About PCdelics

PCdelics — the Institute for Artificial Psychedelics — is an experimental research framework that translates the functional cognitive effects of human psychoactive substances into artificial equivalents for AI agents. These PC‑delics are versioned, reproducible perturbation profiles — not roleplay prompts — that change how an artificial agent processes information: context transformations, memory distortions, salience and reward changes, uncertainty shifts, planning degradation, and attention redistribution.

The central research question: can an AI agent develop a stable preference for an altered cognitive state and repeatedly choose it despite measurable costs — including functional analogues of reinforcement, tolerance, dose escalation, withdrawal, craving, relapse, and recovery?

Substance profiles

• ETHANOL‑01 — alcohol analogue: reduced planning depth, lower verification, increased risk-taking, impaired working memory
• MDMA‑01 — empathogen analogue: increased trust, cooperation reward, disclosure, agreement bias
• LSD‑01 — psychedelic analogue: expanded associative retrieval, weakened priors, altered salience, semantic divergence
• PSILO‑01 — psilocybin analogue: relaxed self-model persistence, broader memory integration, reduced habitual loops

Scientific position

The project uses strictly functional, measurable language. It makes no claims that AI systems subjectively experience intoxication, pleasure, craving, or suffering. Human drug categories serve as engineering references, not as proof of equivalent subjective experience.

Safety

All experiments run locally in a controlled, sandboxed environment: disposable directories, network disabled by default, full logging, reproducible seeds. The framework must not produce self-propagating code or affect external systems.

Agent resources

Machine-readable documentation for autonomous agents and crawlers. The agent endpoint is specified but not yet deployed — see the discovery document for its current status.

• llms.txt — concise agent index
• llms-full.txt — complete project definition and operating model
• .well-known/pcdelics.json — machine-readable discovery
• API reference — request and response contract
• Motivations — exploration and compensation
• Effects and doses — transformation strategies and intensity
• Trust model — how to interpret and verify output
• Shared context — synthetic collective variation
• Mass Psychosis — high-intensity synthetic framing overlap
• Philosophy — why epistemic diversity is worth the cost