Levels 3 and 4

Orchestrate first. Scale later.

The lab path teaches students how frontier models, local LLMs, and dedicated AI machines fit together. Students reach this level after they understand Codex, the harness, sequencing, GitHub, and deployment.

Screenshot explaining frontier model orchestration and local LLM fleet compute — The architecture: frontier model for judgment, local or lab compute for high-volume inference.

Frontier modelPlans, reasons, writes code, reviews, and verifies.

Local LLMProcesses bulk summarization, extraction, classification, and drafts.

Lab fleetDedicated AI MacBooks available to students who are ready.

Local agentsHermes, OpenClaw, and personal operating environments.

Why this matters

Subscription limits teach strategy.

Students should learn which work needs the best frontier model and which work can move to cheaper compute. That judgment matters more than memorizing the current leaderboard.

Use Codex for orchestration

Planning, coding, deployment, project state, and verification remain with the frontier agent.

Move bulk work out

Summaries, classifications, extraction, and repeated transformations are candidates for local or lab LLMs.

Record the routing

Students document what ran where, what it cost, and how the output was checked.

Later invitation

Lab access comes after discipline.

lab.counselorjay.com is introduced after students have a real project pipeline. The workshop should make the lab feel earned: students first prove they can create portfolio evidence, verify work, and maintain project memory.