# Understanding AI & LLMs — Course Package **Workshop:** Monday 18 May 2026 · 2.5 hours · live online **Facilitator:** Florin Bădiță (ai-courses.badita.org) **Audience:** Mixed-experience adults — no coding required, no prior AI assumed **Tiers:** Community ($29) · Coaching ($129) **Package version:** 1.0 · build date 17 May 2026 --- ## What this package is Six artefacts that turn the existing 35-slide impress.js deck into a **results-driven** workshop — meaning participants leave with something they can use Monday afternoon, not just notes they'll never reopen. The package is opinionated about three things: 1. **2.5 hours is short.** Every slide must earn its place. Anything that doesn't change what participants *do* gets cut or moved to pre-work. 2. **AI moves faster than slide decks.** The current deck cites GPT-4o, Claude 3.5, Gemini 1.5 — all of which are now 2–3 generations old. The course design includes a fact-check audit with specific patches needed before Monday. 3. **Results = behaviour change, not knowledge.** Success is measured by what participants *do* in the 30 days after, not by what they could repeat back at the end. --- ## File map | File | Audience | Purpose | |---|---|---| | `README.md` | Facilitator | This file. Orientation + how to use the package. | | `01_course_design.md` | Facilitator | Strategy bible. Outcomes, audience, pedagogy, fact-check audit, risk register, tier strategy. **Read first.** | | `02_slide_deck.md` | Facilitator | All slides with revised content, speaker notes, timing, and integrated interaction prompts. Maps 1:1 onto the existing impress.js deck so you can patch slides individually. | | `03_facilitator_runbook.md` | Facilitator | Minute-by-minute delivery script. Polls, lab transitions, "if running short" branches, contingency for technical failure. | | `04_participant_workbook.md` | Participant | Pre-work (≈20 min), live capture pages, lab worksheets, the 30-day follow-up plan. Distributed before the session. | | `companion.html` | Participant | Interactive web companion participants keep open during the session. Live timer, lab prompts revealed on cue, prompt-library builder, takeaway tracker with local persistence. | --- ## How to deliver this course **T–7 days.** Send `04_participant_workbook.md` (or a rendered PDF) to all registrants. Pre-work is short and sets a baseline so live time can be applied work, not definitions. **T–24 hours.** Read `01_course_design.md` end-to-end if you haven't. Apply the fact-check patches to the deck (15–20 min of edits). Test `companion.html` in the browser you'll be sharing. **T–60 minutes.** Run the facilitator runbook's "pre-flight" checklist (mic, polls, lab links, recording on). Open the companion in a side window. Have a backup tab with a single LLM chat open in case the planned tool fails. **Live session.** Follow `03_facilitator_runbook.md`. The structure interleaves theory and practice — there are three live labs, not one at the end. The companion reveals each lab prompt on cue so participants don't read ahead. **T+24 hours.** Send recording, slide deck, completed workbook template, and the calendar link for Coaching-tier 1:1s. **T+30 days.** Optional cohort follow-up: a 30-minute group call where Community-tier participants share what they used AI for in the last month. Powerful retention tool and natural setup for the advanced course. --- ## Tier-specific deliverables **Community ($29) gets:** - Live workshop attendance - 7-day recording access - Slide deck PDF - `04_participant_workbook.md` rendered as PDF - `companion.html` (works offline once loaded; data persists locally) - Group Q&A during the session **Coaching ($129) gets everything above plus:** - Lifetime recording access - 30-minute 1:1 follow-up call (book in T+7 to T+21 window) - Personal prompt library review — participant submits 3 prompts from their actual work, gets written feedback - A custom system-prompt / `CLAUDE.md` template built around their role (drafted by Florin using the companion's intake data) The Coaching tier's real value is the **personalised system prompt**, not the call. The call is the delivery mechanism for the artefact. --- ## What's intentionally not in this package - **A pre-made slide PDF.** Your existing impress.js deck is good — better than starting over. The slide deck doc shows changes against it, not a replacement. - **A recording infrastructure.** Use whatever you already have (Zoom + cloud recording is fine). - **Marketing copy or landing page.** That's a different project. The course-page copy you shared is solid; only the model names in the "what you'll cover" section need updating to match the corrected deck. - **Translation.** English only. Romanian/Spanish/Italian translation would be a separate pass — the workbook is the highest-leverage piece to translate first. --- ## Credits & versioning - **v1.0** (17 May 2026) — Initial build for the 18 May cohort. - **Built by Florin** using Claude Opus 4.7 + ADR-Driven Development. Every substantive design decision is in `01_course_design.md` so future updates can argue with the reasoning, not just the output. - **Next revision trigger:** when a new frontier model (GPT-6, Claude Opus 5, Gemini 4) ships, the model-landscape slide and knowledge-cutoff slide need a 10-minute patch. Everything else is model-agnostic. --- ## Quick start (if you only have an hour before the session) 1. Read `01_course_design.md` § "Fact-check audit" (10 min) and patch the 6 highest-priority slides. 2. Skim `03_facilitator_runbook.md` for the three lab transitions — those are where the session breaks if you're not ready. 3. Open `companion.html` in a browser. Click through to confirm each section reveals. 4. Send the workbook to participants if you haven't. 5. Go.