Intention
Infer what the user actually wants, why they want it, and what success should look like.
Iris LabxResearch / Large Intention Model
LIM turns intention into action.
A Large Intention Model is a post-trained model architecture designed to understand what a human wants to accomplish, plan the work, execute through tools and agents, and verify that the outcome matches the original intention.
LLMs answer:
What should I say next?LAMs answer:
What should I click next?LIMs answer:
What does the user intend, and how do we complete it?Why LIM exists
Most AI systems are still reactive. They generate text, retrieve answers, or operate interfaces one action at a time. But real users usually do not want an explanation of how to do something. They want the outcome.
LIM closes the gap between response generation and fulfilled intention. It reasons about the goal behind the request, the constraints around it, the tools required, and the evidence needed to know the work is complete.
The IPAV loop
Plan
Decompose the intention into executable steps across tools, agents, APIs, browsers, and enterprise systems.
Act
Execute the plan through MCP tools, APIs, CLIs, AgentNet capabilities, and coordinated agents.
Verify
Check whether the outcome satisfies the original intention, then retry, refine, or escalate when needed.
Architecture
LIM extends foundation models with modules built for intention, planning, execution, and verification. In Iris Labx systems, the model is connected to tools and agents through MCP, APIs, CLI environments, A2A coordination, and AgentNet capability discovery.
- Intent Decoder
- Planner
- Execution Orchestrator
- Verifier
- Multimodal Context Layer
Examples
Play Selena Gomez on YouTube
Start the music now
Open YouTube, search, select the right result, press play.Find 10 SaaS leads
Build a qualified pipeline
Query sources, filter companies, enrich profiles, rank, deliver a structured list.Prepare the Q3 report
Create a business-ready artifact
Pull data, generate charts, write analysis, format the document, verify completeness.Book a flight to Tokyo under $800
Complete a travel booking
Search flights, compare options, select itinerary, book, confirm.From LIM to MLIM
A Multimodal Large Intention Model extends the same idea beyond text. It can use visual, audio, sensor, application, and environmental context to infer intention in physical and digital settings.
That matters for enterprise operations, robotics, manufacturing, healthcare, field services, and any workflow where the system must understand context before it acts.