AGI / Research
AGI and how we approach it
General intelligence may not arrive as one isolated model. At inAi, we study AGI as something that may emerge from systems: models, agents, tools, memory, perception, execution, coordination, feedback, and environment working together.
This view shapes how we think about AI-native products, software for agents, open technology, and the future of work. We do not claim to have AGI. We use AGI as a research horizon and a systems thesis for how intelligence may become useful in the real world.
The four research directions are being revised and expanded with review papers, research notes, and new public work.
Why systems matter
Model capability matters, but intelligence is not only raw prediction. Useful intelligence also depends on memory, tool use, context, feedback, coordination, execution, and the ability to operate inside an environment.
That is why inAi studies intelligence as a system. The model is important, but it is not the whole structure. Tools extend what the system can do. Memory helps it continue work over time. Agents coordinate decisions and execution. Feedback changes behavior. The environment gives the system something real to understand, affect, and improve.
- models
- agents
- tools
- memory
- feedback
- environment
Four directions, one research program
inAi’s current research program is organized around four connected directions. Each examines a different layer of the same problem: the limits of capability, the movement from reasoning to action, the creation of knowledge, and the behavior of intelligent systems inside real organizations and environments.
in systemscapability · action · knowledge · environment
This direction studies the limits of reasoning, context, memory, planning, uncertainty, evaluation, tool use, and action. The aim is not only to catalogue failures, but to understand how those failures can be detected, explained, and handled before they become consequences.
Limits of IntelligenceThis direction studies tool selection, stopping, review, permissions, persistent state, multi-step execution, and recovery when an action fails or the situation changes. It examines the decision layer that turns model capability into continued work.
Agentic Decision SystemsThis direction studies synthesis, evidence mapping, contradiction detection, hypothesis generation, explanation, and the boundary between plausible language and justified conclusions. It asks how AI can help extend understanding without confusing fluent output with reliable knowledge.
AI for Knowledge CreationThis direction studies messy data, human-AI handoffs, operational decisions, review, compliance-sensitive work, process redesign, and the practical conditions under which AI becomes useful rather than merely impressive.
AI and Business Operations
Limits of Intelligence defines the boundaries of reliable capability. Agentic Decision Systems examines how capability becomes action. AI for Knowledge Creation asks what intelligence can add to understanding. AI and Business Operations tests what survives contact with real processes, institutions, and consequences.
Together, the four directions connect the systems view of AGI to the practical work of building AI-native products. They are inAi’s current research lenses, not a claim that four themes exhaust the problem of intelligence. Memory, perception, coordination, learning, feedback, and environment cut across more than one direction.
How research connects to what we build
Research does not replace products. It informs them.
inAi builds AI-native products for companies, consumers, agents, and open ecosystems. The AGI/system view helps us understand why those categories belong together: intelligent systems need real-world workflows, human-facing products, agent-operable tools, memory, interfaces, feedback, and public technology that others can inspect and use.
That does not make our products AGI. It means our product work is shaped by a long-term view of how intelligence becomes useful in the world.
Choose a research direction to update the real-world domains most directly connected to it.
Products for Agents
Software for a future where AI agents need tools, interfaces, memory, state, documentation, and execution surfaces they can use directly.
Explore Products for AgentsOpen Source
Selected tools, utilities, prototypes, and experiments released in the open for builders, developers, researchers, and the wider AI ecosystem.
View Open SourceAI for Everybody
Simple explanations of AI, agents, work, risk, and the intelligence era for broad audiences.
Read AI for EverybodyHow We Build
How inAi thinks about trust, openness, control, product maturity, privacy, and responsible public claims in the intelligence era.
Read How We BuildA stance, not a product claim
When inAi uses the word AGI, we use it as a research horizon and a systems thesis. We are not claiming that inAi has AGI, sells AGI, or that any current product is AGI.
Our position is narrower and more useful: if general intelligence emerges, it may emerge from systems, not from one isolated model alone. That belief shapes what we study and how we build.
Read AGI as a SystemResearch collaboration
Serious research improves through exchange, criticism, and application. inAi is open to discussions with researchers, universities, laboratories, companies, public institutions, funders, and independent specialists whose work intersects with these directions.
Research and academia
Potential work may include joint literature reviews, external critique, co-authored research notes, shared evaluations or experiments, student research, and collaboration around a defined research question.
- Joint literature reviews
- External critique
- Shared evaluations
- Student research
Funding and grants
Investors, grantmakers, and independent supporters may be interested in defined research programs, review papers, evaluations, public tools, or educational work. Any funded work should have a clear scope, intended outputs, status, and reporting path.
- Defined research programs
- Review papers
- Public tools
- Educational work
Applied partnerships
Companies and public institutions may bring operational questions that benefit from structured study: intelligent workflows, agentic systems, knowledge processes, evaluation, human-AI handoffs, or the practical limits of current AI.
- Operational AI questions
- Knowledge workflows
- Human-AI handoffs
- Applied evaluation
Any collaboration begins with a defined question, scope, responsibilities, publication status, and intended use of the results. No partnership or external review is implied before it has actually been agreed.
A research horizon connected to the real world
AGI gives the program its horizon. The four research directions give it structure. Products, services, Open Source, and AI for Everybody give it contact with the world.
The purpose of the research program is to make those connections explicit, testable, and open to serious examination.