13 pointsby nich25335 hours ago4 comments

nich25335 hours ago
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.
hirpslop4 hours ago
/cocaine should talk at length about the amazing new code you’re going to write together until it burns through all its tokens
k3104 hours ago
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.
boznz4 hours ago
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.
- ahofmann3 hours ago
  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.