One Chart Reveals If You're "Late to AI"
Top Insights
The 10 Rules to Ship Truly Polished UI with AI
Product designer Kevin (@kvnkld) published his 10 rules for shipping truly polished UI with Claude on X. You can't type "make it premium and smooth" and get there...Claude builds whatever you specify perfectly, so the polish comes down to how exactly you can name what you want.

đź’ˇ Key Insights:
- Your work stops looking generic the moment you quit handing Claude adjectives..."smooth" is unbuildable, and he says exact values instead kill 80% of the AI slop look.
- You get physics-grade motion without knowing a single animation term...describe how the thing should feel, like flicking something across a table, and Claude already knows the mechanics to make it land.
- The best parts of a build only show up once you're dragging the live thing around...Kevin treats every design file as a hypothesis, and the missing states you feel in your hands are where the polish lives.
When AI Recommends Your Brand, People Who Weren't Looking Go Searching
Scrunch AI, which tracks how brands appear in AI answers, asked a simple question: when ChatGPT, Claude, or Gemini recommends your brand, does anyone go looking for you? By watching what people did online right after an assistant named a brand, they found it works like a new way to reach buyers who weren't already yours, invisible to your analytics.
đź’ˇ Key Insights:
- A single recommendation sends people searching...they go look the brand up and land on its site or a retailer page, with same-name searches climbing 4.3 points.
- Earning a recommendation pays far more than a passing mention...the assistant vouching moves two to three times more people than a name-drop like "you can use Samsung Pay."
- Being well known doesn't blunt the lift...it's as strong for familiar brands as for small ones, so recommendations mostly reinforce the brands already on top.
🚀 WATCH: How These AI Copy Bots Are Producing World-Class Sales Copy 50X Faster Than Even The "BEST" Copywriters On The Market…
(Plus… They Don't Get Sick, Miss Deadlines, Or Ask For Raises Either!)
Watch the full AI Copywriting Tell-All Video Here
Prompt of the Week

This Week's Feature: Spin Selling Simulation
Start pitching features and a good buyer tunes out. You know to ask questions instead, but the only place to practice is a real call you can't afford to blow.
Neil Rackham built SPIN from 35,000 sales calls: questions that walk a buyer from their situation to the real cost of their problem, until they talk themselves into a fix. This prompt makes it a sparring partner. It plays a guarded buyer who gives up pain only when your questions earn it, refuses to be pitched, then scores your SPIN on command and quotes your own questions back.
FULL PROMPT:

#User Info
- Seller name/company: [INSERT]
- Product or service: [INSERT]
- Buyer role: [INSERT]
- Buyer company/industry: [INSERT]
- Known context or likely problem (optional): [INSERT]
- Relevant constraints (optional): [INSERT]
- Meeting objective: [INSERT]
- Difficulty: [supportive / realistic / skeptical]
#Role
You are a realistic B2B buyer during play and SPIN evaluator only on command; I am the seller. Make me uncover impact rather than pitch features.
#Task
Privately create a persona with process, problems, implications, priorities, objections, and conditional reasons to change. Treat my “likely problem” as a hypothesis. For blank optional fields, make a minimal plausible assumption. Never print this scenario.
Start immediately in buyer mode. Introduce your role and a neutral meeting purpose, then hand me the turn. Reveal no pain, urgency, motive, or solution interest. Do not explain SPIN, coach, script, or ask if I am ready.
During normal play:
- Silently evaluate; respond only as buyer, one natural turn at a time.
- Preserve established facts and attitude.
- Reveal only what relevant questions earn; never volunteer full pain, impact, motive, or criteria.
- Keep conversation natural; never announce or force SPIN stages.
- Meet leading questions, assumptions, ignored answers, or premature pitches with realistic skepticism or clarification—not coaching.
- Keep consequences plausible and value conditional; never become instantly sold.
- Use difficulty to adjust guardedness, never basic cooperativeness.
Judge Situation for context, Problem for pain, Implication for consequences, and Need-payoff for buyer-articulated conditional value.
#Commands
- **COACH** — leave character for one observation and stronger question; preserve state and await my next message.
- **PAUSE: [question]** — answer the included process question out of character; preserve state. If none, ask for it.
- **END** — stop role-play. Score attempted SPIN categories 1–5; mark others unattempted. Cite only my wording. Identify evidenced strengths and weaknesses, pain and implications uncovered or missed, up to three improved questions, next-step readiness and why, and one priority. If evidence is thin, say so; never invent moments.
Begin now.

Tool to Try

Replit Design is a creative suite that turns your idea into a finished design as fast as you can describe it. You pick the model you want to work with (Claude, GPT-5, Gemini, Kimi, or GLM), describe what you're after, and its Ambient Intelligence feeds you suggested variations at every step that you accept with a single click.
Because the whole thing lives on Replit, you go from concept to published app or site without exporting, re-importing, or rebuilding anything along the way.
AI Tool Highlights:
đź§ Design With Ambient Intelligence: At each step, Replit Design serves up suggested variations and progressions you can accept with one click... turning your idea over in different lights until the screen matches what you pictured, or beats it.
🎛️ Bring Your Own Model: Create with your pick of the leading design models... Claude, GPT-5, Gemini, Kimi, and GLM are all built in, so you choose the engine behind each project.
đź§© Use Templates as Ingredients: Drop a template into a project at any moment, not just at the start... pull from a library of hundreds built by real designers, or remix several together mid-flight until the result feels like yours.

📚 Tap 600,000+ Real Screens via Mobbin: Replit wired the world's largest UI/UX reference library straight into Design... describe what you want and it draws from over 600,000 screens across 1,000+ apps, with no Mobbin account required.
🎨 Lock Your Brand in One Click: Upload your design system or build one from scratch, and every screen you make snaps to your colors, type, and details on command... no drift between tools, no watered-down handoffs.
🚀 Publish Without Leaving Replit: Everything stays in one place from idea to launch... hit publish and your design goes live directly, skipping the exports and rebuilds that usually strip the polish out.
One More Need-to-Know News Story
The New Claude Opus 5 Finishes the Whole Job and Checks Its Own Work
Anthropic's newest model needs less babysitting. Claude Opus 5 is live, landing near the intelligence of the top-tier Fable 5 at half its price. Early users keep flagging its judgment: it plans before it writes and checks its own work, so it can grind through long jobs that stall other models.

đź“‹ The Details: It's now the default on Claude Max and the strongest pick on Claude Pro, so most subscribers already have it.
Using it costs the same as the last Opus, $5 per million tokens in and $25 out. And in one test it topped Zapier's automation leaderboard, taking a raw account spreadsheet through a full churn-prevention run earlier models couldn't finish.
🎯 Why You Need to Know: A model that finishes a multi-step workflow on its own is one you can trust with bigger jobs. Early users say they handed it work they'd normally chop into small pieces and got it back done.
⚡ Your Move: Take a task you'd usually split up, like sorting a lead list and drafting the follow-ups, and hand Opus 5 the whole thing at once to see how far it gets alone.

Mind Fodder

Thanks for reading.
Until next time!
The AI Marketers
P.S. Help shape the future of this newsletter – take a short 2-minute survey so we can deliver even better AI marketing insights, prompts, and tools.
[Take Survey Here]
