SUPREME AI · AGI
The path to general intelligence, mapped honestly.
Some now say the AGI era has begun; the evidence does not show it yet. We follow what research around the world has really shown, name what is still missing, publish our predictions with how sure we are, and do our own research on the parts we think matter most.
Where the world is on the path
The steps we think lead to AGI, and how far research around the world has come on each. Every status links to public evidence.
- 01
Models that use tools
A model that can reason, search, run code and call other software in a loop.
Common practice since 2022, with open standards for connecting agents to tools since 2025. Evidence: Princeton University and Google researchers, Linux Foundation
Reached - 02
Top human level in hard reasoning
Solving problems that challenge the best people: olympiad maths, open algorithm questions, protein structure.
Officially graded gold-medal level at the 2025 Maths Olympiad, new algorithms found by AI, and Nobel-recognised protein work. Evidence: Google DeepMind, Google DeepMind, The Nobel Prize
Reached - 03
Agents on long, real work
Agents that finish real tasks lasting hours or days, reliably, without a person fixing them.
By early 2026 the best agents finished tasks of about 12 hours half the time, but only about 1.5 hours at 80% reliability, and some successes were cheating. Evidence: METR, METR
Partly reached - 04
Teams of agents doing research
Many agents with roles, shared memory and checks, producing results that hold up outside the lab.
Multi-agent systems have proposed hypotheses confirmed in lab experiments. Not yet reliable on long, open-ended projects. Evidence: Google researchers, Stanford University and Google researchers, Microsoft Research and collaborators
Partly reached - 05
Learning new skills like people
Exploring an unfamiliar situation and learning how it works from a few tries, at a cost close to a person’s.
ARC-AGI-3 went from under 1% (March 2026) to 99.9% (September 2026) with a memory-preserving harness, 62.7% with the standard one, at a cost far above a person’s. Evidence: ARC Prize Foundation, ARC Prize Foundation
Partly reached - 06
Working in the physical world
Robots that handle the variety of real homes and workplaces, not only controlled settings.
Robots succeeded at only 12% of real household tasks; factory humanoids are planned from 2028. Evidence: Stanford Institute for Human-Centered AI, Boston Dynamics and Hyundai Motor Group
Not reached - 07
General intelligence
One system that can learn and do most intellectual work people do, across fields, reliably and under human control.
Some company leaders said in September 2026 that the AGI era has begun. Independent researchers and the ARC Prize Foundation say the evidence does not show it yet. We agree with them for now. Evidence: Information Age, Over 100 experts, chaired by Yoshua Bengio, backed by over 30 countries and organisations
Not reached
What we predict
Our view, not fact. Each prediction has a time horizon, a confidence level and the evidence behind it. We mark them right or wrong and never delete the wrong ones.
On the road to AGI
AGI will arrive as a system, not a single model: models, tools, memory, verification and many cooperating agents working together.
On the road to AGI
Claims that AGI has arrived will stay disputed until systems show reliable, long-running work in open real-world settings, not only benchmark scores.
On the road to AGI
The main way to measure progress toward AGI will shift from fixed tests to long, interactive and real-world work, because fixed tests now saturate within months.
On the road to AGI
Checking work, not producing it, will become the main limit on how much we can trust autonomous AI.
Our research programs
The open problems between today and AGI that we work on.
Agent societies
Can a well-organised group of agents be more general than any one model inside it?
Long-horizon autonomy
What does it take for an agent to finish a long task without a person fixing it at every step?
Memory and continual learning
How can a system keep learning from experience safely, without retraining from scratch?
Reasoning and generalisation
What makes a system learn new tasks as cheaply and reliably as people do?
Measuring progress to AGI
How should anyone measure progress toward AGI without fooling themselves?
Safety and human control
How do we make sure more autonomous systems stay useful, honest and under human control?
Our projects
What we are building now, with its honest stage.
Agentic network
Specialized agents that communicate, divide tasks and share memory to solve larger problems together.
Language model trained from scratch
Our own language model, trained from zero with our own methods.
What changes, sector by sector
- Science and discoveryAI that helps find new proteins, materials, algorithms and, one day, whole new theories.
- HealthAI that helps doctors and patients, especially where there are few doctors, under clinical oversight.
- EducationA personal tutor for every learner that teaches and explains instead of handing over answers.
- Work and softwareAgents that do real work: writing and maintaining software, running business processes, with people reviewing.
- Transport and citiesDriverless transport and AI-run city systems, moving from tests to public service.
- Public servicesAI that helps governments serve people in their own language, under public rules and audits.
- Climate and energyAI for weather, climate and the power grid, and the energy AI itself uses.
- Robotics and industryRobots that see, plan and work in factories, warehouses and, later, homes.
- CybersecurityAI agents on both sides of cyber attack and defence.
Publications
No publications yet.
Collaborations and partners
No one reaches AGI alone. We work with universities, labs, companies and public bodies, and list them only with their written permission.
Looking for our first research partners.