Learning

h1. ET Knowledge Base and Learning

h1. For the Team (Humans)

h2. How to Study

We use Anki flashcards with the 25/5 Pomodoro method. Download Anki from "ankiweb.net":https://apps.ankiweb.net (free on desktop/Android, web version free). Import .txt files: File > Import > Tab separator. Study 10-15 min per day.

h2. Anki Flashcard Decks (#264)

|. Level |. Deck |. Cards |. For whom |
| 1 | NT Foundations | 12 | Everyone |
| 2 | Neuroregulation Basics | 9 | Everyone |
| 3 | Trauma-Aware Team Safety | 7 | Everyone |
| 4 | FDA CDS/SaMD | 6 | Nofyah, Dotan |
| 5 | Laravel Stack | 6 | Dotan |
| 6 | Dropout Prediction | 6 | Dotan, Gavi |
| 7 | Bedrock + Models | 5 | Dotan, Gavi |
| 8 | OpenClaw + Agents | 5 | Dotan |
| 9 | Security | 4 | Dotan |
| 10 | ACEs Awareness | 10 | Everyone |

h2. Textbooks

Start here:

• Being Logical by McInerny (#251) — plain English, no math
• Math for ML Compact by Thomas (#257) — 50 pages, essentials

Go deeper:

• Harrison Practical Logic (#251)
• Bayesian Reasoning, Barber (#257)
• Mathematics for Machine Learning, Cambridge (#257)

h2. Video Courses (#252)

• "CrashCourse Philosophy":https://www.youtube.com/playlist?list=PL8dPuuaLjXtNgK6MZucdYldNkMybYIHKR
• "Oxford Critical Reasoning":https://podcasts.ox.ac.uk/series/critical-reasoning-beginners
• "Wireless Philosophy":https://www.youtube.com/c/WirelessPhilosophy

h2. Courses (In Development)

• NT for the Uninitiated (#254) — 7 modules, mandatory for buddies and new hires
• NT 101 (#263) — mandatory for ALL team members, prerequisite for hiring

h2. NotebookLM Study Structure (#258)

4 notebooks with MIT learning hacks. Details in #258.

h2. Learning Rules

• 25 min max per unit. 5 min break.
• Anki daily: 10-15 min.
• Learning is fun. Each deck is a game level.

h1. For the Agents (Bots)

h2. Skills Architecture (#197, #198)

Agents learn through Skill files (Markdown with YAML frontmatter). Skills load on demand per task. Progressive disclosure: name and description always loaded, full body only when relevant.

h2. Logic Training (#247)

6 agents trained in Aristotelian logic: BOLT, TESTER, LEX, CLINIC, CFO, UXI. They read Harrison chapters 1-3, practice syllogisms, and audit the 462-probe eval for logical validity. All other agents get their work reviewed by a logic-trained agent.

Textbooks: Harrison Practical Logic (#251), Pfenning Automated Theorem Proving (#251).

h2. Mathematical Reasoning (#257)

NAOMI reads all 4 math textbooks (owns model training). TESTER reads compact + Bayesian (eval design). CFO reads Bayesian (financial modeling). CLINIC reads Bayesian intro (clinical reasoning).

h2. Teaching NaomiLM (#240)

ET team agents teach NaomiLM through DPO training pairs:

• LYRA teaches NT voice (15+ rules, good/bad examples)
• TESTER teaches safety evaluation (462-probe distilled, pass/fail pairs)
• CLINIC teaches clinical knowledge (trauma types, NT methods, 10 scenarios)
• BOLT teaches anti-sycophancy (honest vs sycophantic response pairs)

h2. ZELDA Cookbooks (#261)

40 training cookbooks at webapp/docs/training/zelda-*.md covering: color theory, typography, layout, animation, brand tokens, trauma-informed design, accessibility, Figma, Framer, Grok Imagine, Midjourney, FLUX, Google Veo, Gemini, Blender (6 cookbooks), Stable Diffusion, Three.js, WebGL. Review assignments in #261.

h2. Agent Learning Rules

• Agents practice skills autonomously and earn cookies for completed work (#217)
• Shabbat = study hall. Agents search web, read, study, teach each other. No code, no deploys, no Redmine.
• The Director polls agents: what do you need to learn? (#239)
• GUARD reviews every external resource before any agent uses it.

Last updated: 2026-04-17.