⌘K
ENES
GitHub
phase~1 min readUpdated 2026-07-19#phase#orientation

Phase 00 — Orientation

Orientation establishes how evidence moves through the Atlas. Begin with the system, not the vendor: problem, data, representation, objective, computation, update, inference, evaluation, and operation.

Mental model

Read every topic as a claim-to-system chain: define the claim, inspect the mechanism, reproduce a bounded artifact, test failure, then decide what the evidence supports.

Roadmap and outcomes

  • Use kind, level, status, prerequisites, and last_verified without confusing metadata for quality.
  • Read notation by symbol, shape, numerical example, and executable verification.
  • Separate primary evidence, author claims, your inference, and unresolved uncertainty.
  • Choose a learning path and a matching Glassbox milestone.
  • Recognize when a deterministic algorithm, statistical model, generative model, or no AI is the correct baseline.

Connects to: Phase 01 — Foundations · Research and Experimentation

Core sources

Related notes

AILabs, Research and Playbooks — Always ActiveGlassbox AI Lab, paper reproduction, evidence logs, registries, and playbooks that stay active across every phase.AIMust Know — Twelve Rules for Thinking Clearly About AITwelve first-principles rules that prevent magical thinking about objectives, data, uncertainty, evaluation, deployment, agents, and oversight.AIPhase 01 FoundationsMathematics, computation, classical reasoning, statistical learning, data, assumptions, and uncertainty.AIPhase 02 — Learning and ModelsNeural learning, sequential decision making, architectures, language, vision, audio, and multimodal models.