Building AI with intuition.
Frontier labs are scaling reasoning. We're teaching machines the other half of intelligence: judgment, instinct, the sense that comes before the logic.
Not another frontier model. A different kind of intelligence.
The Thesis
Scale isn't the whole answer.
Every major lab is racing toward the same horizon: bigger models, longer chains of thought, more compute. The bet is that intelligence emerges from brute force deliberation. System 2 thinking, scaled to infinity.
But human cognition doesn't work that way. Most of what we call intelligence, recognizing a face, sensing when something is off, knowing which move to make, happens in milliseconds without explicit reasoning. Kahneman called it System 1. We call it intuition.
Astryn Labs is exploring this other half. Not by making models think harder, but by teaching them to sense. To develop fast, implicit judgment that complements deliberate reasoning. The goal isn't a bigger brain. It's a more complete one.
Our Approach
What training intuition actually means.
Three research pillars that translate cognitive science into machine learning. Concrete, measurable, and built for the real world.
Pattern sense
Training models to recognize deep structural patterns without explicit feature engineering. The kind of perceptual fluency that lets a chess grandmaster see a winning line in a glance.
Implicit judgment
Building decision systems that weigh context, uncertainty, and prior experience in a single forward pass. Not by enumerating options, but by arriving at the right answer the way intuition does.
Fast heuristics
Developing lightweight cognitive shortcuts that trade perfect accuracy for speed and adaptability. Enabling AI to operate in real time, under pressure, with incomplete information.
Who We Are
A small team asking a big question.
Astryn Labs is a research group exploring whether machines can develop intuition. Not through more compute, but through a different kind of learning altogether.
We're early, we're focused, and we're working on problems most labs aren't looking at yet. No product roadmap. No growth metrics. Just rigorous research on a question we think matters.
Current focus
- Cognitive architectures
- Implicit learning systems
- Intuition in machine models
- Early stage research
What We Believe
A manifesto for a different kind of AI.
These aren't slogans. They're the principles that guide every experiment, every architecture choice, every hire.
- 01
Intelligence is not one thing. Reasoning and intuition are complementary systems, and AI needs both.
- 02
The best models won't be the biggest. They'll be the most complete: fast where speed matters, deliberate where it counts.
- 03
Cognitive science isn't marketing. System 1 and System 2 isn't a metaphor. It's an engineering blueprint.
- 04
Research should be honest. No fake metrics, no vaporware demos. Just rigorous work on a hard problem.
- 05
The companies that win won't just scale compute. They'll understand what kind of intelligence they're actually building.