AIX Universal Markdown · v0.1.3 Draft

Validate before you ship.

Every chatbot produces different markdown. None of them agree on heading hierarchy, bullet markers, table format, or what survives a copy operation. AIX is the open standard for what AI-generated markdown must emit, render, and round-trip without loss.

The Markdown Problem

Same prompt. Four chatbots. Seventeen formatting dimensions of disagreement.

Markdown is the most widely used lightweight markup format on the web — and the most fragmented. Twenty-two years after John Gruber introduced it on Daring Fireball, no two parsers agree on how to interpret it, no two platforms agree on which extensions to support, and as of the AI era, no two generators agree on how to produce it.

We fed the same prompt to Claude, ChatGPT, Gemini, and Perplexity, asking each for a structured response with paragraphs, bullet lists, and a comparison table. The outputs differed on seventeen separately measurable formatting dimensions. Heading style. Bullet markers. List spacing. Section separators. Table cell encoding. One chatbot’s table didn’t even survive the copy operation — it existed on screen and not in the clipboard.

Every prior markdown standardization effort has been about parsers: given a markdown file, what HTML should come out. CommonMark and GitHub Flavored Markdown answer that question rigorously. But the question of what markdown an AI tool should emit in the first place — and what an interface must preserve as it renders and copies — has never been specified. That seat is empty.

AIX Universal Markdown fills it. Read the full history · Open the spec

The Evidence

How four chatbots format the same data.

Below are excerpts of how Perplexity, Gemini, ChatGPT, and Claude formatted a single item in a five-platform comparison. Same prompt. Same data. Four house dialects.

Perplexity H2 · dash bullets
## LangGraph (with LangChain) LangGraph sits on top of LangChain and has become a de facto standard.[2][4][5] **Key features** - Graph‑based orchestration - Deep LangChain ecosystem - Strong state management ***
U+2011 hyphens · *** separators · numbered refs
Gemini H3 · asterisk bullets
### 1. LangGraph LangGraph has established itself as a premier orchestrator. **Key Features:** * Treats workflows as cyclical graphs * First-class human-in-the-loop * Built-in state management
Skips H2 · blank lines between bullets
ChatGPT H2 bold · --- separators
## **2. LangGraph / LangChain Platform** **LangGraph** has become one of the most important developer frameworks. **Key features** * Graph-based agent orchestration * Durable execution * Human-in-the-loop support ---
Bold-on-heading · padded table cells
Claude Bold pseudo-headings
**1\. LangChain / LangGraph** LangGraph is the low-level workhorse of the LangChain ecosystem. Key features: * Stateful multi-agent workflows * Durable execution windows * Python and JavaScript support ⠀
U+2800 Braille separators · escaped 1\.

Captured May 2026 · Full variance catalog →

The Architecture

Three layers. Three failure surfaces. Three fixes.

Variance between AI markdown outputs lives in three distinct places. AIX Universal Markdown specifies normative requirements at all three.

01
Model Output Layer

What the AI model emits as raw markdown. Failure mode: heading inconsistency, escape sequences, invisible characters, non-canonical separators. Fixed by prompt engineering, fine-tuning, and model-side conformance rules.

02
Interface Render Layer

How the chatbot’s UI displays the markdown. Failure mode: lossy transforms — glyph substitution for cell values, structural elements rendered as components that can’t serialize back to markdown. Fixed by the platform vendor.

03
Copy / Export Layer

What survives a clipboard or export operation. Failure mode: silent loss of structural elements. The most damning kind of failure — the user has no way to know what was dropped. Fixed by vendor preservation guarantees and the AIX Markdown Cleaner as consumer-side defense.

The larger system

SoloTrillion.io is the wedge.

AIX — the AI Interchange Standard authored by Michael Tchong at Ubertrends LLC — defines the envelope, naming convention, persona registry, task schemas, and provider comparison format for an interoperable generative AI ecosystem. The full standard lives at aix-standard.org. This site is its developer-facing wedge: a Validator and Cleaner you can run in thirty seconds, plus the Markdown subset of the spec rendered inline.

The Solo Trillion Orchestra applies AIX across an entire agent system — platform-level orchestration with AIX as the wire format between every component. Same standard, expanded to platform scope.

AIX

AIX helps your output survive the trip between tools — a much-needed open standard for generative markdown, data, and agent workflows. Keeps every heading, table, and field intact from one tool to the next, so you render once and never re-fix the same output by hand.

Read the standard →