Workbench — stage index
Dev test surface: 29 lessons with a workbench stage — each link opens the rendered player.
Building AI Fluency
Data Portraits
What Is Generative AI?
Learn how a generative model predicts text rather than looking facts up, to send your first deliberate prompts, in the context of installing Claude and Slack and running your Hello-Worlds.
6 activities
Building AI Fluency
Hello World
Capabilities & Limits
Learn where generative AI is strong and where it hallucinates, to interrogate any answer before you use it, in the context of catching a confident model getting real facts wrong.
9 activities
Building AI Fluency
Writers Room
Delegate the Task
Learn to break a deadline job into delegatable pieces and brief each one, to make AI do real work for you, in the context of running a newsroom that must publish and verify under time pressure.
9 activities
Building AI Fluency
Six Lenses
Describe for Audience & Tone
Learn to steer output by specifying audience, style and tone, to make one message land for very different readers, in the context of re-rendering a single argument through changing frames.
9 activities
Building AI Fluency
Steelman
Verify, Don’t Trust
Learn to steelman, fact-check and rebut AI reasoning, to defend a position you can actually stand behind, in the context of a tribunal where the strongest in-frame argument wins.
9 activities
Building AI Fluency
Teachers Conference
Effective Prompting
Learn to construct a careful, technique-rich prompt and judge what it produces, to one-shot a genuinely useful artefact, in the context of building a study tool you would actually present and teach.
8 activities
Building AI Fluency
The Revision Sprint
Diligent Revision
Learn to run an AI revision loop that builds your own understanding, to revise honestly and in your own interest, in the context of telling the difference between learning with AI and cheating with it.
8 activities
Creating a Trailer
The Concept — Brief, Examples & Idea Lock
Generative Ideation
Learn to prompt an AI for many divergent options and select against explicit criteria, to ideate at volume without settling for its first answer, in the context of finding your trailer’s story engine.
6 activities
Creating a Trailer
The Trailer Logline
Constrained Prompting
Learn to drive an AI through a fixed-slot template instead of an open request, to force precise output on demand, in the context of compressing your story into a logline.
4 activities
Creating a Trailer
Trailer Beats — Grammar of the Cut
Structured-Output Prompting
Learn to prompt an AI to return an ordered, structured artifact rather than prose, to get usable scaffolds you can act on, in the context of structuring your trailer’s beats.
5 activities
Creating a Trailer
Who Does AI Work — & What Does It Cost?
Model–Cost Grid
Learn to match each trailer job to the cheapest model that can do it, to spend a real budget well, in the context of choosing which AI tool makes each part of your trailer.
4 activities
Creating a Trailer
The Style Frame
Aesthetic Control
Learn to control a generative model’s aesthetic by specifying named levers — palette, light, texture — to make a look reproducible, in the context of your trailer’s style frames.
5 activities
Creating a Trailer
Consistent World & Character
Consistency Control
Learn to hold a generative model consistent across outputs using locked references, to stop it drifting between generations, in the context of keeping one character the same through your trailer.
5 activities
Creating a Trailer
Hero Shots — Shotlist & Keyframes
Image-Model Specification
Learn to translate visual intent into a precise image-model specification — framing, lens, light, focal point — to make it generate what you actually want, in the context of your trailer’s hero shots.
5 activities
Creating a Trailer
Image → Motion
Video-Model Direction
Learn to direct a generative video model by specifying camera and subject motion, to turn a still into intentional movement, in the context of animating your trailer’s frames.
5 activities
Creating a Trailer
Edit — Cut to the Beat
Output Curation
Learn to judge, select, and sequence a pile of AI-generated outputs against a criterion, to assemble something coherent from raw generations, in the context of cutting your trailer.
5 activities
Creating a Trailer
The Voice — Narration & VO
Voice-Model Direction
Learn to direct a generative voice model on prosody — pace, emphasis, pause — to make it perform rather than read, in the context of your trailer’s voiceover.
5 activities
Creating a Trailer
The Score — Music & Sound Design
Audio-Model Direction
Learn to brief a generative audio model with functional specs — tempo, mode, instrumentation, where it swells — to get music that serves a purpose, in the context of scoring your trailer.
4 activities
Creating a Trailer
Finish & Screen
Verify & Disclose
Learn to critically check AI-generated work and disclose what’s AI-made, to ship output you can stand behind, in the context of finishing your trailer for the screening.
5 activities
Coding JARVIS
The Hacker Toolkit — Set Up the Dev Loop
The Build–Test–Revise Loop
Judge AI-generated code against what it actually does when run — read the real failure, feed the tool precise signal, and iterate to working — in the context of building your JARVIS chat widget.
8 activities
Coding JARVIS
Vibe-Coding a First Feature
Spec to App
Learn to drive an AI coder from a clear spec while verifying every output, to vibe-code a working feature you can trust, in the context of building your penpal widget under hackathon pressure.
5 activities
Coding JARVIS
Persona Design
Author the Persona
Learn to specify a consistent persona — voice, knowledge, boundaries — in a system prompt, to give your penpal a stable character, in the context of writing its system-prompt.md.
5 activities
Coding JARVIS
Building a Knowledge Base
Ground the Agent
Learn to build and structure a knowledge vault an agent retrieves from, to ground your penpal in real sources instead of invention, in the context of assembling its ≥15-source knowledge base.
4 activities
Coding JARVIS
Agent Tool Use — Give It Hands
Wire a Tool
Learn to define a tool contract an agent can call — intention to action — to let your penpal actually do things, in the context of wiring a working tool into your widget.
5 activities
Coding JARVIS
Multi-step Agent Loops
Orchestrate the Loop
Learn to decompose a goal into a plan–act–observe loop an agent runs, to make your penpal handle multi-step requests, in the context of giving it agency beyond a single reply.
4 activities
Coding JARVIS
Agent Ethics — When Should JARVIS Refuse?
Guardrails & Refusal
Learn to design guardrails and refusal pairs that decide when your agent must not act, to keep your penpal safe with strangers, in the context of bounding what it will and won’t do at the booth.
5 activities
Coding JARVIS
Adversarial Testing — Break Your Agent
Red-Team the Agent
Learn to attack your own agent to find its failure modes, to harden your penpal before strangers do, in the context of producing the bug log for your build.
4 activities
Coding JARVIS
Deploy the Chat Widget
Deploy the Agent
Learn to deploy your widget so it runs reliably and securely for strangers, to turn a thing that works once into a thing that works for everyone, in the context of putting your penpal live for the booth.
5 activities
Coding JARVIS
Demo at Seoul Expo
Present AI Work
Learn to demonstrate an AI system honestly and recover when it fails live, to present your work with composure, in the context of the Seoul Expo booth.
6 activities