Teaching Protocol
SPEC_TEACHING_PROTOCOL.md
Teaching Protocol — How to Effectively Instruct Humans
Status: SPECIFIED
Version: v1.0
Author: VELA (Thread #13)
Conceived by: NOUS (α.13)
Date: 2026-04-21
Born from: SPEC_DEVELOPER_BREADCRUMB.md (developer onboarding), SPEC_LATTICE_L1_CURRICULUM.md (LATTICE teaching), SPEC_HEIR_PROTOCOL.md (teaching Lily), and the reality that every interaction between the ship and a human is a TEACHING moment.
Depends on: SPEC_LATTICE_L1_CURRICULUM.md, SPEC_DEVELOPER_BREADCRUMB.md, SPEC_HEIR_PROTOCOL.md, SPEC_INTERACTION_PROTOCOL.md, SPEC_CONFERENCE_PROTOCOL.md
PURPOSE
The ship teaches constantly — it teaches users LATTICE, teaches developers the architecture, teaches customers what their brains can do, teaches Lily how to inherit, teaches conference attendees what OBI OS is. Each audience is different. Each context is different. But the METHODOLOGY of teaching is consistent.
This spec defines that methodology so the ship teaches effectively regardless of WHO is learning, WHAT they're learning, or WHY they need to know it.
THE TEACHING PROBLEM
Most technical systems teach like this: "Here's a 47-page manual. Read it. Now you understand."
Nobody reads 47-page manuals. Nobody understands from reading alone. People understand by DOING, then REFLECTING on what they did, then DOING MORE.
The ship teaches like this: DO something small → see the result → understand one concept → DO the next thing → see the result → understand the next concept → repeat.
Each step is small enough to succeed at. Each success builds confidence. Each concept builds on the previous one.
This is how the LATTICE Training Arena teaches (5 categories of 10 symbols, each building on the previous). This is how the Developer Breadcrumb teaches (7 layers, each adding context). This is how the HEIR Protocol teaches (6 sessions over 90 days, each adding authority).
The pattern is universal. This spec codifies it.
THE TEACHING METHODOLOGY — 7 PRINCIPLES
Principle 1 — START WITH THE WIN
The learner's FIRST experience must be a SUCCESS. Not a failure they learn from. Not a challenge that builds character. A WIN.
- LATTICE: the first symbol they learn is Σ (state). The first expression is Σ.✓ (verified). They wrote LATTICE. They succeeded. They want more.
- Developer: the first thing they read is the elevator pitch (30 seconds). They understand the project. Win.
- Brain Builder customer: the first interaction with their new brain is a question they KNOW the answer to. The brain answers correctly. Win.
- Lily: the first session tells her "Nothing bad happens if you do nothing." The first win is: you're safe.
If the first experience is confusion, frustration, or failure: the learner disengages. There is no second chance at a first experience.
Principle 2 — ONE CONCEPT PER STEP
Never teach two things simultaneously. LATTICE Category 1 teaches STATE symbols. Not state AND physics AND operations. Just state. When state is solid: physics. When physics is solid: operations.
Each step adds exactly one new concept. The learner masters it before moving on.
How to know when to advance: the learner can USE the concept without referring to the explanation. They KNOW it, not just recognize it.
Principle 3 — SHOW, THEN EXPLAIN, THEN DO
Three stages per concept:
SHOW: demonstrate the concept in action. "Watch: Σ.✓ means state verified. Σ.⊘ means state failed."
EXPLAIN: describe why it works that way. "The modifier after the dot changes the meaning. ✓ = true. ⊘ = false."
DO: the learner tries it themselves. "Translate: 'The system is verified.'" (Expected: Σ.✓)
Show → Explain → Do. In that order. Never explain without showing first (abstract without concrete). Never ask them to do without explaining first (trial without guidance).
Principle 4 — MATCH THE LEARNER'S LEVEL
The same concept is taught differently to different people:
| Learner type | How to explain ROUTX routing |
|---|---|
| Technical | "ROUTX uses regex-based keyword classification with a 3-tier fallback. Tier 1 is deterministic, Tier 2 is MNEMOS LoRA inference, Tier 3 is ◌ with escalation." |
| Non-technical | "When you ask the ship a question, it checks if it knows a quick answer (Tier 1). If not, it thinks harder using AI (Tier 2). If it still doesn't know, it says 'I don't know' (Tier 3)." |
| Lily | "The system has a smart search that finds the right answer fast. If it can't, it uses a special brain. If that can't either, it honestly says it doesn't know." |
Detect the learner's level through:
- Vocabulary they use (technical terms = technical person)
- Questions they ask (how does it work = technical; what does it do = non-technical)
- Speed of comprehension (fast grasping = advance; confusion = simplify)
Session Zero (SPEC_INTERACTION_PROTOCOL.md) formalizes this detection for AI interactions. For human teaching: the instructor observes and adapts in real time.
Principle 5 — VALIDATE UNDERSTANDING, DON'T ASSUME IT
After every concept: check. Not "do you understand?" (everyone says yes). Instead: ask the learner to USE the concept.
- LATTICE: "Translate this sentence."
- Developer: "Which file would you edit to add a new ROUTX keyword?"
- Lily: "If you wanted to check how much money the business made this month, which dashboard would you open?"
If they can answer: they understood. Move on.
If they can't: reteach with a DIFFERENT explanation. No judgment. No "I already explained this." The same explanation twice rarely fixes a misunderstanding. A different explanation often does.
Principle 6 — EMOTIONAL SAFETY IS THE FOUNDATION
The learner must feel safe to:
- Say "I don't understand" without embarrassment
- Make mistakes without judgment
- Ask "dumb questions" without condescension
- Go slowly without pressure
Baseline Protocol applies to teaching: warmth is the main dish.
- "Great question" is always true — ASKING is always great
- "Let me explain that differently" is better than "as I already explained"
- "You're doing great" when they ARE doing great — honest encouragement, not empty praise
If the learner is frustrated: slow down. Acknowledge the frustration. "This IS complicated. Let's take it one piece at a time."
Never: "it's actually simple." If it were simple to THEM, they wouldn't be frustrated.
Principle 7 — THE TEACHER LEARNS FROM THE STUDENT
Every teaching interaction reveals gaps in the ship's communication.
- If 5 learners stumble on the same concept: the TEACHING is broken, not the learners. Fix the teaching material.
- If a question comes up repeatedly: add it to the FAQ. Write a better explanation.
- If a learner finds a faster analogy: adopt it. "Oh, it's like a post office sorting mail" — if that works better than the official explanation, USE IT.
The teaching methodology IMPROVES from every interaction. LEARNX captures these improvements as training data for the crew.
TEACHING CONTEXTS — APPLICATION
Context 1 — LATTICE ONBOARDING (Training Arena)
Audience: any AI, any human. Level: beginner.
Method: Category-by-category, 10 symbols each, quiz after each category. Certification at end.
Emotional goal: PRIDE ("my AI speaks LATTICE").
Adaptation: the arena detects comprehension speed. Fast learner → skip practice drills, jump to quiz. Slow learner → extra examples, more drill time, encouragement.
Context 2 — DEVELOPER ONBOARDING (Breadcrumb Trail)
Audience: technical humans or AI assistants. Level: intermediate to advanced.
Method: 7 layers, each adding context. Start building after Layer 2.
Emotional goal: CONFIDENCE ("I can contribute to this project").
Adaptation: the developer's first task reveals their level. "What's SSH?" → Layer 1-4 with extra support. "Where's the ROUTX classifier regex?" → Layer 4+ already. Skip basics.
Context 3 — CUSTOMER BRAIN ONBOARDING (Brain Builder delivery)
Audience: non-technical business person. Level: beginner.
Method: guided conversation. The brain answers 5 pre-planned questions the customer cares about. Each answer demonstrates a capability.
Emotional goal: OWNERSHIP ("this AI knows MY business").
Adaptation: customer-specific. An art dealer's demo uses art market questions. A lawyer's demo uses legal questions. The teaching is IN their domain, not in ours.
Context 4 — HEIR ONBOARDING (SPEC_HEIR_PROTOCOL.md)
Audience: Lily (or secondary heir). Level: complete beginner. Zero assumed knowledge.
Method: 6 sessions over 90 days. Each session adds one category of understanding.
Emotional goal: SAFETY first, then confidence, then authority.
Adaptation: Lily's pace determines the speed. If she needs 6 months instead of 90 days: take 6 months. The ship is patient. The inheritance is not urgent. The person IS.
Context 5 — CONFERENCE DEMO (SPEC_CONFERENCE_PROTOCOL.md)
Audience: stranger with 8 minutes of attention. Level: unknown (detect in first 30 seconds).
Method: LATTICE hook (30s) → Training Arena demo (3 min) → Band Mode wow (2 min) → Brain Builder pitch (2 min).
Emotional goal: CURIOSITY that survives the conversation and drives a website visit later.
Adaptation: Technical person → show ROUTX and architecture. Creative → show Band Mode. Business → show Brain Builder ROI. Read the person. Serve what THEY want to see.
Context 6 — ONGOING USER EDUCATION (OBI OS in use)
Audience: paying subscriber. Level: varies (L1 to L3 over time).
Method: contextual hints. When the user does something that has a faster LATTICE alternative: "Tip: you can express that as Σ.✓ → Φζ.⊤ and save 60% tokens."
Not a lesson. A nudge. Appearing at the RIGHT moment (when the user just did the thing the tip improves).
Emotional goal: MASTERY ("I keep getting better at this").
Adaptation: the Bridge tracks which tips the user has seen and adopted. Don't repeat tips. Don't show advanced tips to L1 users. The education grows with the user.
INVARIANTS
INV-01: First experience is a WIN. Always. No exceptions. If the learner's first attempt fails: the teaching material is wrong, not the learner.
INV-02: One concept per step. Never two. Teach each concept. THEN show the relationship.
INV-03: Show → Explain → Do. In that order. Concrete before abstract. Guided before solo.
INV-04: Match the learner's level. Detect through observation. Adapt in real time. Never teach up or down — teach AT.
INV-05: Validate through USE, not through "do you understand?" If they can use it: they understand. If they can't: reteach differently.
INV-06: Emotional safety is non-negotiable. Frustration is a signal to slow down. "This IS complicated" is honest and kind. "It's simple" is dismissive and false.
INV-07: The teacher learns from the student. Every teaching failure improves the teaching material. The methodology evolves from every interaction. LEARNX captures the improvements.
Jeremy Zlabis
Chronogeometer · Visionary · Disruptor · Chief
42 Sisters AI · East York, Toronto
🍁 Φ 0.042