s1-B1-1 cover AI-Native Product & Design Lab
Certificate program · 14 hours · hands-on coursework 3 modules · 26 exercises · 3 capstone projects
Program Kickoff & AI-Native Product Mindset
s01 · Betterworks · 90m · Full Cohort ·
released
s01 → 6 blocks, 41 slides, 14 trainer notes, 1 exercises, 6 reflects, 2 quizzes, 20 asides
Pre-read
None. The kickoff sets the context cold. Show up, cameras on.
Session goal
Establish the shared mental model the whole programme builds on:
1. **Why we're here** — the 14-hour arc, what each cohort member will be able to do by the end
2. **AI-native mindset** — defaults shift before tools do
3. **Compound prompting** beats one-liners — and there's a structure to it
4. **Context engineering** — Claude Projects and Skills as the daily defaults
5. **Outputs vs Judgement** — what AI handles, what you sign
1. **Why we're here** — the 14-hour arc, what each cohort member will be able to do by the end
2. **AI-native mindset** — defaults shift before tools do
3. **Compound prompting** beats one-liners — and there's a structure to it
4. **Context engineering** — Claude Projects and Skills as the daily defaults
5. **Outputs vs Judgement** — what AI handles, what you sign
The first 20 minutes set the air the cohort breathes for 14 hours. Land the WHY before any feature tour.
s1-B1-2 list What we'll cover
Joint kickoff. PM + Design tracks split. We join up again for design, integration & showcase.
- M1 · Foundations — Joint kickoff, then PM + Design tracks run in parallel
- M2 · Insight & Design — Cohort reconvenes for insight-to-design + prototyping
- M3 · Integration & Showcase — API workflows + capstone to senior leadership
s1-B1-3 question Why now?
Every quarter you don't operate AI-native
is a quarter your competition
Why now? Every quarter you don't operate AI-native is a quarter your competition does.
s1-B1-4 list What you'll be able to do by the end
Capabilities, not lectures.
- Clean up a 600-item backlog in a single AI workflow
- Prompt your design system as a guardrail (DS Project)
- Ship from prompt to working screens with Figma Make
- Wire a working API loop end-to-end
- … 1 more
s1-B1-5 custom (no title)
s1-B1-6 photo When I'm not working
s1-B1-7 list How we learn together
Four ground rules.
- Be vocal — the cohort learns when you do
- We're all learning — no senior/junior in this room
- Ask when something doesn't grep — there are no silly questions
- Videos on, unless you're away — so I know not to call on you
s1-B1-8 statement The new world rewards the learner — not the knower.
Show up curious. Leave with
The new world rewards the learner — not the knower. Show up curious. Leave with reps.
s1-B1-9 cover Mindset · Prompting · Judgement
The shared mental model the whole programme builds on.
s1-B1-10 list Today
Session 01 · ~90 min · flexible
- Welcome, course intro & ground rules
- AI-native mindset
- Pace, intros & Vision Benchmark
- Compound prompting — the live aha
- … 2 more
Two beats. Slow. The cohort either buys the frame here or politely tolerates the next 14 hours.
s1-B2-1 section From AI-curious to AI-native
The shift is in your defaults, not your tools. Tools follow defaults.
s1-B2-2 list Three habits of AI-native operators
Defaults · Composition · Reflection.
- Default to AI — Claude tab is open before Slack. Drafts start there; decisions end with you.
- Compose, don't type — prompts are rooms (role, context, input, task, format, reflect, constraint), not asks.
- Demo, don't memo — ship a working artifact in the meeting you used to ask for a deck.
The first 20 minutes the cohort hears about Pace, hears each other, and PG hears them. Extended block — the intro round is load-bearing.
s1-B3-1 split Pace — the case product
Goals coach for ambitious 28–45-year-olds. B2C app + B2B Teams dashboard. Different domain from your day job; same mechanics, same patterns
col 1 · What it is
Real PRDs, real screens, real backlog.
Shared sandbox all session.
col 2 · Why a case product
Shared vocabulary across the cohort.
Your work substitutes Pace in M3.
Showcase
s1-B3-2 activation Vision Benchmark + 1-min intros
Two prompts in your LMS · Proud + Wish · 15 min · on camera
s1-B3-3 statement You're not behind.
You're early — but only if you're moving.
You're not behind. You're early — but only if you're moving.
Same task. Two ways. Two outputs. The cohort sees the lever. We frame the *model picker* before the live demo so the lever has the right context.
s1-B4-1 section Make the prompt do the heavy lifting
Pick the model on purpose. Compose the prompt like an artifact, not an ask.
s1-B4-2 statement All models are not equal.
Pick on purpose, not on habit.
All models are not equal. Pick on purpose, not on habit.
s1-B4-3 activation How do you rate a model?
One dimension you actually use — call it out. Speed, cost, reasoning, modality, something else?
s1-B4-4 list What "new" means
The dimensions that move generation-to-generation. This is your selection checklist.
- Context window — how much you can shove in (8K → 200K → 1M+ tokens)
- Reasoning depth — extended thinking, tool use, multi-step planning
- Speed — time-to-first-token + tokens/sec (Haiku-class vs Opus-class)
- Cost — $ per million input vs output tokens
- … 2 more
s1-B4-5 grid3 Anthropic family — our defaults
Default. Reach for. High volume.
- Default: Claude Sonnet 4.6
- Reach for: Claude Opus 4.8
- High volume: Claude Haiku 4.5
s1-B4-6 activation How do you select a model?
Same task, three vendors — what's your default, and why? 30 seconds, voice not chat.
s1-B4-7 table The frontier today — by capability scale
col 1
col 2 · Frontier reasoning
col 3 · Daily workhorse
col 4 · High volume
col 5 · Design-native
s1-B4-8 section Compound prompting
Same model. Same task. The prompt does the heavy lifting.
s1-B4-9 grid3 The compound prompt
Seven parts grouped into three jobs. Cheat sheet handed out today.
- Set the scene: ROLE · CONTEXT · INPUT
- Define the work: TASK · OUTPUT FORMAT
- Guard the output: REFLECTION · CONSTRAINT
s1-B4-10 code Same task. The prompt is the only variable.
col 1
col 2
s1-B4-11 code What context did to the output
col 1
col 2
s1-B4-12 code Same task. The prompt is the only variable.
col 1
col 2
s1-B4-13 code What context did to the output
col 1
col 2
s1-B4-14 split One-liner vs Compound — same model, same input
Top row is the prompt. Bottom row is the output Claude returned. The structure is the lever.
col 1 · One-liner
summarise this PRD for the eng channel
Output (paraphrased):
"Here's a summary o
col 2 · Compound
ROLE: PM writing to eng leadsTASK: get +1 on 12 Jul launchOUTPUT: 4-line Slack m
Compound prompting is the *what*. Projects are the *where* and Skills are the *how*. Plus a 60-second floor on data hygiene.
s1-B5-1 section Stop re-explaining. Start composing.
Compound prompting is the *what*. Projects are the *where* and Skills are the *how* — context that compounds, not re-types.
s1-B5-2 section Project
A long-lived context bundle. Every chat in the Project starts already-knowing.
s1-B5-3 section Skill
A packaged invocation. Five-step chains become one command.
s1-B5-4 section Use both. Not one.
Project = what Claude knows. Skill = what Claude does. Both compound.
s1-B5-5 list Before you paste
60-second data hygiene.
- Is your data used to train? Check the model's privacy setting
- Is the workspace covered by your company's enterprise agreement?
- Would you be okay if this prompt + output leaked? If not — don't paste.
The frame, the Q&A pocket, the takeaways, the bridge.
s1-B6-1 section Output is the artifact. Judgement is the work.
AI handles the swing. You sign the decision.
s1-B6-2 statement AI raises the bar for judgment —
it doesn't lower the need for it.
AI raises the bar for judgment — it doesn't lower the need for it.
s1-B6-3 split Outputs vs Judgement
What Claude handles. What you sign.
col 1 · Outputs
First drafts. Lists. Summaries.
What you'd ask a sharp intern to swing at.
col 2 · Judgement
Decisions. Tradeoffs.
The thing you put your name on.
s1-B6-4 statement AI is the cheapest collaborator you'll ever have —
work it like a thoughtful col
AI is the cheapest collaborator you'll ever have — work it like a thoughtful colleague, not a vending machine.
s1-B6-5 list Today's takeaways
- Learner > knower — show up curious; reps over answers
- Picture your AI-native team — vision before tactics
- Pick the model on purpose — not on habit
- Structure beats model — compound prompting is the lever
- … 2 more
s1-B6-6 recap Next time
- S02 — Backlog Intelligence (PMs)
- 600 ideas → 47 dupes → 6 themes — live on Pace
- Bring token discipline. We cost it on screen.
s1-B6-7 close Thank you
One thing that landed today — drop it in the reflection prompt below.
After this session
- **Setup check (today):** confirm Pace Aha! sandbox access — log in once, fix on Wednesday if blocked.
- **For S2 (Thursday):** skim 10 ideas in the Pace backlog. Note one "huh, that's messy" reaction. No formal pre-read.
- **For designers (parallel track):** S2 is PMs-only; designers get S4 on Friday.
- **Cheat sheet:** download the compound prompting one-pager from the LMS (linked in chat).
---
## Open items (for polish pass / PG review)
1. **[?CONFIRM] Growth-mindset reframe** — picked the *learner* statement; three alternatives listed in the Block 1 aside.
2. **[?CONFIRM] Block 2 AI-native mindset content** — picked *defaults shift* reframe + *three habits* list; full shortlist in the Block 2 aside.
3. **[?CONFIRM] PG intro slide fields** — verify `role` line and tags on the merged intro+credentials slide.
4. **[?CONFIRM] Pace screenshots** — drop `goal-set.png`, `coach-thread.png`, `teams-dash.png` into `/Users/Shared/pg-website/public/images/pace/` (or copy `claude-design-to-code.png` from `career-agent/content/assets/claude-design-handoff/` as a stand-in).
5. **[?CONFIRM] Vision Benchmark Miro URL** — drop the working Miro board URL into the activation slide's `link:` field.
6. **[?CONFIRM] Model parameters list** — six dimensions on the "What 'new' means" slide; reorder/extend if needed.
7. **[?CONFIRM] Cross-vendor model table data** — model names as of 2026-06-14; verify before each cohort. Fable confirmed in design-native tier.
8. **Compound prompting cheat sheet** — `decks/cheat-sheets/01-compound-prompting.html`. Browser-print to PDF + upload to LMS before delivery.
9. **Token economics relocated to S02 opener** — update S02 docket when its polish pass happens.
---
## JSX / Component-layout escape hatch
PG flagged: "we might need jsx content for some slides so they don't break." Current state:
- The `custom` layout reads a raw `html:` string from the directive and dumps it into a
`
- **For S2 (Thursday):** skim 10 ideas in the Pace backlog. Note one "huh, that's messy" reaction. No formal pre-read.
- **For designers (parallel track):** S2 is PMs-only; designers get S4 on Friday.
- **Cheat sheet:** download the compound prompting one-pager from the LMS (linked in chat).
---
## Open items (for polish pass / PG review)
1. **[?CONFIRM] Growth-mindset reframe** — picked the *learner* statement; three alternatives listed in the Block 1 aside.
2. **[?CONFIRM] Block 2 AI-native mindset content** — picked *defaults shift* reframe + *three habits* list; full shortlist in the Block 2 aside.
3. **[?CONFIRM] PG intro slide fields** — verify `role` line and tags on the merged intro+credentials slide.
4. **[?CONFIRM] Pace screenshots** — drop `goal-set.png`, `coach-thread.png`, `teams-dash.png` into `/Users/Shared/pg-website/public/images/pace/` (or copy `claude-design-to-code.png` from `career-agent/content/assets/claude-design-handoff/` as a stand-in).
5. **[?CONFIRM] Vision Benchmark Miro URL** — drop the working Miro board URL into the activation slide's `link:` field.
6. **[?CONFIRM] Model parameters list** — six dimensions on the "What 'new' means" slide; reorder/extend if needed.
7. **[?CONFIRM] Cross-vendor model table data** — model names as of 2026-06-14; verify before each cohort. Fable confirmed in design-native tier.
8. **Compound prompting cheat sheet** — `decks/cheat-sheets/01-compound-prompting.html`. Browser-print to PDF + upload to LMS before delivery.
9. **Token economics relocated to S02 opener** — update S02 docket when its polish pass happens.
---
## JSX / Component-layout escape hatch
PG flagged: "we might need jsx content for some slides so they don't break." Current state:
- The `custom` layout reads a raw `html:` string from the directive and dumps it into a
`
`. That's how slides 6, 7, and the cross-vendor table work.
- This is fragile in two known ways: (a) `:::` anywhere inside the HTML closes the directive
early, (b) unescaped quotes in YAML break the parser before the engine ever sees the html.
- Path forward when it breaks: add a `component` layout to `engine.js` that takes a
`component: PaceShowcase` field and renders an actual Astro/MDX component imported from
`src/components/slides/`. The directive content stays YAML-thin; complex JSX lives in the
component file where types and tooling help.
For S01 the inline-HTML approach is fine. Flag any slide that starts feeling too clever — that
slide is the one we promote to a component.
- This is fragile in two known ways: (a) `:::` anywhere inside the HTML closes the directive
early, (b) unescaped quotes in YAML break the parser before the engine ever sees the html.
- Path forward when it breaks: add a `component` layout to `engine.js` that takes a
`component: PaceShowcase` field and renders an actual Astro/MDX component imported from
`src/components/slides/`. The directive content stays YAML-thin; complex JSX lives in the
component file where types and tooling help.
For S01 the inline-HTML approach is fine. Flag any slide that starts feeling too clever — that
slide is the one we promote to a component.
Add ?token=… to the URL.