Mike Litman
50 Decks Before Claude Design
A TALK BY MIKE LITMAN

50 Decks Before
Claude Design

What it actually takes to ship presentations before the tools caught up. A first-hand account, written on launch day.

50+

Presentations built entirely in HTML, CSS, and JavaScript. No PowerPoint. No Figma. No design tool. Just code, a custom template, and a nine-step checklist. Across these topics:

AI WORKFLOWS VOICE AGENTS BRAND STRATEGY CULTURAL INTELLIGENCE PRODUCT PITCHES TASTE SYSTEMS DESIGN PHILOSOPHY FOUNDER THINKING
01
THE PURPOSE

Thinking out loud in public is the portfolio

These presentations aren't filed away after a client meeting. They're live, deployed, searchable, and tracked. Every one of the 50 decks is proof of work: a public record of how I think, what I'm paying attention to, and how fast I can move from idea to published argument. In a career built on cultural intelligence, the intelligence has to be visible.

PUBLIC + TRACKED + DEPLOYED PROOF OF WORK THINKING IS THE PRODUCT
02
THE IDEATION

It starts with something I can't stop thinking about

Something interesting lands -- an article, a product launch, a pattern I keep noticing. I save it to my #presentations-claude-code Slack channel. Then I open Claude Code and ask one question: is there enough here for a standalone presentation and a blog post? If yes, we start building. If not, it goes back in the pile.

CURIOSITY FIRST SLACK AS INBOX CLAUDE CODE AS CREATIVE DIRECTOR
03
THE INPUTS

The presentations start long before the Slack channel

The raw material comes from living in the work. Following the right voices. Reading deeply in specific areas. Noticing when different threads start converging. My information diet is deliberately narrow: AI and its cultural implications, brand strategy, the economics of creative work, human behaviour at scale. When something arrives at the intersection of two or more of those, it's usually worth saving.

AI + CULTURE BRAND + BEHAVIOUR THE INTERSECTIONS ARE WHERE THE DECKS ARE
04
THE CONVERSATION

Finding the angle, not just the topic

The conversation with Claude Code isn't just "is there enough here?" It's: what's the actual argument? Who is this for? What would make someone change their mind? The first instinct is rarely the sharpest one. The topic might be Claude Design -- but the angle is "what 50 decks taught me before it existed." That reframe is where the deck gets interesting.

TOPIC ≠ ANGLE THE REFRAME IS THE WORK FIRST INSTINCT IS A DRAFT
05
THE MOMENT

Knowing when to ship is as important as knowing what to ship

This deck is being written on Claude Design's launch day, before I have access to it. That's a deliberate call. The best time to publish a take isn't when you've fully digested something -- it's when the topic is live and the field is still open. A day-one perspective from someone with receipts is worth more than a retrospective from someone who waited.

PUBLISH NOW RECEIPTS BEAT RECENCY DAY-ONE IS AN EDITORIAL CHOICE
06
THE GATE

Sometimes Claude Code says no

Not everything that lands in the Slack channel becomes a deck. Sometimes there isn't enough for a standalone argument. Sometimes the angle isn't distinct enough. Sometimes it's too close to something already in the portfolio. That rejection is part of the quality control. It's why 50 presentations don't feel like noise -- because a lot more than 50 ideas were considered and declined.

NOT EVERYTHING SHIPS THE NO IS AS IMPORTANT AS THE YES COHERENCE IS A CURATORIAL CHOICE

The curation is the hardest part.
The build is proof that the idea was worth it.

Every deck that ships passed a test most ideas don't survive. When it does make it through: nine steps, no shortcuts.

07
THE WORKFLOW

Nine steps to ship one talk

Not a checklist: a production pipeline. What a team of three would have managed in 2020 now runs in a single session.

01Write the HTML deck
02Generate the OG image via Playwright screenshot
03Capture the thumbnail Playwright at 1280×720, visual verify
04Update decks.json and add the pretty URL redirect
05Mobile check at 375px
06Run the LinkedIn carousel 12 slides, PDF output
07Write the companion blog post
08Generate the X Article Claude Opus, 1,200 words
09Claims audit: every unattributed number must be sourced
08
THE FOUNDATION

Built the system once. Every deck inherits it.

All 50 decks run on a single shared HTML template. It ships pre-wired with Plausible analytics, a grid and mosaic view, full mobile support, share buttons, keyboard navigation, print styles, and a resume-from-last-position feature. That's not set up per deck -- it's inherited. The infrastructure decision was made once. Every deck that follows pays nothing for it.

ONE TEMPLATE 50 DECKS ANALYTICS ON EVERY SLIDE ZERO SETUP PER DECK
09
THE OUTPUT

One idea becomes four pieces of content automatically

The nine steps aren't just a build process -- they're a content strategy. One conversation, one set of slides, and you get: a deployed presentation, a companion blog post, a LinkedIn carousel (12 slides, PDF), and a 1,200-word X Article written by Claude Opus. All from the same source material. The idea does the work once. The system distributes it everywhere.

PRESENTATION BLOG POST CAROUSEL X ARTICLE ONE SOURCE. FOUR FORMATS.
10
THE SHIFT

A strategist's primary output used to take two weeks

For most of my career, a proper presentation meant research, narrative architecture, visual production, and at least one round of client revisions. One to two weeks minimum, if you were doing it properly. Now the same output -- researched, structured, designed, deployed, with a blog post and LinkedIn carousel -- takes days. Sometimes a weekend. The thinking hasn't got faster. Everything around the thinking has. This presentation is the proof: conceived, built, and published on the day Claude Design launched.

WAS: 1–2 WEEKS NOW: HOURS THE THINKING STAYED THE SAME THIS DECK IS THE PROOF

The hardest part isn't the thinking.
It's everything that comes after the thinking is done.

By the time a presentation is live, you've spent as much time on logistics as on ideas. That's the gap the tools haven't closed. Until now, possibly.

11
ANTHROPIC LABS – 17 APRIL 2026

Collaborate with Claude to make visual work

"Gives designers room to explore widely and everyone else a way to produce visual work." Powered by Claude Opus 4.7. Available on Pro, Max, Team, and Enterprise.

INText, images, documents, code
OUTDesigns, slides, prototypes
Auto-applies your brand system
Fine-grained editing + live adjustments
Inline comments
Org-scoped sharing
Export: HTML, PDF, PPTX, Canva
Handoff to Claude Code
CLAUDE OPUS 4.7 ANTHROPIC LABS anthropic.com/news/claude-design-anthropic-labs
FROM INSIDE ANTHROPIC — @FLOMERBOY

7 tips from someone on the verticals team who serves 7 products with it

01Set up your design system and core screens first — an hour here is worth it
02Iterate with your engineers live — Claude is fast enough to riff in the meeting
03Use the Comment tool for surgical edits — point and crit, don't describe
04Ask Claude to make video demos of your ideas
05Use connectors (docs, Slack) — prompt from meeting notes, go for a walk
06Ask for bespoke on-the-fly tools — it's not a canvas tool, experiment
07Know when to slow down — icons, illustrations, naming are still craft
ANTHROPIC VERTICALS TEAM LAUNCH DAY, 17 APRIL 2026
FROM INSIDE ANTHROPIC — @HELLOITAUSTIN · GROWTH MARKETING

The marketer's version of the same story

01Claude scopes the brief before building — asks what you're actually testing, what you need to see, before a single pixel moves
02Describe a new component in plain English — Claude builds it to your design system and drops it into the live page
03Tweaks panel generates custom controls per page — toggle sections, drag to reorder, A/B test live, pressure-test ideas in chat
04Share the Claude Design URL with your design team to refine — or hand it straight to Claude Code to ship

"I used to write up my page ideas in a doc and hope they survived the handoff. Now I can show my design team."

ANTHROPIC GROWTH MARKETING LAUNCH DAY, 17 APRIL 2026
12
MY PREDICTION

The nine steps become a conversation

Every script I've written to generate thumbnails, OG images, and carousel exports exists because there was no tool that could do it conversationally. Claude Design isn't just faster -- it's a different model entirely. Instead of automating a checklist, you describe what you need and iterate. That's not an efficiency gain. It's a category change.

FROM CHECKLIST TO CONVERSATION ITERATION REPLACES AUTOMATION
13
THE LIMITS

The editorial judgement is still yours

The interesting question isn't whether Claude can generate a slide. It's whether it can know which slide to cut, which idea isn't ready yet, and when the whole thing is trying too hard. That's not a tool problem. That's experience built across 50 presentations.

SEQUENCING IS STILL HUMAN POINT OF VIEW IS STILL HUMAN KNOWING WHEN TO STOP IS STILL HUMAN

Better tools lower the cost of a bad idea.
Only better thinking raises the floor.

I don't have access to Claude Design yet. This talk was written on its launch day, before I'd used it once. That's the point: 50 decks in, I already know what I need to think better. The tools are catching up. The question is whether I am too.

14
WHAT I ACTUALLY WANT TO KNOW

The questions I can't answer yet

?What does the nine-step checklist actually collapse to: one step, three?
?Can it apply a brand system it's never been shown before?
?Does the output need as much editing as any first draft always does?
?Does it know when to stop, or does it just keep generating?
?Is it better at layout, or at thinking about what the layout should say?
WRITTEN ON LAUNCH DAY, BEFORE FIRST USE
WATCH THIS SPACE

Update coming
when I get access.

Until then: 50 decks, nine steps each, all live at mikelitman.me/decks

mikelitman.me · hello@mikelitman.me

Continue from slide ?