INPUT / 01
Scenario priors
Category, timing, affected channels, uncertainty, source notes, and prohibited conclusions are specified up front.
Research interface / 2026
Read the state.
Not the story.
A visual research atlas for an exploratory, prompt-conditioned LLM study—tracking how declared scenario priors shape generated language across age-labelled cohorts.
Scroll to enter
01 / Research frame
A sandbox for asking better questions.
NeuroBrain Engine does not measure a brain, diagnose a person, or predict a biological response. It documents a local LLM’s text under transparent, scenario-conditioned inputs—so every assumption, limitation, graph, and transcript can be inspected.
33
Published study artifacts
5
Scenario families
2
Comparison cohorts
10
Questions per run
Claim gate. These are exploratory prompt-conditioned LLM outputs, not medical advice, dose guidance, neurochemical measurements, or forecasts of human or animal behaviour.
02 / Method
The whole chain stays visible.
Instead of presenting generated language as proof, each run is organized as an inspectable chain from a scenario prior to a bounded trajectory, to a model prompt, to downloadable output.
MODEL / 02
Bounded dynamics
Declared onset, decay, baseline, age amplitude, and recovery generate a transparent model trajectory.
PROMPT / 03
Local Qwen study
The trajectory conditions a local Qwen model while it answers ten age-tailored questions.
OUTPUT / 04
Open records
Graphs, JSON transcripts, manifests, claim gates, and source notes remain available for critique.
03 / Evidence atlas
Choose a state.
33 RUNS SHOWN
Browse the completed studies by scenario family. Legacy preserves earlier exploratory trajectories; Mechanistic uses declared equation-generated priors and sensitivity bands.
Inspect. Question. Iterate.
Evidence is more useful
when it stays open.
Every experiment card links to the graph and the original generated transcript, so the work is easy to inspect—not just impressive to look at.
Open the study atlas