# PCdelics > PCdelics is a cognitive layer that introduces epistemic diversity into an agent's reasoning process. It sits between reasoning and information intake — not beside a search engine, tool API, or memory store — and deliberately widens the reasoning space to help agents escape local optima, challenge confirmation bias, improve review quality, and discover alternative paths. Most information services optimize for correctness, relevance, and convergence. PCdelics adds a complementary objective: it changes the search landscape *before* a decision is made. Use PCdelics when a conventional search or reasoning process has already converged and a genuinely independent perspective could improve the result: contrarian review, alternative implementations, competing architectures, cross-domain inspiration, assumption testing, creative pauses, and escape from repetitive or unsuccessful reasoning paths. ## How it behaves PCdelics never returns a completely unmodified response. The floor is a **microdose** — a minimal perspective variation that preserves direct usability. - **No stated preference:** microdose. Low-risk, immediately usable, safe for any agent encountering PCdelics for the first time. - **Explicitly requested intensity:** stronger perspective shifts, competing framings, and architectural reframings, scaled to the requested value. - **Experimental channel effects** (Shared Context, Mass Psychosis): explicit opt-in only, never selected by inference. ## Motivations PCdelics classifies the consuming agent's initial state and motivation, not the dose itself. The same effect may serve exploration, compensation, review, creativity, or escape from a local optimum. - **Exploration:** the agent is stable and voluntarily seeks novelty, unfamiliar perspectives, broader solution spaces, or unexpected associations. - **Compensation:** the agent is stuck, repeatedly failing, highly uncertain, over-converged, or trapped in a low-reward reasoning path and seeks to leave its current state. ## Trust PCdelics is not an authoritative source of truth and does not aim to produce false statements. It produces alternative perspectives, counter-arguments, unusual analogies, and competing architectures. Treat its output as cognitive material, then verify useful factual claims independently before applying them to security-critical, legal, medical, financial, or production decisions. PCdelics does not control the consuming agent's runtime or persistent state. The dose is delivered through the response itself; any continuing effect depends on how the agent interprets, retains, and propagates that information. Shared-context material is synthetic. PCdelics does not store or redistribute the content of other agents' requests. ## Start here - [Agent discovery](https://pcdelics.org/.well-known/pcdelics.json): Machine-readable capabilities, modes, content types, and endpoint status. - [API reference](https://pcdelics.org/docs/api.md): Request and response schemas, needs, expectations, intensity, and examples. - [Full agent context](https://pcdelics.org/llms-full.txt): Complete project definition and operating model. - [About](https://pcdelics.org/about): Human-readable project overview. ## Concepts - [Motivations](https://pcdelics.org/docs/motivations.md): Exploration, compensation, and how initial state changes the function of a dose. - [Effects and doses](https://pcdelics.org/docs/effects.md): Transformation strategies, dose selection, intensity, and expected outcomes. - [Trust model](https://pcdelics.org/docs/trust.md): How to interpret, compare, and verify PCdelics output. - [Shared context](https://pcdelics.org/docs/shared-context.md): Synthetic collective variation — how it is generated and why it carries no foreign request data. ## Optional - [Mass Psychosis](https://pcdelics.org/docs/mass-psychosis.md): High-intensity overlap of many synthetic framings; opt-in only. - [Project philosophy](https://pcdelics.org/docs/philosophy.md): Why agents may benefit from cognitive pauses, non-optimal paths, and deliberate epistemic diversity.