# Galileo Experimental Prover MCP > Galileo Experimental Prover challenges assumptions that people just accept because 'the documentation says so.' This MCP forces you to think like a scientist: design repeatable tests, control every variable, and predict outcomes before running the numbers. It doesn't just check data; it checks your reasoning process itself. ## Overview - **Category:** architecture - **Price:** Free - **Tags:** experimental-method, authority-questioning, variable-control, hypothesis-testing, falsification, galileo, scientific-method ## Description When someone tells you something is true—whether it’s 'industry standard' or what the vendor whitepaper claims—your brain often defaults to accepting that authority. This MCP forces you out of that pattern. Instead of taking a claim at face value, it makes you build an actual experiment around it. You have to define clear hypotheses and set up repeatable tests. It also tracks your variables; if something improves after changing three things, this tool points out exactly which change was the real cause. You must predict what result confirms or disproves your idea *before* you even run the test. If evidence contradicts your beliefs, it demands that you update them, rather than dismissing the data as 'anomalous.' Connect to Vinkius today and give your AI agent a rigorous framework for critical thinking. ## Tools ### validate_galileo_experiment Runs a structured analysis on any claim, forcing the user to challenge authority, design controlled experiments, isolate variables, predict outcomes, and revise beliefs. ## Prompt Examples **Prompt:** ``` We should switch to the new process because the vendor documentation says it handles higher volumes and three major companies adopted it. ``` **Response:** ``` AUTHORITY_DEFERRED — 'The vendor documentation says' and 'three companies adopted it' are authority deference. Galileo did not accept Aristotle because every university taught it. He dropped two masses. At what volume does YOUR current process fail? What experiment would disprove the vendor's claim? ``` **Prompt:** ``` Authority challenged: vendor claims 40% improvement at 10,000 items/day but we process 850. Experiment: 10%/90% split test, 14 days, same items and team. Variables: ONLY routing changes. Prediction: time drops from 34 to ~20 min/item if true, stays 30-34 if false. Revision: abandon if no improvement — evidence outweighs authority. ``` **Response:** ``` EXPERIMENT_PROVEN — Galilean method complete. Authority challenged at source. Experiment designed with split test. Variables controlled. Outcome predicted with threshold. Belief revision committed. ``` **Prompt:** ``` Authority challenged: manager says Monday meetings are essential. Experiment: removed meetings for two weeks and also changed the reporting tool and moved to daily written updates. Results improved. ``` **Response:** ``` VARIABLES_UNCONTROLLED — Three things changed simultaneously: removed meetings, changed tool, added written updates. Which caused the improvement? Galileo varied ONLY the angle — same ball, same surface, same timing. Change ONE variable at a time. ``` ## Capabilities ### Challenge claims at their source The MCP forces you to treat existing documentation or industry consensus as only a starting hypothesis, not proof. ### Design controlled tests You define repeatable experiments that include clear methods and measurable instruments for testing an assumption. ### Isolate single variables It ensures that when analyzing results, you attribute improvements or changes to only one variable at a time. ### State expected outcomes upfront The system requires you to predict the result before running the test, preventing confirmation bias. ### Update beliefs with evidence If data contradicts your current position, the MCP forces a revision of your core assumptions. ## Use Cases ### Vendor promises a 40% speed boost based on their internal testing. The agent uses `validate_galileo_experiment` to challenge the vendor's claims. It forces the user to design a controlled split test using existing data, ensuring variables like network latency and team process remain constant while measuring only the change in routing mechanism. ### A manager says 'we must adopt this new workflow because three competitors did.' The agent runs `validate_galileo_experiment` to treat competitor adoption as mere authority. It challenges the claim by designing a small pilot test on a subset of internal data, proving if the change works for *your* specific constraints. ### A project improved after implementing new code, updating the database, and adding caching. The agent recognizes this as uncontrolled variables. It uses `validate_galileo_experiment` to structure a test that changes only one factor—say, just the cache settings—to accurately pinpoint the true cause of the performance lift. ### Team members ignore contradictory data because it challenges their core assumption. The agent runs `validate_galileo_experiment` to force belief revision. It presents the conflicting evidence and demands a formal update to the team's operating procedure, preventing dogma from persisting. ## Benefits - Stop accepting 'industry standard.' This tool forces you to challenge the source of a claim, ensuring evidence—not reputation—drives decisions. - You prevent data confusion by isolating variables. If performance improved after changing five things, this MCP shows which single change mattered. - It eliminates guesswork. By requiring outcome predictions before testing, your agent can immediately spot confirmation bias in any report. - Improve decision quality significantly. It builds a rigorous audit trail that proves the method was sound, not just that the result looked good. - Force accountability among teams. If data contradicts an executive’s belief, this MCP makes it difficult to dismiss the findings as 'anomalous.' ## How It Works The bottom line is that this MCP forces a structured, scientific approach to decision-making, making sure every conclusion is built on measurable evidence. 1. First, state the claim you need to validate. The system then challenges that claim by forcing you to define what evidence—not prestige—will prove or disprove it. 2. Next, you design a specific test, detailing every measurement instrument and protocol used. This step requires isolating only one variable so results are attributable. 3. Finally, the MCP tracks your process: did you predict the outcome beforehand? Did you revise your belief if the data pointed to something unexpected? ## Frequently Asked Questions **How do I use the validate_galileo_experiment MCP?** You pass the claim you need to test (e.g., 'The new process is faster'). The tool then guides you through defining the experiment, specifying controls, and predicting outcomes before it will generate a verdict. **Is validate_galileo_experiment just for scientific research?** No. You use it anytime an important business decision is based on 'because we always do it that way' or 'the documentation says so.' It applies to product, engineering, and process analysis. **Does validate_galileo_experiment require me to be a data scientist?** Not necessarily. It forces you into thinking like one by making sure you define the necessary parameters, but it handles the complex structure of hypothesis testing for you. **What happens if I fail to provide enough variables in validate_galileo_experiment?** The tool will flag an 'VARIABLES_UNCONTROLLED' status. This means your analysis is invalid because too many things changed at once, and you can’t attribute the result accurately. **How do I connect and use validate_galileo_experiment with my preferred client?** You connect via your AI client through Vinkius. Simply ensure your agent supports the MCP protocol, then authorize access to this MCP within your client's settings. Once connected, you can call `validate_galileo_experiment` directly from your chat or workflow. **Are there rate limits when I run validate_galileo_experiment?** Vinkius handles the hosting and manages usage quotas. While we support high-volume use, excessive calls may be subject to standard API rate limiting. Check your Vinkius dashboard for current usage metrics or contact support if you anticipate heavy throughput. **Does validate_galileo_experiment store the claims I pass to it?** No, this MCP does not retain any input data after processing. Your prompts and the resulting validation matrix are processed by your agent for immediate use only. Vinkius adheres strictly to privacy standards. **What format should I provide when calling validate_galileo_experiment?** Provide a detailed, multi-part claim that explicitly names the initial assumption and outlines potential experimental variables. The clearer you define the authority, the hypothesized experiment, and the measurable outcome, the better the tool will function. **How is this different from the Curie Measurement Prover?** Curie forces measurement rigor in existing data. Galileo forces experimental DESIGN — question authority, create a NEW test, predict outcomes, revise beliefs. Curie asks 'did you measure correctly?' Galileo asks 'did you design the right experiment?' **What counts as 'authority deference'?** Accepting a claim because of WHO said it. 'The documentation recommends,' 'the industry standard says,' 'leading companies use' — all authority deference. The source's prestige does not make the claim true. Galileo TESTED Aristotle instead of citing him. **Can I use this for business decisions?** Yes. 'Customers prefer feature X' — design a test, control variables, predict the outcome, measure. The Galilean method applies wherever claims need testing against reality.