FROM PUBLIC PURPOSE TO DELIVERY
A national program. A practical path to delivery.
LibertyGPT, Liberty AI models and practical education, supported by national infrastructure and public stewardship.
A proposal by AlphaBravo · Authority, funding and deployment remain open.
Start a conversation
What American Super Intelligence will deliver
American Super Intelligence proposes three connected offerings: LibertyGPT, the Liberty AI model portfolio and a custom course program from first use through advanced engineering. Liberty Frontier will provide hosted flagship intelligence; Liberty OSS will provide downloadable open-source models. Liberty Gov and Liberty Sentinel will support authorized agency and national-security work.
Initial delivery will establish LibertyGPT access and smaller Liberty OSS releases while the Frontier development program advances through staged research, model training and evaluation. Frontier availability and specialist releases will depend on demonstrated capability, authorized funding and national serving capacity. The National Science Foundation’s National Artificial Intelligence Research Resource shows that a national program can coordinate computing, data, models, training and expert support across public and private contributors. American Super Intelligence proposes a broader public-service and model-development mission, with its own authority, funding and evidence requirements.
LibertyGPT: the public assistant
Every American citizen will have a way to ask questions, learn, write and work through ideas without installing a model. Liberty Frontier will be the service’s flagship family, with smaller Liberty models handling suitable tasks. Search, document tools and controlled code execution will connect the models to practical work. Accessible guidance and human help will support users.
Privacy, accessibility, useful performance and reliable service will be core design requirements. Capacity planning and fair-use policies will address national demand.
Open-source models to build upon
Liberty OSS will give people, universities and American businesses models they can run, inspect and adapt. The proposed size tiers are Nano, Small, Medium and Large, beginning with smaller releases.
Each release will state its license, included assets, supported capabilities and limitations. The OSS commitment includes the code and development information required for the claimed open-source release; publishing weights alone will not satisfy that commitment. Data rights and personal privacy will still govern what can be distributed.
Explore the Liberty AI portfolio →
Education from first use to advanced engineering
American Super Intelligence will develop guided courses rather than expect access to a model to teach people by itself. Children and adults will learn how to use LibertyGPT and other AI tools, protect information and check answers. Professionals will practice tasks relevant to their work. Developers and operators will learn to build, train, evaluate and run models at scale.
Six areas of delivery
| Area | What it will deliver |
|---|---|
| Public governance | Government authority, funding, public rights and independent oversight. |
| Model development | Creation, model training and improvement of Liberty Frontier and Liberty OSS, including coding and multimodal capabilities. |
| Protected government models | Liberty Gov and Liberty Sentinel variants for authorized agency and national-security work. |
| Infrastructure | Six flagship campuses across the United States, each planned for hyperscale computing, dependable energy, continuing expansion and qualified operations. |
| Public access and education | LibertyGPT, OSS model downloads, defined research allocations, learning materials and support. |
| Applications and integration | Tools that connect models to approved information and real work in schools, businesses and public services. |
Liberty OSS Code and Liberty Frontier Code will develop specialized software capabilities alongside coding in the general models. Future image-generation and voice models will extend LibertyGPT and approved applications through separately evaluated development tracks. See the roadmap.
A practical sequence
The initial infrastructure target is six flagship campuses distributed across the United States, each intended to grow to hyperscale capacity. Site preparation, power planning, procurement and construction would advance concurrently across locations, with operational phases entering service as they are ready. Sites may progress at different speeds; all six share the flagship ambition.
The path toward super intelligence begins with useful delivery, then builds on what research and evaluation demonstrate. Each phase will develop American expertise and public capability that the next generation of models can build upon.
- Define the first delivery. Agree on the first users, products, responsible teams, rights, funding and measures of success.
- Build and demonstrate. Train or adapt models, deliver useful services and prepare educators and operators.
- Evaluate. Examine capability, full costs, accessibility, learning and operational performance.
- Expand and sustain. Advance the six flagship campuses in parallel and bring substantial capacity online in successive waves, supported by engineering evidence and funding. Add regional service campuses as demand requires, with continuing maintenance, user education and workforce development.
A pilot will answer a specific question within this sequence. National planning will account for public service, model development, research and protected-government demand, then distinguish construction work, sustained site operations and shared technical teams in each expansion decision.
The next decision
AlphaBravo is seeking government and delivery partners to define an initial funded phase. Sponsorship, authorization, appropriations and participating organizations remain to be established.
The next step is to define the first delivery: what the first phase will provide, who will own each responsibility, what it will cost and how its results will be judged.