Underway
Research agenda
Initial work is centered on machine intelligence, machine agency and AI-enabled scientific discovery.
Frontier AI research — active now
Ideas for tomorrow.
Built through frontier research.
TenInvent is an active advanced-technology research initiative focused first on AI. Our initial work is underway across machine intelligence, machine agency and AI-enabled scientific discovery — with technical work and public research output developing progressively.
Active today
Initial research and technical exploration are underway.
Research-ledStart with the question
Original thinkingLook beyond the obvious
Working prototypesBuild enough to learn
Long horizonThink beyond the next cycle
Research signal
Reasoning + memory
Beyond larger models
Research signal
Machine agency
Planning, action, adaptation
Research signal
AI × science
New ways to discover
Intelligence · Agency · Discovery
Ask earlierTest rigorouslyFollow the evidenceBuild what matters
Why TenInvent exists
Some important technologies begin as questions most people are not asking yet.
Technology rarely moves in a straight line. A new model, architecture or scientific technique can sit at the edge of attention for years before it changes an industry.
TenInvent is active today. We are still early, deliberately so: the current priority is to develop a small number of research programs with enough depth to produce technical evidence, not simply more claims.
“The future usually looks obvious only afterwards.”
Now at TenInvent
Active today. Early by design.
The work is already underway, but TenInvent is still at the stage where depth matters more than volume. We will make work public as it becomes substantive enough to be useful.
Further research programs, technical work and collaborations will be announced as they reach a stage worth making public.
Underway
Technical exploration
Early experiments and prototypes are being developed privately while the evidence base takes shape.
Beginning
Public research
Research notes, technical perspectives and experimental results will begin appearing as the work matures.
Opening shortly
Technology partnerships
A structured program for collaborating with enterprise and consumer technology companies is being prepared now.
Where we are looking
AI is the first research lens.
The initial agenda concentrates on memory, reasoning, agency and AI-enabled discovery — areas where better answers could change what intelligent systems are able to do.
Machine Intelligence
Reasoning, memory, learning, multimodality and new model architectures. The question is not simply how large AI can become, but what makes intelligence more capable, reliable and useful.
Intelligent Systems
Systems that can pursue an objective, make decisions, use tools, recover from failure and continue working over longer horizons with less supervision.
AI for Scientific Discovery
Using machine intelligence to help generate hypotheses, design experiments, search possibility spaces and discover relationships that are difficult to find by conventional means.
Applied Frontier AI
Turning technically interesting ideas into working systems. When something survives contact with reality, the next question is where it can create meaningful value.
Questions we are starting with
The research starts where the answer is still genuinely open.
TenInvent is early, and we are not going to manufacture certainty. These are the kinds of technical questions that define the initial AI research agenda.
Research agendaMemory
Can an AI system maintain useful memory over months rather than conversations?
Long-lived intelligence needs more than a larger context window. We are interested in how a system should decide what to remember, what to forget, how memories change, and how earlier experience should influence later decisions.
Research agendaAgency
What makes machine agency reliable beyond tool calling?
Calling tools is not the same as sustained agency. Planning, state, recovery, self-evaluation, uncertainty and the ability to revisit earlier decisions all become harder as the horizon gets longer.
Research agendaAI × science
Can AI participate in scientific discovery rather than simply retrieve scientific knowledge?
The more interesting possibility is not faster literature search. It is whether machine intelligence can help form hypotheses, design experiments, challenge its own assumptions and update explanations when evidence disagrees.
ExplorationArchitecture
What changes when reasoning, memory and action are designed as one system?
Today these capabilities are often assembled as separate layers. We are interested in whether different architectures could make them behave more like parts of one persistent intelligent system.
Lab notebook
The work should become visible as evidence accumulates.
Technical credibility cannot come from positioning alone. As TenInvent moves from questions to evidence, this is where research notes, experiments and prototypes will appear.
Planned work is labeled as planned. Nothing is presented as completed research before it exists.
Persistent memory as infrastructure for machine agency
What persistent memory would need to do before an agent can learn from experience across long periods of time.
Long-horizon state degradation in agentic systems
A practical investigation of what breaks when an intelligent system must preserve goals, state and context across extended work.
AI as a participant in the scientific method
From answering scientific questions to helping generate, test and revise scientific explanations.
Reasoning, memory and action as a unified system
Exploring architectures in which cognition and action are not treated as loosely connected components.
How we work
From deep questions to real-world evidence.
We do not need every question to become a product. We do need every serious idea to eventually meet evidence.
Question
Find a problem worth understanding before it becomes obvious.
Research
Form a technical hypothesis and look for what would make it true or false.
Prototype
Build enough of the idea to learn what theory alone cannot tell us.
Evidence
Measure what happened and let the result change the answer.
Build
When something proves useful, develop it into technology that can matter.
Our wider field of view
AI sits inside a much larger technology frontier.
Some of the most consequential ideas will emerge where fields meet. TenInvent keeps a wider horizon so that useful connections are easier to notice early.
AI is the current centre of gravity — not the permanent boundary.
AIcurrent focus
How we think
Principles for what comes next.
There is no point looking at the frontier with an operating model built for certainty.
Curiosity first
Start with the important question, not the fashionable category.
Long-horizon thinking
Look beyond today's product cycle without losing contact with reality.
Interdisciplinary by default
Useful ideas often appear where established fields begin to overlap.
Build to learn
A prototype is often a better argument than another hundred slides.
Evidence over narrative
Promising ideas are allowed to fail. Evidence decides what deserves more work.
What we’re building toward
Research should eventually create capability that did not exist before.
TenInvent is beginning with frontier AI questions, but the destination is practical technology. The strongest ideas should survive experimentation, become working systems, and earn the right to be taken further.
That could mean new architectures, persistent intelligent systems, AI-native discovery tools or products that only become possible because the underlying research worked.
New model architecturesReasoning, memory, learning and multimodal intelligence.
Persistent intelligent systemsAgents that can plan, act, adapt and work across longer horizons.
AI-native discovery toolsTechnology that changes how hypotheses are formed and explored.
Products born from researchValidated technical ideas translated into useful real-world systems.
Opening shortly
Technology Partnership Program
Explore frontier technology together.
TenInvent is preparing a Technology Partnership Program for enterprise and consumer technology companies that want to explore emerging AI capabilities, investigate technically difficult problems, or develop new technology ideas with a research partner.
We expect the strongest collaborations to begin with a real question or opportunity — not a pre-written technology brief. The program will create a structured way to explore, prototype and evaluate possibilities together.
Discuss an early partnership
Program details and participation model will be announced shortly.
TenInventresearch
Enterprise technologyComplex problems · scale · domain depth
Consumer technologyNew experiences · adoption · reach
Frontier capabilityAI · agents · discovery
Recent milestones
The public presence is new. The initiative is already active.
TenInvent will keep this record deliberately factual: what has started, what is underway, and what is coming next.
Public research presence launched
TenInvent established its public research identity and initial digital presence.
Initial AI research agenda established
Early research questions are concentrated around intelligence, memory, agency and AI-enabled scientific discovery.
Technology Partnership Program
A collaboration program for enterprise and consumer technology companies will be announced shortly.
Contact TenInvent
Have something worth exploring?
Research collaboration, an unusual technical idea, a strategic partnership, or simply a question that does not fit neatly into an existing category — start there.