Claude Code got me my first customer
TL;DR
The stack is the story: Idea Browser + Claude Code + Paper + Humblytics turns one idea into a live, testable funnel — Amir starts with an AI B2B sales sparring partner, generates positioning and a freight-software lead magnet, designs the page in Paper, ships it, and launches an A/B test in about 30 minutes.
The real bottleneck isn’t building pages anymore — it’s maintaining context and getting customers — Amir says the missing piece after “vibe coding” is tracking ICP, positioning, growth experiments, and results over time so you can compound decisions instead of starting from scratch every session.
Good AI design comes from taste, references, and constraints — not magic prompts — instead of saying “improve the design,” Amir feeds Claude screenshots, style guides, and Tailark components, then uses prompts like “subtle animation” and “consistent layouts” to avoid the generic purple landing-page look.
Paper matters because it fills the gap between Figma and raw code — Amir frames it as the missing middle layer where designers can iterate visually while staying connected to Claude Code, rather than tossing static mocks over the wall or editing everything blindly in code.
Humblytics’ big trick is no-code experimentation on live pages — after installing analytics, Amir has Claude create a headline test using real site data like 40% average scroll and 25% bounce rate, then dynamically swaps copy such as “Every lost deal started with an objection your rep wasn’t ready for” without redeploying code.
Amir and Greg think this stack is an arbitrage window for marketers and agencies — Greg compares it to early Facebook ads at $0.05 CPC, arguing that most people still don’t know how to use these tools together, which creates room to sell landing-page optimization and managed growth services for $5k-$20k per month.
The Breakdown
The promise: no more ugly vibe-coded landing pages
Greg opens by teeing Amir up to give “all the sauce”: how to go from idea to validation to a polished landing page that actually makes money. The hook is blunt and relatable — plenty of tutorials promise this, then leave you with a weird purple vibe-coded site instead of something you’d trust with a customer.
Starting from context, not chaos
Amir’s first big point is that finding an idea isn’t enough; you need a living record of how the business evolves. Using Idea Browser’s MCP connection to Claude Code, he pulls project files like offer definition, ICP, positioning, and growth strategy for an “AI sparring partner for B2B sales teams,” narrowed to freight software reps practicing objection handling.
Turning the idea into a lead magnet
He uses an Idea Browser skill called “lead magnet legend” to create a specific offer: “5 objections that kill freight software deals.” That becomes the foundation for a PDF guide and landing page, and Amir keeps emphasizing why this matters: everyone can ship pages now, but the hard part is knowing what to pitch, who to target, and how to get early customer confidence.
Why Paper is the missing middle between Figma and code
Amir explains Paper as the workflow layer that was missing once everyone started building directly in Claude Code. Instead of static Figma mocks handed to engineering, or messy blind iteration in code, Paper lets you ideate and refine sections visually while staying tightly connected to the codebase.
How to make AI-generated design actually look good
His design advice is practical: give Claude reference images from sites you like, ask it to extract the design system, and reuse that style guide in future sessions. He also pulls in Tailark UI components as references, showing how a solid existing block plus careful prompting like “subtle animation” and “consistent themes” can turn raw output into something that looks genuinely polished.
The live build, plus a tiny ant cameo
As Claude builds the lead magnet page section by section inside Paper, the whole thing feels delightfully scrappy — including a brief detour about an ant wandering across the table and “getting the direct sauce.” It’s a funny moment, but it also reinforces the episode’s energy: this is fast, informal, and very much happening live.
The bigger thesis: the terminal is becoming the interface for work
Greg and Amir zoom out from landing pages to a broader claim they’ve apparently been making for a while: first it looked like Cursor was the interface for work, now it’s the terminal itself. They riff on agents getting wallets, inboxes, and website access, and Amir predicts more agents will visit websites than humans; Greg cites Gartner saying 20% of internet commerce by 2030 could involve agents making purchases.
From page views to revenue experiments
Once the page is live, Amir installs Humblytics and has Claude create an A/B headline test based on real behavioral data. The system pulls metrics like 40% average scroll and 25% bounce rate, creates a variant headline, updates the page without redeploying code, and starts tracking clicks and conversions — which Amir says is exactly how his team now runs personalized campaign pages and managed growth work for customers paying $5k to $10k+ per month.