Sparks — Curriculum ReferenceFifteen Sparks. One fluent professional.
Every Monday, a new Spark lands in the student’s inbox — a 30-minute session on one skill, applied to one real task, inside the AI tool they already use. Across fifteen weeks, the skills compound and the habits stick. The student graduates with a working AI playbook — four personal context files, around ten customised prompts, fifteen crib sheets — and the professional judgement to know when to trust AI and when to push back.
Sequenced so the skills compound. Early Sparks build the context files and prompting habits that later Sparks rely on; mid-course Sparks apply those skills to high-stakes work; the final Sparks consolidate everything into a configured workspace and a documented workflow.
The foundation. AI starts every conversation blank — here the student fixes that, building me.md (their first personal context file) and walking out with a personalised plan for the fifteen weeks ahead.
Role assignment. The same question asked blind, then with a role assigned — the delta is the lesson. First reusable prompt in the library: a role-assignment template tuned to a recurring task, stacked on me.md.
The three-stage meeting workflow — before, during, after. The student prepares for, captures, and follows up a real upcoming meeting, and walks out with customised meeting prompts and the EXTRACT pattern.
The second context file. Builds org.md — culture, audiences, constraints, unwritten rules — and stacks it with me.md so responses are tailored to where the student actually works.
AI’s default is to agree. The student casts it as a bounded adversary to stress-test a real position they hold — with sycophancy named as the reason AI usually agrees instead of pushing back, and steel manning as the move that makes it challenge properly.
First lesson of the AI Professional tier. Scope a question, have AI source it, then run the three cautions — it invents, it’s incomplete, it agrees with itself. The spine: AI’s output is a lead, not a verdict. Hallucination named and made memorable.
Audience-first framing for a real upcoming presentation — start with who is listening, not what you want to say — then rehearse the toughest questions with AI before the room asks them.
The student draws weeks of practice into their own rules: a five-question pre-flight check and a short personal AI-use policy, in their own words. Anchored by "expert in their field first, AI delegator second".
AI makes the thing itself — a Word document, then a PowerPoint slide, on the student’s own work. The skill is the handoff: when to keep working through AI and when to take it out and finish by hand. Walks out with a reusable artefact brief.
Advanced context engineering as a three-level ladder — paste-it-yourself, a configured workspace, connected files. Builds project.md and sets context up once so AI starts every session already knowing the student.
First lesson of the AI Fluent tier. AI reads the emails the student already writes, learns their voice, and builds voice-and-style.md — a fourth context file — so drafts start sounding like them, not like everyone.
A conversation where the other person pushes back. AI builds the other side’s strongest case, maps the objection the student would least want to face, and plays the counterpart so they can rehearse the opening before the real thing.
Delegation made deliberate. Three modes — automation, augmentation, agency — matched to a real task the student has this week, with a calibrated call on what to hand over and what stays theirs.
The AI mechanics of working with data: check it is sound, let AI read and summarise it, get a recommendation, pin down how the numbers were computed (in code, not guesswork), and produce a board-ready output — then verify what is load-bearing before trusting it.
The inversion — AI in the questioner’s chair. The student assigns it a teaching role (a Socratic partner that questions, or a patient tutor that explains), then closes the programme knowing how to keep getting better on their own.
Sparks is not fifteen disconnected lessons. Five threads run end-to-end, each surfacing in the session where it matters most — experienced first, named second, then revisited until it becomes instinct.
A playbook that grows
Every Spark adds something concrete to the student’s playbook — a context file, a customised prompt, a crib sheet. By Spark 15 the student owns around 28 working documents, all of them tailored to how they actually work. The framing is deliberate: not a finished product but a toolkit they keep refining as their work changes — which is why every personal file carries the header Living document — refine as you go.
Vocabulary, in context
Around twenty-eight field terms — co-intelligence, context window, hallucination, alignment, anthropomorphism, reasoning models — landed where they matter, so graduates can hold their own in any AI conversation.
Build, then use when natural
Artefacts from early Sparks get reused in later ones, not as ritual but where the connection is genuine. The me.md built in Spark 1 grounds the stress-testing work in Spark 5, and the steel-manning move from that session returns to rehearse a hard conversation in Spark 12. By the late Sparks the playbook is a working set the student reaches for when the task fits — not a folder they remember to open before each session.
Responsible AI, every Spark
Professional judgement about when and how to use AI is built into every session, not packaged as a compliance module. Source integrity in Spark 6, the pre-flight check and a personal AI-use policy in Spark 8, data sensitivity in Spark 14. Spark 8 consolidates the universal AI-use principles into a framework anchored by ‘expert in their field first, AI delegator second’ — a stance the student internalises through weeks of real practice, not a memo they file away.
Confidentiality as muscle memory
Students paste real work into AI throughout the course. Confidentiality is a running practice, not a one-off lesson — introduced in Spark 1 at the first upload-and-paste moment for me.md, rehearsed at every paste moment thereafter, consolidated in the responsible-AI principles in Spark 8. Names, numbers, identifiers — describe the shape, not the fingerprints. By Spark 15 the habit is instinct.
Want to see a Spark in action?
Book a walkthrough or ask for a pilot scoped for your team.