Unified façade over MathNN attention/curiosity, RL intrinsic motivation, and cognitive-drive ODEs.

Stack

[text]
Salience (candidates) → Attention (curiosity-biased) → Focus top-k
        ↑                      ↑
   weather/hazards      β_eff from drives + body
Curiosity (count + episodic mem) → intrinsic reward / shape
Drives ODE: attention, arousal, curiosity, fatigue, confidence

Commands

[text]
layers status|doctor|metrics|selftest
layers mode task|explore|safety|social
layers capacity 5 | beta 1.5
layers link body on|off | link drives on|off
layers seed | candidates | clear | add <id> <kind> [sal]
layers focus [query]
layers cycle [novelty] [success] [query]
layers drives | drives step <nov> <suc> | drives reset
layers curiosity bonus <key> | shape <r> <i>
layers attn-demo | curiosity-demo
layers rl <agent_id>
attention focus|mode|capacity|demo
curiosity status|beta|bonus|reset|demo|mode
layerskit status|selftest|focus|cycle

Coupling

Input	Effect
Mode explore	Higher β, novelty bias
Mode task	Goal relevance bias
Mode safety	Safety candidates prioritized; low β
Drive curiosity/fatigue	Scales beta_eff
Body restorative/recover/safe	Suppresses exploration β
Body equilibrium	Calls layers cycle

Building blocks reused

• _Attention, _CuriosityAttention (MathNNLayers)
• _IntrinsicCuriosity (RLExplore)
• _CognitiveDrives (MathDynamics)
• _RLAgentNet::SetCuriosityScale via layers rl
• Weather storm risk as optional salience

Self-test

[text]
layers selftest

---

Filed under: Uncategorized - @ July 18, 2026 6:25 am