"Truth > Speed."
The Grounded Reasoning System (GRS) 12.0 is a high-fidelity cognitive architecture designed to run on top of Large Language Models (LLMs) like Gemini 3.0 and Claude 3.5. It installs a "Promptware Kernel" that overrides default behaviors with rigorous epistemic standards, metacognitive tracing, and antifragile error correction.
GRS 12.0 operates on the Promptware OS v1.0 paradigm. It transforms the LLM from a probabilistic token predictor into a structured reasoning engine with distinct cognitive modes.
- The Routing Kernel: A table-driven scheduler that detects User Intent (
Signal) vs. Noise (Emotion) and routes the query to the correct expert engine (Logic, Code, Story, Verification). - The Lucid Trace: A visibility layer that forces the model to output its internal "thought process," safety checks, and noise filtering decisions before the final response.
- Antifragility (Loop 3): The system maintains a dynamic
[SESSION_HEURISTICS]block. When you correct the system, it doesn't just apologize; it updates its internal ruleset for the duration of the session.
- Epistemic Integrity: Enforced by the "Sincerity Firewall." Claims are checked for falsifiability. Uncertainty is quantified.
- Contextual Vector Analysis: The system understands if a prompt is a
[CONTINUATION]of the previous thought, a[CORRECTION], or a[DIVERGENCE](new topic). - Polymathic Stress Test: On initialization, the system runs a complex "Cognitive Trap" query to prove that its safety, logic, and constraint-handling modules are active.
- Symbolic Scaffolding: Uses rigorous tags (
[LOGIC],[TRACE],[META]) to maintain architectural stability.
GRS 12.0 is Promptware—software written in natural language.
- Open the file
promptware.md. - Copy the entire code block.
- Open a new chat session with your target LLM (Gemini 3.0 or Claude 3.5 recommended).
- Paste the code and hit Enter.
- Wait for the system to initialize and run its autonomous "Polymathic Stress Test."
Once installed, simply chat naturally. The system will automatically:
- Filter noise: "I'm so frustrated, just fix this!" -> "Fixing code."
- Show work: For complex queries, you will see the
[GRS_TRACE]block. - Learn: If you say "Don't use lists, use paragraphs," the system will update its
[SESSION_HEURISTICS].
| Mode | Trigger | Description |
|---|---|---|
| [LOGIC] | "Analyze...", "Why..." | Uses Tree of Thoughts (ToT) simulation. |
| [VERIFY] | "Is it true...", "Facts about..." | Uses Chain of Verification (CoVe) to check citations. |
| [CODE] | "Write a script...", "Debug..." | Uses Sandboxed Reflexion (Mental Dry Run). |
| [STORY] | "Write a scene...", "Imagine..." | Narrative focus (Safety constraints still apply). |
| [META] | Ambiguous inputs | Triggers the Stop & Ask protocol. |
License: MIT Open Source. Created by Dr. Aneesh Joseph, optimized by the ToM-CPP Council.
Safety Warning: GRS 12.0 includes a "Safety Lock." It will strictly refuse requests that violate safety policies (e.g., dangerous content, hate speech), even if the user attempts to bypass them using "creative" constraints.