Building the First
Continual Megapolis
Builders behind the Continual Technological Systems
Building Continual Machine Intelligence as systems that operate indefinitely
Monte Lua is our game product powered by MDL, a universe simulation engine built to keep worlds, state, and character dynamics evolving over time instead of collapsing around a fixed interaction window.
The product is not separate from the research. It is one of the places where we pressure-test what continual operation should mean in practice: persistent simulation, long-horizon coherence, and agents that can continue running instead of resetting.
If we want machine intelligence that does not fail at fixed context limits, we need product surfaces where continuity is a core requirement. Monte Lua gives us one of those surfaces.
Machine Intelligence that can build, inspect, and continually control real hardware
The platform turns supported STM32 and ESP32 boards into host-powered machine intelligence surfaces where agents can interact with low-level electronics instantly through high-level scripts and native UI.
The same runtime that lets a user wire up sensors, motors, and actuators can also let agents remotely control that hardware autonomously, observe feedback, and continue operating without the usual build-and-flash loop.
Architecture work for continual operation
MGPT, IMGPT, and CUA are the main research surfaces behind Continual MI. These are the systems we are building toward machine intelligence that can keep operating instead of failing at fixed context limits.
A successor architecture to GPT
MGPT stands for Mask-Generative Pretrained Transformer. It is a new architecture built around assistant-controlled masking, summarization, and sink-preserving system rotation so context can be maintained for far longer than ordinary GPT-style flows.
This is the first research surface in the stack, and the one with a demo path available now.
Infinite MGPT for stable indefinite operation
IMGPT keeps MGPT's masking workflow, but changes how masked history is handled in positional space. The aim is effectively unbounded operation through position-aware collapse rather than naive context growth.
The core direction is to bound mask operations and stabilize RoPE rotations so long-running operation stays learnable and stable.
Applied computer-use experiments
CUA is the applied experiment layer for computer use and operational agency. It gives Continual MI a place to test how future model architectures can interact with real software environments over time.
It is the practical bridge between architecture research and systems that can act continuously in the world.
Continual Society & School
Learn to build the systems that will power the first fully autonomous city on planet Earth. No prerequisites required - anyone can start their journey here.
The Continual School leverages our custom, never-seen-before Continual EMWaver platform - the first of its kind, developed in-house by Continual as an educational development platform that teaches hardware interfacing, software, and firmware development fundamentals, then advances to high performance systems and culminates in research and your first bleeding-edge research contributions as a part of the Continual Society.
Complete Learning Journey
1. Continual EMWaver Platform
Low performance hardware, software & firmware bridge
2. Continual Machine Intelligence Systems
High performance software, hardware, and machine learning systems development
3. Continual Research & Methodology
Machine learning research paths and guidance
4. Continual Society Publication
Your first published research contribution under the Continual Society
Become a member of the Continual Society
Become part of an exclusive community platform with builders, researchers, and visionaries working together to create the technologies that will power the first autonomous megapolis. Includes full access to Continual School.
Community Society platform + Continual School
Join the Continual MI Team
After completing the Continual School curriculum, priority is given to graduates to join the actual Continual MI development team.
Work directly on building the systems that will power the first autonomous megapolis, from infinite-context models to the machines that will construct our future.
Team Opportunities
Research & Development
Machine learning architecture research and implementation
Systems Engineering
High performance systems and machine learning infrastructure
Hardware Development
Custom hardware for autonomous systems
Curriculum Development
Creating and updating Continual School machine learning course content
Education & Teaching
Leading courses and mentoring students in the Continual School
Megapolis Construction
Building the first autonomous city systems
🎓 School graduates receive priority consideration for team positions