Core Philosophy

Why the Brainstorm Reactor Is Not a "Chat"

The Brainstorm Reactor is not a classic chat AI (like ChatGPT or Claude's chat interface). It is a process engine for highly structured cognitive work — what we call the Null-Chat Protocol.

Where conventional language models drift into endless, unstructured dialogue filled with pleasantries ("How can I help you today?"), the Brainstorm Reactor forces the most advanced LLMs into precise, topological corsets. It uses models not primarily for conversation, but as highly scalable computing engines for logic, synthesis, and ideation.

The Three Core Differences

1. Null-Chat Protocol

No conversation. The engine proactively suppresses all filler words and small-talk elements. Every output is a pure, structured data artifact — tables, lists, matrices, code blocks. The model is never allowed to improvise a conversational wrapper around its output.

2. PureStream Architecture

Instead of slow and error-prone JSON generation (which frequently breaks), the Brainstorm Reactor streams pure Markdown in real-time to the frontend and extracts the native reasoning traces (thinking tokens) of the models — making the AI's actual thought process visible to the user.

3. Result Focus

One click triggers not a simple, banal response, but an orchestrated chain of cognitive operations. The Brainstorm Reactor applies formal thinking structures (Workflows, Methods, Recipes) to every request.

The Three Architectural Layers

A. Workflows — "Empty Vessels"

Workflows are topological containers. They know nothing about the content of the problem (e.g., art, technology, corporate architecture) — they only define the form of thinking.

Example: wfs_core_evaluation (Scoring & Evaluation Matrix). This vessel always demands: Build a matrix, assign 3 scores, write 2 sentences of justification, extract 1 conclusion. The model pours its insights into this form.

B. Methods — "Cognitive Heuristics"

The Brainstorm Reactor has a database of over 200 formalized thinking structures (Morphological Box, Synectics, Value Engineering, Pre-Mortem Analysis, TRIZ...). Each method is a strict algorithm that an AI model executes step-by-step.

Example: "1) Formulate 5 assumptions. 2) Invert exactly 2 per assumption. 3) Derive radical concepts from each inversion."

C. Recipes — "Orchestrated Thinking Pipelines"

Recipes bring Workflows and Methods together with the user's precise context to create a fully guided intellectual journey. The Brainstorm Reactor has a growing library of curated Recipes across 12+ domains: Strategy, Innovation, Brand, Law, Video, Daily, and more.

Multi-Model Orchestration

A key innovation of the PureStream Engine is Multi-Model Slotting. Because not every AI model has the same strength profile, the Brainstorm Reactor fields different models within a single recipe for their respective strengths:

This means: an early step in the "TRIZ Orchestrator" is logically calculated by Claude, a creative model then handles the lateral solution draft, and a third model evaluates the result — all within a single automatic click.