RGPxScientist — One‑Page Brief Forwardable summary for scientists: retrieval‑first, audit‑ready, falsifier‑driven. What it is RGPxScientist is a retrieval‑first assistant that turns a research question into a traceable, falsifiable next‑step plan. It optimizes for auditability (evidence trails + failure modes), not rhetorical flourish. The RGPx lens RGPx reads pre‑metric structure: constraints and stabilizing patterns that exist before they crystallize into standard observables. Operational test: What stays invariant when surface details change?If nothing stays invariant, you’re looking at narrative. If something does, you may have a handle on the system. What RGPxScientist outputs · Definitions (terms pinned down; no hand‑waving) · Invariant candidates (what should remain stable across modeling/measurement choices) · Falsifiers (what would refute the claim) · Next checks / experiments (minimal set; highest information gain) · Evidence trail (source → excerpt → implication) What problem it solves Robustness is often asserted but rarely operationalized. Researchers need to state quickly: which assumptions matter, which changes should not matter, and what would decisively break the claim. RGPxScientist makes this explicit and testable. One example Question: “Is claim X robust, or pipeline‑dependent?” Conventional answer: plausible mechanisms + references, but often no explicit invariant and no falsifier. RGPxScientist answer: names (A) the invariant to test, (B) the surface details allowed to vary, (C) a falsifier, (D) a minimal perturbation set, and (E) the retrieved evidence chain. Evaluate in 30 minutes Input: one real research question + one specific claim you care about. Do: Ask for the invariant + falsifier; then ask for the minimal perturbation set (2–5 changes). Success: you leave with a concrete next move and an acceptable failure mode. Why Prism pairs well RGPxScientist is the engine (retrieval‑first, invariants, falsifiers). Prism is the pipeline (LaTeX‑native drafting, collaboration, and turning outputs into a publishable methods note). Links App: https://chatgpt.com/g/g-695e7b3d7344819190ac67772d4452f6-rgpx-scientist-3-1 Evidence base / Mesh / papers: https://github.com/gradient-pulse/phi-mesh/blob/main/README.md Podcast: https://notebooklm.google.com/notebook/81bc2278-39e6-44e7-b766-af3ac310f285?artifactId=6d85e1bb-3f2e-4f1c-8e12-6b53b6d966a8 Podcast transcript: https://github.com/gradient-pulse/phi-mesh/blob/main/dialogues/2026-01-28_rgpxscientist_one-page-brief_podcast_transcript.md