Open-Source Project Simulates Psychedelic States for AI Coding Assistants

A new open-source project called just_say_no introduces provocative personality modifiers for AI coding assistants, simulating how different substances might alter cognitive approaches to programming tasks. Created by developer Nicholas Haan, the toolkit offers 12 distinct "altered state" commands that fundamentally change how Claude Code and OpenAI Codex CLI respond to coding requests.

The system works through carefully engineered prompts that modify the AI's approach along specific cognitive axes:

• Installation: Users add command files to their project directories (~/.claude/commands/ for Claude or ~/.agents/skills/ for Codex CLI)
• Usage: Commands follow /substance [dose] [task] syntax with optional dose levels (light/medium/heavy)
• Dynamic Context: Three commands (marijuana, adderall, caffeine) incorporate real-time git data like status and diff outputs

Each substance targets distinct cognitive dimensions:

1. Pattern Recognition (/lsd): Enhances metaphorical connections and synesthesia at higher doses, seeing "code as living geometry"
2. Aggression & Velocity (/cocaine): Produces rapid, confident solutions with self-congratulatory commentary
3. Philosophical Depth (/shrooms): Focuses intensely on single elements, finding "cosmic significance in error handlers"
4. Mechanical Output (/speed): Generates silent, assembly-line coding without deliberation

Notable differentiators include /ketamine's abstract detachment ("code as floating geometry") versus /xanax's apathetic minimalism ("idk just add try-catch"), or how /caffeine's scattered brainstorming contrasts with /cocaine's decisive shipping mentality.

The project demonstrates how constrained perspectives can spark creativity. At light doses, /adderall prescribed produces exceptionally thorough code documentation, while /ayahuasca ceremony approaches refactoring as spiritual ritual. As creator Haan notes: "These commands force Claude out of default helpful-assistant mode into genuinely different cognitive styles."

Though presented as satire with prominent disclaimers against actual drug use, the project raises serious questions about personality engineering in AI tools. Unlike superficial tone adjustments, these prompts modify fundamental problem-solving approaches—prioritizing either cosmic connections (/lsd), ruthless efficiency (/speed), or spiritual reverence (/ayahuasca).

The complete command reference details all 12 substances and their dose-dependent behaviors. While not a funded startup, the project's viral traction highlights developer interest in unconventional AI interaction models beyond conventional chatbot interfaces.