there something that is absolutely necessary. Things like "This project uses Typescript" is absolutely unnecessary because its something claude will auto infer. Something like "Do not push into master branch" should be there. Also as it gets loaded on every conversation, you should keep it short or else your token usage blows up.
You can have something like "YOU MUST STRICTLY FOLLOW GUIDELINES LISTED IN commit_guidelines.md BEFORE YOU RUN ANY GIT COMMANDS" if you feel your CLAUDE.md is getting too large.
But honestly, you should ask CLAUDE itself to write CLAUDE.md with strict and proper guardrails rather than yourself
Yes because most people who hate AI dont even know the ABC of AI or that they hate AI for a completely wrong reason. I dont like AI everywhere either but if we continue hating this greatest achievement of 21st century so far unconditionally we will never get to the light
First you write your requirements properly in a text or a markdown file. It should properly explain what you want and more importantly what you dont want. Then you give that to AI and dont forget to include "ask any number of clarifying questions before we begin"
Next, be prepared to answer all of its clarifying questions. If you dont know you can ask it to suggest recommended option and pros and cons of each approach
Yeah it will take an entire day. Deal with it. If your requirement is not to just create a few pages of html and css it may take upto half day to a full day but once you are clear. Ask it to write specs and plan and it will.
Then when you are happy with the plan it will start building it. At that point you can step off and let it do its thing
MX, antiX and Gentoo are the only distros that support a non systemd init system without making a big deal about it like its normal to have a different choice as a linux user
Last week in Cabo, Mexico, an adult friend thought it would be hilarious to add El Mencho to a picture of our dinner outing, text it to his parents, and tell them we met a new “friend”. He asked ChatGPT to add El Mencho to a photo. It refused. My under-13 child said “oh, I got this”, found a photo of El Mencho, asked ChatGPT to add the person from the photo to the dinner party photo and voilà, we’re enjoying drinks with El Mencho. Our friend's parents asked what’s wrong with him. I was an impressed Dad.
This is indeed some very touching story, dinner with a leader of CJNG? At that point I would ask how the hell do you even know his name but
Do you know something? When Chinese copy AI they open source it but your American companies would never do that
It could be someone copy pasting the entire reddit thread into Claude and asking it to generate a response
"You are right to raise this and we want to address this directly"
Isnt that how Claude Sonnet or Opus writes?
Rage bait?
We get it for $6.5 lol cuz we from poor region
KDE like a real grown up
How do you know?
Yess
Where do you define it? Do you define it in CLAUDE.md? That thing gets loaded on every conversation so it should pick up.
Ohh and one thing, you are only supposed to put there something that is absolutely necessary. Things like "This project uses Typescript" is absolutely unnecessary because its something claude will auto infer. Something like "Do not push into master branch" should be there. Also as it gets loaded on every conversation, you should keep it short or else your token usage blows up.
You can have something like "YOU MUST STRICTLY FOLLOW GUIDELINES LISTED IN commit_guidelines.md BEFORE YOU RUN ANY GIT COMMANDS" if you feel your CLAUDE.md is getting too large.
But honestly, you should ask CLAUDE itself to write CLAUDE.md with strict and proper guardrails rather than yourself
Yes because most people who hate AI dont even know the ABC of AI or that they hate AI for a completely wrong reason. I dont like AI everywhere either but if we continue hating this greatest achievement of 21st century so far unconditionally we will never get to the light
First you write your requirements properly in a text or a markdown file. It should properly explain what you want and more importantly what you dont want. Then you give that to AI and dont forget to include "ask any number of clarifying questions before we begin"
Next, be prepared to answer all of its clarifying questions. If you dont know you can ask it to suggest recommended option and pros and cons of each approach
Yeah it will take an entire day. Deal with it. If your requirement is not to just create a few pages of html and css it may take upto half day to a full day but once you are clear. Ask it to write specs and plan and it will.
Then when you are happy with the plan it will start building it. At that point you can step off and let it do its thing
If you ask it to it will
Using AI properly is a skill guys it does not come automatically
Also you need to demand it performs regression tests properly.
Honestly popular models are so good these days if it produces wrong result its the fault of the one in front of the screen.
We are not at the point of AGI so dont expect AI to read your mind
Wise words master
You are only allowed to post Goat Head Smoothie on Reddit
https://old.reddit.com/r/StupidFood/s/SGqbcQjdUB
MX, antiX and Gentoo are the only distros that support a non systemd init system without making a big deal about it like its normal to have a different choice as a linux user
A bug in Beta that they will fix for sure. Bugs happen people
Well if you are from outside Mexico but know his name they will question the kind of crap you watch and the websites you visit
This is indeed some very touching story, dinner with a leader of CJNG? At that point I would ask how the hell do you even know his name but
I think Quantum was last redesign which is older than half a decade now, time flies