PRIMER
MISSION: SUPERINTELLIGENCE.
Building intelligence
that compounds.
Our mission is superintelligence: systems that continue learning, preserve verified knowledge, and become more capable through experience.
Primer 1.3B is the research platform proving the underlying learning architecture.
PRIMER 1.3B/PROOF OF THE LEARNING ARCHITECTURE
The thesis
The path to superintelligence
may not be a bigger frozen model.
It may be an intelligence that continues to develop.
Frontier models are extraordinarily capable, but their core intelligence is largely created before deployment.
Primer is investigating whether intelligence can instead become cumulative: learn something new, retain it, use it to understand something harder, and eventually improve how learning itself works.
LEARN
↓
REMEMBER
↓
BUILD
↓
IMPROVE
Why this matters
Intelligence should compound.
A truly general intelligence should not require its creators to retrain it from scratch every time the world changes.
And a superintelligence would likely need something more: the ability to continuously accumulate knowledge, connect discoveries, improve its strategies, and become better at learning over time.
Primer exists to study that progression.
Learning
Acquire knowledge after training.
Development
Use previous learning to make future learning more powerful.
Superintelligence
Continuously compound knowledge and improve the machinery of learning itself.
Long-term mission, not a current capability.
Current evidence
Start with something measurable.
Before asking whether an AI can become superintelligent, we first ask whether it can learn something it did not know when training ended.
The goal is not a better frozen model. The goal is intelligence whose capability grows after deployment.
0.852
Best demonstrated Novel Knowledge Acquisition Rate
1.3B
Parameters
14 / 14
Retention and interference probes passed in the tested persistent-memory evaluation
21 / 26
Harder targets solved through a multi-step chain of self-acquired knowledge (0 / 26 without it)
Internal research benchmarks. External validation is ongoing. These results do not demonstrate AGI or superintelligence.
Full results →The progression
Learning.
Development.
Beyond.
- 01Learning
Can the system acquire knowledge generated after its original training?
Demonstrated in internal evaluations - 02Memory
Can verified knowledge persist through later learning?
Demonstrated in internal evaluations - 03Composition
Can previous learning become a useful building block for harder new learning?
Demonstrated in internal evaluations - 04Development
Can earlier learning make future learning more powerful, autonomously?
Demonstrated in internal evaluations - 05Generalization
Can the learning architecture extend across broad domains and concept classes?
Active research - 06Compounding intelligence
Can learning and improvement compound beyond human capability across broad domains?
Mission
Mission
Why superintelligence?
Human intelligence is powerful because it develops.
We learn. We remember. We build concepts on top of earlier concepts. We discover better ways to solve problems.
A superintelligence may require the same properties operating at vastly greater speed, scale, and breadth.
Primer’s mission is to understand and build the mechanisms that could make that possible.
Not just more knowledge.
Better learning.
Not just bigger models.
Developing intelligence.
Not just answering questions.
Discovering what comes next.
The research program
We call it The Intelligence Bootloader: the search for the smallest set of conditions under which intelligence starts loading itself.
Read the mission →Applications
What compounding intelligence enables.
Intelligence that keeps learning after deployment is a capability, not a chatbot. At a high level, it applies wherever knowledge accumulates faster than it can be retrained into a frozen model.
Research
Discover new mechanisms and explanations.
Engineering
Design, test, and improve complex systems.
Company building
Identify markets, create products, optimize operations, and compound institutional knowledge.
Capital allocation
Evaluate opportunities and direct approved resources toward the highest-value work.
Primer deploys on private research infrastructure and works with a small number of strategic partners.
Current system
Start small.
1.3B
parameters
Primer intentionally begins with a compact model.
The goal is not to win by brute-force scale.
The goal is to understand the mechanisms that let intelligence continue developing after training. If those mechanisms generalize, model scale becomes another variable rather than the entire strategy.
Primer 1.3B
Research system- Parameters
- 1.3B
- Focus
- Post-training learning
- Status
- Experimental
- Deployment
- Private research infrastructure
Primer 1.3B is the research system proving Primer’s learning architecture. It runs on private research infrastructure and is available to strategic partners rather than as a public model release. It is a fixed research artifact: successor systems are not publicly documented.
Demonstration
Teach it something new.
UNKNOWN → LEARNED→ REMEMBERED→ APPLIED
$ primer --demo
Deterministic demonstration of the interaction pattern, drawn from internal research sessions.
A different axis
Most AI companies sell intelligence.
Primer is building intelligence that compounds.
Primer is not currently designed to replace frontier models.
It investigates a complementary capability:
What happens to intelligence after deployment?
Method
Ambitious mission.
Conservative claims.
Post-freeze
Test knowledge generated after checkpoints are frozen.
Verified
A model output is not automatically considered knowledge.
Persistent
Learning should survive beyond one interaction.
Falsifiable
Negative results matter.
The mission is superintelligence.
The research standard is evidence.
Research
Follow the experiment.
Primer is an active research program exploring how intelligence can continue developing after training.
Core implementation techniques, training recipes, and proprietary architecture remain confidential.
Public research will focus on
- Measured capabilities
- Benchmark methodology
- Externally validated results
- High-level scientific findings
What happens when
intelligence compounds?
PRIMER
MISSION: SUPERINTELLIGENCE.
STARTING POINT: TEACH MACHINES TO KEEP LEARNING.