01
Measure
Design evaluations that separate meaningful novelty from noise, remixing, and memorized style.
Creative intelligence lab
Machines can answer. Can they imagine?
Vulsar AI is building the science and systems to measure, understand, and improve creativity in AI assistants.
Research focus / 01
Measuring meaningful novelty
The problem
Fluency is not creativity.
AI can produce an endless stream of plausible answers. But volume can hide imitation. Creativity asks for something harder: ideas that are original, useful, coherent, and right for the moment.
Our central question
How do we know when an AI has made a genuine conceptual leap?
Answering that question requires new evaluations, better models of human judgment, and a deeper view of how assistants explore ideas.
Our approach
From intuition to instrument.
We are turning creativity from a vague aspiration into something AI researchers can observe, test, and deliberately improve.
02
Understand
Trace how models explore, combine concepts, revise, and recognize their own strongest ideas.
03
Improve
Build training signals and tools that help assistants produce more original, valuable work.
Research thesis
Creativity is a system, not a spark.
A creative assistant must do more than sample unlikely words. It must navigate a possibility space, recognize what matters, and choose a direction with purpose.
- 01 Novelty
- Does the idea move beyond the statistically comfortable?
- 02 Value
- Is the surprise useful, resonant, or illuminating in context?
- 03 Agency
- Can the assistant explore, critique, and steer its own process?
What we're building
A foundation for creative intelligence.
Our early research agenda connects rigorous measurement with the systems needed to help AI assistants think more expansively.
Evaluation
Creative benchmarks that reward the rare good idea.
Open-ended tasks, structured human judgment, and robust signals for originality and value.
Learning
Models of taste and self-critique.
Teaching assistants to recognize promising directions before we do.
Systems
Agents that explore before they answer.
New workflows for divergence, reflection, selection, and refinement.
Our work is early. The ambition is not.
Start a conversation
Let’s make AI worth wondering about.
We want to hear from researchers, builders, and curious minds who believe creativity is a frontier worth measuring.
Your message is delivered privately and used only to respond to your inquiry.