Learn AI Design with Phoebe

Two tracks into design where taste still decides

Generation got cheap, so judgement got expensive. A generator will hand you forty screens in a minute and not one opinion about which is right - and the average of everything on the internet is, by construction, not a point of view. A leader track gives design and brand heads the tool map, the IP and provenance reality, the craft-and-careers question, and a rollout that raises quality instead of just throughput. A designer track takes working designers through one real feature - the shared summary in Cadence, an AI meeting-notes app - from constraint brief to a spec'd, system-conformant, handed-off screen set. A live critique lab runs on the key pages: it renders a real screen, measures it for real (WCAG contrast ratios, computed spacing, hue families), and climbs 17 → 100 as you turn on the five craft levers - then loses 19 points the moment you reach for the sixth.

2tracks
16sessions
5craft checks, measured
45min per session

The leader track 🤝 · for design, brand & creative leads · 6 x 45 min · decisions, not demos

Judgement mode: what actually changed in design work, how to judge output like a director, which tools earn a seat and which are theatre, the IP and provenance exposure your legal team will eventually ask about, what AI does to craft and careers, and a rollout that is measured on quality rather than volume.

🌀 Leader 1 · easy

What AI changes in design work

Where generation genuinely collapses cost, where it quietly does nothing, and why the bottleneck moved from making to deciding.

▶ Leaders start here 👁️ Leader 2 · easy

Taste is the moat

Judging output like a director: the measurable half you can delegate to a check, the judgement half you cannot, and how homogenisation happens to whole categories.

The critique standard 🗺️ Leader 3 · medium

The tool and workflow map

Figma, Adobe, Canva, image generators, prompt-to-UI - what each actually outputs, where it lands in a real pipeline, and the buy-versus-ignore call.

Buy / adopt / ignore ⚖️ Leader 4 · medium

IP, provenance and brand risk

What you can register, what vendor indemnities actually promise, what Content Credentials prove, and the transparency rules for synthetic imagery.

The legal conversation 🪜 Leader 5 · deciding

Craft, juniors and career ladders

If AI does the work juniors used to learn on, where does the next senior come from? Hiring signals, portfolio review, and a ladder that still develops taste.

Team design 📈 Leader 6 · deciding

Rollout and measuring quality

A 12-month adoption plan whose success metric is design quality, not asset count - with the review ritual, the guardrails, and what to stop doing.

Strategy session

The designer track 🎨 · for product, brand & design-adjacent builders · 10 x 45 min · no coding needed

One running feature, taken further every session: Cadence's shared summary. The critique lab climbs 17 → 32 → 46 → 66 → 86 → 100 as the craft levers come on, and b10 does not ship until the capstone screens score 100 with an idea in them. Every session leaves you an artifact - a brief, a moodboard, a screen, a token set, a spec.

🔄 Designer 1 · easy

The designer's AI loop

Brief, generate, critique, direct, ship - and the five-check lab you will spend ten sessions climbing, starting at a brutal 17 out of 100.

▶ Designers start here 🎯 Designer 2 · easy

Art-direction prompting

The brief that produces usable work: constraints instead of adjectives, references that carry weight, and the iteration loop that fixes the brief rather than the pixels.

The core skill 🧭 Designer 3 · medium

Research and moodboards

Concept exploration at speed without converging on the average - divergent prompting, reference discipline, and a moodboard that states a position.

Exploration 📱 Designer 4 · medium

UI drafting: flow to screens

From a flow to Cadence's shared-summary screens with prompt-to-UI tools - and the states generated work always forgets.

Cadence v1 ships 🔬 Designer 5 · the heart

The critique lab

The signature session: the five-check script, the WCAG criteria to cite by number, the AI-as-reviewer prompt, and the 19-point cost of "can we add a colour".

▶ Score it live 🧩 Designer 6 · hands-on

Design systems and tokens

Turn the five checks into tokens and components a machine can enforce - so consistency stops depending on whether you were paying attention.

Consistency, not vibes 🖼️ Designer 7 · hands-on

Illustration and imagery

Generated assets that look like your brand rather than like a generator - style references, consistency techniques, and the licensing you must check.

Assets that fit ✍️ Designer 8 · hands-on

Content design and microcopy

Words are interface: labels, empty states, errors, and the AI-disclosure copy that builds trust instead of hedging - paired with the visuals, not bolted on.

Words as design 🤝 Designer 9 · hardest

Handoff to code

Specs an engineer can build from, AI-assisted front-end drafts, and the review that catches the gap between the design and what shipped.

Design to build 🏆 Designer 10 · hardest

Capstone: the Cadence feature

Assemble everything - brief, exploration, screens, tokens, imagery, copy, spec - and do not ship until the score is 100 and you can name the idea in one sentence.

Ship it at 100
easy - everyone medium hands-on / deciding hardest

Choose your path 🗺️

Two doors into the same conviction: the tool raises your floor, and only direction and a standard raise your ceiling.

🤝 Leader (decide & govern) a1→ a2→ a3→ a4→ a5→ a6
🎨 Designer (ship the feature) b1→ b2→ b3→ b4→ b5→ b6→ b7→ b8→ b9→ b10
⚡ Non-designer who must ship visuals b1→ b2→ b5→ b7→ b8
🚀 Full journey (lead AND design) a1-a6→ b1-b10

The knowledge map 🧠

Both tracks at a glance - hover a session to spotlight its concepts, click any node to jump in.