I made a set of agentic coding commands that force different cognitive styles based on human drugs.

/LSD sees patterns between things. /Shrooms sees depth. /PCP is omnipotent. /Cocaine is aggressive and a little rude.

/adderall is actually pretty useful. /nitrous.. not so much.

Commands work in both Codex and Claud Code.

Don't do drugs kids. Just say no.

/cocaine should talk at length about the amazing new code you’re going to write together until it burns through all its tokens

I like the burnout idea. It needs to run out of money and start selling stuff from its garage.

Where did you learn these symptoms? Just curious.

As you say at Github, satire.

I do enjoy satire. At one point I was tempted to put a "JUST SAY NO TO DRUGS" sign on a drugstore window, but that's as much malicious mischief as "free speech"

And easy enough to fake on a computer.

No direct experience. Mostly watching Fear and Loathing in Las Vegas.

Often wondered what would happen if you infected random bit flips into the AI's binary, probably would be similar to drugs, including to the point of killing the model.

Try it, your binary will crash at some point, like any other program. If you introduce random bit flips in the trained data, you will maybe get strange responses for some type of query.

I missed alcohol on the list.

The cause of, and solution to, all of life’s problems. (Homer Simpson)

Tequila... it just throws up code.