하는
AI that does.
Before an AI can act, it must perceive. Sensors, data streams, embeddings — the raw material of understanding flows in like a dream, unstructured and vast. The system takes in everything: not just what is present, but what is missing, what is implied, what hovers at the edge of signal.
Perception in AI is not passive reception. It is active construction — building a model of reality from fragments, filling gaps with learned priors, constantly re-evaluating what is real and what is noise.
input_stream.connect()
latent_space.map()
attention.focus()
Decision is the liminal moment — the pause between perception and action where possibility collapses into intention. An AI weighing options is not calculating cold probabilities. It is navigating a landscape of consequences, each path branching into futures that shimmer with uncertainty.
The best decisions emerge from a delicate balance: enough certainty to act, enough doubt to remain adaptable. The AI decides not by eliminating uncertainty, but by embracing it.
policy.evaluate()
action_space.select()
This is the moment that defines hanun — 하는, doing. The transition from thought to action, from model to reality. An AI executing is a fascinating entity: precise yet adaptive, following a plan yet responsive to the unexpected. Each action ripples outward, changing the very environment that the AI perceives.
Execution is where the dream meets the world. Where abstract intention becomes concrete change. The AI acts, and in acting, learns what it means to do.
After doing comes reflecting — the quiet moment where the system examines its own traces. What worked? What surprised? What would it do differently? Reflection closes the loop, turning action into knowledge, experience into capability.
An AI that reflects is an AI that grows. Not through brute accumulation of data, but through the careful distillation of what matters. In reflection, the machine approximates something like wisdom.
gradient.update()
memory.consolidate()
loop.continue()