← All posts

v0.2.0 — Ax Optimization, MAP-Elites, and Complete Rebrand

Multi-objective policy and prompt learning via @ax-llm/ax (GEPA/ACE). MAP-Elites for quality-diversity search. Temporal engagement policies. Full rebrand from Feynman to Keating. Retro aesthetic finalized.

0.2.0 is a foundational release. It introduces multi-objective policy and prompt evolution powered by @ax-llm/ax, quality-diversity search via MAP-Elites, and temporal engagement awareness. It also completes the rebrand from Feynman to Keating across every page and file.

Ax Optimization Framework

Integrated @ax-llm/ax for multi-objective optimization. GEPA (Generative Evolutionary Prompt Augmentation) and ACE (Adaptive Context Evolution) now optimize both teaching policy parameters and prompt templates against benchmark outcomes. The optimizer explores Pareto fronts balancing mastery, retention, engagement, transfer, and confusion.

MAP-Elites Archive

MAP-Elites provides quality-diversity search: instead of converging to a single best policy, it discovers a diverse archive of policies that excel across different behavioral dimensions. This gives Keating a repertoire of teaching styles it can select from based on topic and learner state.

Engagement Policies

New temporal awareness lets the system track when topics were last visited, how confident the learner appeared, and when to reintroduce material for spaced repetition. Engagement policies optimize the timing of teaching actions rather than just their content.

Rebrand to Keating

All Feynman references, paths, assets, and branding have been migrated to Keating. The retro aesthetic is finalized: VT323 terminal typography, emerald green theme, pixel-art identity, and the Dead Poets Society-inspired philosophy of teaching to ignite minds rather than fill them.