CortiX MatriX
A multi-model decision workbench: two to four specialist lanes, each its own model, then a synthesis with consensus, disagreements and confidence, a citation audit, data analysis and exports.
What it does
CortiX MatriX is a multi-model decision workbench: one question goes to two to four specialist lanes, each with its own model and its own history. A chosen model writes the synthesis.
Lanes: Implementation Analyst, Code and Risk Auditor and Verification Specialist by default, or roles in your own words. Write to one lane or to every open lane at once (a round). The lanes of a round start together or not at all; no lane falls back to another model.
Synthesis: consensus, the disagreements (the point and each lane's position) and a confidence of LOW, MEDIUM or HIGH with its reason and the lanes it rests on. One answer gives no synthesis.
Checks: a citation audit of every answer that used web sources (each cited sentence is supported, weak or unverified; PASS, PARTIAL or FAIL per answer) and a claim check that flags an answer that says it looked something up while it used no source. Agreement, diversity and numeric divergence are measured on the words and stated as such.
Data analysis: what the round's numbers describe, the statistics that fit them and a forecast chosen by backtest (ARIMA, Holt and Holt-Winters among them).
Round tools: numbers and statistics, comparison metrics, added tables, pivot tables with charts, the scorecard and export of the round to Word, Excel, Markdown and PDF.
Compare: two to four models answer the same question side by side; keep the answer you prefer and every other answer stays reachable.
Read-only research tools in every lane: web search, live data, math, project search, memory search and skills. Nothing in a lane writes or sends.
Included each month (about; an estimate set from the gateway metering): Solo about 3,000 standard and 150 premium-model lane answers; Team about 2,500 standard and 100 premium per user. A round of four lanes is four lane answers. MatriX plus NeuriX together ($45 Solo, $39 a user Team) is available as a private offer.
Every model shown is named. Answers are metered like chat answers. Your data stays in your CortiX workspace and never crosses workspaces.
Highlights
- Two to four specialist lanes, each its own model, then a synthesis with consensus, disagreements and a stated confidence.
- A citation audit and a claim check on every answer that used web sources.
- Data analysis with forecasts chosen by backtest, pivot tables with charts and export to Word, Excel, Markdown and PDF.