1
Hook + objectives
Attention-grabbing opener — a scenario, question, or surprise — paired with 2–4 action-verb objectives.
No “in this section you will learn…” openings. No “understand” or “learn about” objectives.
Most AI authoring tools generate.BREW also rejects.
One Claude authors a course. A second Claude — adversarial — checks the work against the methodology you picked. When the author has skipped a step, vagued out, or trotted out generic platitudes, the critic sends it back. We've watched this happen.
BREW is in beta. Invite-only — opening to a small group later this year.
BREW produces traditional e-learning courses — SCORM 1.2, SCORM 2004, xAPI, HTML5, and printable PDF — that ship to any LMS or run standalone. Multi-pass authoring up to ~25 generation steps. Methodology followed in close detail. Visuals generated alongside the content. Made by Alt Shift Lab, the team behind CoNoggin.
First, the brief
Generation doesn't start with a prompt. It starts with a conversation.
You give BREW the raw material — paste content in directly, or upload PDFs, Word documents, Markdown, plain text, structured data. BREW reads it and asks clarifying questions: who's the audience, what change you want, what context the learners are in, where this fits with what they already know.
You pick a methodology. Evidence-Based(Gagné's Nine Events + Merrill's First Principles, the default), Problem-Based Learning, Action Mapping, 4C/ID, or Competency-Based. Bringing your own is a first-class option — paste your methodology description, or upload a course you want to emulate. The pipeline follows whatever you choose.
Then BREW drafts an outline. You review it. You ask for changes. You keep iterating until the shape is right. Only then does the multi-pass pipeline begin.
How a course gets made
Each section runs through a five-pass pipeline. Then the whole course goes through three review passes. The work takes minutes, not seconds. What emerges has been argued with on the way in — by the same standards an instructional designer would apply on a third pass.
Per section · 5 passes
2
Demonstrate
Worked examples and multi-modal content. Capped at 20% of section time.
No walls of text. Multiple modalities, every section.3
Practice activities — the big one
3–6 varied interactive blocks with feedback. 50–60% of section time, enforced. At least two different interaction types per section, including one scenario, branching, or case study.
Feedback explains why, not just correct/incorrect.4
Assessment
3–5 quiz blocks, each with three question-bank variants. HTML5 and SCORM exports randomly select primary or variant on page load — different attempt, different test.
Tests application, not recall. Wrong answers are plausible.5
Transfer
A real-world application prompt and reflection. Specific to the learner's context, not generic.
Whole course · 3 review passes
6
Adversarial review
A “critical reviewer” persona evaluates each section against a 13-point quality rubric. Returns specific fixes — not pass/fail.
Practice ratio. Interaction variety. Hook quality. Action-verb objectives. Feedback that explains. Methodology fidelity. The lot.7
Fix pass
Applies the reviewer's fixes to each section that failed checks. One call per failing section.
8
Coherence pass
Reads the whole course end-to-end. Checks flow between sections, spaced retrieval, no repetition, Bloom's progression, consistent tone and terminology.
~22–26 Claude callsper 4-section course. Minutes, not seconds. A course that's been argued with before it gets near a learner.
And keep editing — forever
A course isn't done when generation finishes. The editor is the same surface for the next year. Want to refresh language? Edit a paragraph by hand. Want to update a methodology citation, swap an example, replace an entire activity? Tell the AI editor what you want — it returns the surgical change. The full course stays as context; only the bit you asked about moves.
The course you ship today, you can sharpen six months from now without rebuilding from scratch.
Methodology, in close detail
Pick the framework. BREW will follow it — not just label the output with it. The reviewer in pass 6 checks for fidelity to the methodology you chose, not to a generic instructional template.
Evidence-Based (default)
Gagné's Nine Events plus Merrill's First Principles, with spaced retrieval and meta-analysis-backed practice ratios.
Action Mapping (Cathy Moore)
Activity-first, minimal content. Every element supports a measurable behaviour change.
The critic flags content that's interesting but doesn't serve a named action.4C/ID
Worked examples → completion tasks → full tasks. Scaffolding fades. Task classes progress simple to complex. For complex skills with parts that must serve the whole.
Problem-Based Learning
The course is organised around real problems to solve, not topics to cover. Learners discover knowledge through inquiry.
Competency-Based
Define target skills. Assess against them. Mastery-paced.
Custom — bring your own
Paste your own methodology description (or upload an example course). Claude uses it as the structural guide; the critic learns what to check from the description itself.
In practice: “this course is designed using ADKAR”in BREW means the activities address Awareness, then Desire, then Knowledge, then Ability, then Reinforcement, in that order, with each stage's required moves present. Not a label. Not a slide. Actually-followed.Visuals, generated with the course
Courses don't ship as walls of text. BREW generates the visuals as part of the multi-pass system, so they fit the methodology, the topic, and the brand.
Images
Generated for the topic and the activity. Not stock photos. Brand-themed from your seed colour. Pedagogically grounded — what should this image teach, not decorate.
Mermaid diagrams
Sequences, decision trees, hierarchies — anything structural. Flowcharts render in a hand-drawn aesthetic via Excalidraw, brand-themed. Live, editable.
Infographicscoming
Designed alongside the content, not pasted on after.
Talking-head videocoming
AI-generated, for the sections where presenter delivery suits the material.
Ships anywhere
SCORM 1.2 / SCORM 2004
Packaged for any compliant LMS. Score tracking, progress, suspend_data.
xAPI / Tin Can
For modern learning record stores. Dual-tracking architecture lets SCORM data flow to your LMS while xAPI statements flow to a separate LRS in parallel.
HTML5
Standalone web courses. No LMS required. Single-file or ZIP.
Printable PDF
For offline, print-only, or hybrid deployments. Answer keys, Georgia serif.
Brand-customisable: colours, typography, logo, voice. Your courses, made by BREW, look like they were made for you.
What BREW doesn't try to do
BREW courses are static once authored. SCORM, xAPI, HTML5, PDF — none of those formats were designed for live AI interactivity at delivery time. So BREW doesn't pretend.
When the work calls for a tutor that adapts in conversation, a session generated on demand, or a mini-course that responds to the learner in real time — that's CoNoggin, where delivery isn't constrained by an external format.
Many BREW users only ever need BREW. That's fine. BREW is its own product, not a teaser for the rest of the family.
Who BREW is for
Instructional designers who want AI to multiply craft, not replace it
The methodology is real. The critic catches what you'd catch on a third pass.
L&D specialists at organisations whose LMS isn't going anywhere
Courses still need to live in the LMS — properly. SCORM-deliverable, brand-customisable, methodology-grounded.
Course authors at consultancies producing courseware for clients
Brand-themed per-client. SCORM-compliant by default. Methodology fidelity you can defend in a review meeting.
Compliance training producers who need SCORM and don't trust generic AI
Adversarial review catches the regressions that one-shot generation reliably produces. Audit-able output.
What BREW is not
- A one-prompt course generator (we have ~25 passes for a reason)
- A drag-and-drop authoring tool with AI buttons (BREW writes the course)
- A SCORM cloud you upload finished courses to (BREW makes the courses; you upload them where they live)
- A live AI tutor in disguise (we don't pretend SCORM supports it)
BREW makes the courses. The depth is the difference.
See it argue with itself
BREW is in beta. Invite-only right now; opening to a small group later this year. If you're building courses you'd want a thoughtful instructional designer to argue with, get in touch.