Protocol v0.1 (Experimental Draft)
Artificial Intelligence Universal Reference Marker
Why AIURM?
As artificial intelligence becomes increasingly present in our tools and workflows, we need to rethink how we interact with it. AIURM proposes a lightweight, universal layer to organize and structure these interactions. It transforms scattered prompts and responses into clear, modular flows, bringing traceability and control, from chat sessions to more complex system integrations.
It is a fundamentally simple concept that, when combined with the power of current and future Large Language Models (LLMs), can unfold into different levels of complexity and possibilities still being explored.
Markers themselves are not new. What AIURM introduces is a systematic use: they are not just labels, but reference anchors for data, instructions, and results. They enable modular, auditable, and reusable flows, reducing ambiguity and inference effort for both humans and AIs.
AIURM defines a clear syntax for assigning and referencing markers. This allows any content, logic, or result to be referenced, reused, compared, or exported, both by humans and AI.
Example:
Question: What was the Big Bang?
Answer: The Big Bang was the initial explosion that gave rise to the universe. [*2]
Why markers?
Markers are reference points for data, instructions, and results. They enable modular, auditable, and reusable flows. They reduce ambiguity and inference effort for both humans and AIs. With clear syntax for assigning and referencing markers, any content, logic, or result can be referenced, reused, compared, or exported, both by humans and AI.
Intention Suffixes: Controlling the Response
To further structure outputs, AIURM defines Intention Suffixes (#0, #1, #2, #3), allowing control over the level of detail and type of AI response.
- #0 Silent operation / confirmation only
- #1 Short / concise response
- #2 Intermediate response
- #3 Most detailed response possible
- No suffix: The AI decides
Example 1: Controlling response granularity using intention suffixes
Question: What was the Big Bang? #1
Answer: The Big Bang was the initial… (concise response)
Question: What was the Big Bang? #3
Answer: The Big Bang was the initial… (Most detailed response possible)
Example 2: Silent operation and marker assignment
Question: { JSON data… } [*analysis_data] #0
Answer: Done [*analysis_data] [*26]
DLR: Data, Logic, Result
The DLR methodology encourages organizing information into three distinct and complementary blocks:
Data… [*data_x] #0
Logic… [*logic_x] #0
Result… [*result_x]
Example of operation:
Apply *logic_x to *data_x and generate [*result_x]
Reference, Auditing, and Versioning
Instead of repeating content, reference the marker for comparison, transformation, or export:
compare *result_x with *other_result_y
For auditing, use commands like:
show dependency tree *marker
For more advanced audits:
generate full marker dependency tree in Graphviz/DOT format
This makes it easy to audit relationships, dependencies, and versions of each step in the flow.
Each update or output receives a new marker, forming a natural history of versions and dependencies.
Applicable in Any AI Interaction
AIURM is not limited to a specific domain. The same logic can be applied in finance, legal, commercial, API integration, and many other contexts where structured, auditable, and reusable AI outputs are needed.
Current Limitations
AIURM, in its experimental stage, operates within the limitations of current AI technologies. The rules are transmitted to the AI via prompt or API, requiring the model to interpret and follow them in each new session. Markers exist only within the session unless externally persisted. The protocol depends on the AI’s ability to understand and follow the instructions and maintain context, which LLMs already do consistently depending on the model level.
Its full potential will be realized as AI platforms support larger contexts and more advanced persistence mechanisms.
Complementary to Other Solutions
AIURM does not replace existing integration, automation, or orchestration solutions. Its goal is to complement those platforms by providing an additional layer of structured control, traceability, and governance over AI interactions. It can coexist with and enhance pipelines, agents, integrations, and workflow systems.
Contribute to Its Evolution
These are concepts too broad to be developed and evolved by a single person.
Participate by testing, studying, creating new applications, and sharing examples of use and integration. Your contribution can help expand AIURM’s possibilities.
Try It in Practice
Explore the step-by-step onboarding, with practical tests and real workflows:
Onboarding Page
For full details, definition, and hands-on onboarding:
https://aiurm.org
AIURM Protocol v0.1 (Experimental Draft)
Public domain (CC0). Open for community discussion, testing, and evolution.
Top comments (0)