Skip to content
AI Dev Cases
Laper

"I Didn't Write a Single Line of Code" — A Former Screenwriter Built a 100,000-Line AI Screenwriting Tool With Claude Code Alone, Hitting $40K+ MRR in 8 Months

Zhao Chunxiang, a former screenwriter at Han Han's film studio, built Laper — an AI screenwriting tool — after a suggestion from a Meituan co-founder. Its 100,000+ line codebase was written with zero manual coding, entirely by Claude Code, and it reached $40,708 MRR with 1,165 paid users 8 months after its invite-only beta opened.

Published: Sep 3, 20264 min readPrimary-source verified · 4
Monthly (est.)
$41k/mo
Time to grow
8 months
Launched
2025
赵纯想赵纯想@chunxiangai"I Didn't Write a Single Line of Code" — A Former Screenwriter Built a 100,000-Line AI Screenwriting Tool With Claude Code Alone, Hitting $40K+ MRR in 8 Months

Key takeaways

  • Turn a dull, industry-specific friction you've personally endured for years into a product — with a resolution only an insider has
  • Import software-engineering concepts (version control, filesystem) into a completely different creative domain to differentiate
  • When an expert's suggestion stings in the moment, reinterpret it later against your own specific expertise instead of dismissing it for good

The pain point, and how they found it

Most screenwriting software forces writers to fight invisible labor: industry-standard formatting. Characters, scenes, and locations end up tracked in separate spreadsheets and notes, so even a small edit means manually chasing consistency across the whole script. Zhao Chunxiang lived that friction for years as a working screenwriter — an unglamorous pain essential to professional screenwriting that nobody wanted to fix — and it became Laper's starting point.

Background & product

Laper bills itself as a "creative OS" for screenwriters, directors, and storyboard artists — an AI screenwriting and pre-production tool that handles scene auto-complete, dialogue and action-line suggestions, automatic character-relationship mapping, and storyboard creation inside one editor. Its defining design choice: characters, locations, and scenes are each treated like independent files with their own Git-like version history — a "filesystem for storytelling." AI isn't a separate chat window you have to context-switch into; it's woven into the editor as auto-complete that should feel like typing faster.

It was built by Zhao Chunxiang (赵纯想, birth name Zhao Xiangyu). Before writing code, he worked for several years as a screenwriter — and for about a year as an actor — at Tingdong Pictures, the film studio founded by Han Han, doing shot-by-shot script breakdown work and screenwriting. He later became an indie developer and had a hit with Belly Book, an LLM-powered food-diary app.

Laper's origin story starts with an unlikely introduction: a mutual contact put him on a video call with Wang Huiwen, a co-founder of Meituan who went on to found the AI startup Guangnian Zhiwai. Zhao was pitching a messaging-app idea; Wang suggested he instead build an AI screenwriting tool that drew on his own screenwriting background. By his own account, he was annoyed enough in the moment to delete Wang's contact — but he later found himself searching for "AI screenwriter," and that search became the seed of Laper.

An alpha appeared in fall 2025, and an invite-only beta opened on December 18, 2025. Invite codes were distributed through an unusual mechanic: forward a designated tweet to WeChat Moments, keep it visible for at least five minutes, then screenshot it — a system that brought in more than 1,800 beta participants. A referral program also rewarded posts about the experience on X, Xiaohongshu, and TikTok with extra codes and up to 60 days of free membership.

What stands out most is the build itself. By Zhao's own account, Laper is a live, production web app with a codebase exceeding 100,000 lines — written with zero manual coding, entirely by Claude Code. The team grew from a solo founder to five people within about half a year, and roughly 8 months after the invite-only beta launched, monthly recurring revenue reached $40,708 with 1,165 paid subscriptions.

Pricing

Free Junior plan: full editor, 60 daily credits, up to 2 script projects. Paid tiers: Senior $20/mo, Elite $60/mo (2,000 AI chats/mo, 2,800 credits/mo, 20 script + 20 film projects), Master $100/mo (unlimited AI chats, 6,000 credits/mo, unlimited projects), and Legend $400/mo for professional productions. Annual billing is 80% of 12 months' worth. (Confirmed on the official pricing page.)

From the founder (primary source)

Laper growth channels and tech stack

How it was built

  1. Productize the unglamorous friction of an industry at a resolution only an insider has — years writing screenplays at Han Han's film company and training in shot-by-shot study revealed that managing industry-standard formatting (scene headings, dialogue, action indentation and line breaks) is the invisible labor that actually costs writers their time
  2. Reinterpret an expert's suggestion later against your own specialty even if you reject it in the moment — irritated by Wang Huiwen's suggestion to build an AI screenwriting tool, he deleted the contact, then found himself searching 'AI screenwriter' days afterwards
  3. Import a software-engineering concept into an unrelated creative domain — treat characters, locations and scenes as independent files, each with Git-like version history, in a filesystem for narrative
  4. Don't make AI a tool opened in a separate chat window — dissolve it into the editor as autocomplete that simply feels like typing faster
  5. Delegate coding entirely to an AI coding tool — a production web app of over 100,000 lines written wholly by Claude Code with no hand-written code
What AI did
Zero hand-written code — Claude Code wrote the entire production codebase of 100,000+ lines for a founder who came from screenwriting and acting, not engineering

How it got users

  1. Build an act of public exposure into how invite codes are earned, turning signup itself into word of mouth — forward a specific tweet to WeChat Moments, leave it up for at least five minutes, and screenshot it
  2. Design that friction deliberately — it manufactures minutes in front of friends' eyes that a plain signup form never buys, attaching exposure to every single invite code
  3. Spread posts across several platforms at once (X, Xiaohongshu, TikTok) with referral points, where sharing your experience earns extra codes or up to 60 days of free membership
  4. Create scarcity through invite-only beta while admitting over 1,800 people in stages
  5. Make the story that 100% of development was delegated to AI part of the marketing itself — the development story landed harder than the product's own AI features

The repeatable playbook

  1. 1Turn a dull, industry-specific friction you've personally endured for years into a product — with a resolution only an insider has
  2. 2Import software-engineering concepts (version control, filesystem) into a completely different creative domain to differentiate
  3. 3When an expert's suggestion stings in the moment, reinterpret it later against your own specific expertise instead of dismissing it for good
  4. 4Bake 'time visible to friends' into the invite-code requirement itself, turning sign-up into a word-of-mouth device
  5. 5Spread referral incentives across multiple platforms (X, Xiaohongshu, TikTok) so posts multiply in parallel
  6. 6Hand 100% of the coding to an AI coding tool with zero manual lines, so someone without a formal engineering background can ship production-grade SaaS

Channels they actually used

The hard parts

The reported $40,708 MRR and 1,165 paid users come from TrustMRR, an independent tracker Zhao himself connected to Stripe — not a fully independent audit. Aggregator sites show an earlier, lower snapshot from the same period ($30,200 MRR, 898 subscriptions), so the numbers swing significantly even amid rapid growth. Whether there are formal co-founders, who the five-person team actually is, and where the company is really based (officially registered in Hong Kong, though Zhao's own activity looks mainland-China-centered) remain unconfirmed by independent sources.

Deep dive

【Deep dive】Here's how Zhao Chunxiang went from screenwriter to AI founder, and the design philosophy and growth mechanics behind turning Laper into "the tool screenwriters actually wanted."

Seeing the "invisible labor" because he'd lived it

Most screenwriting software sells AI text generation while ignoring a duller problem: managing industry-standard formatting (scene headings, dialogue and action-line indentation and spacing rules). Having spent years writing scripts at Han Han's film studio and drilling through shot-by-shot script breakdowns, Zhao understood that formatting management — not prose generation — was the real bottleneck eating writers' time. Because he designed it as the person who lived the pain, rather than an engineer imagining it from outside, Laper could focus squarely on unglamorous automatic formatting and auto-complete.

Importing software-engineering ideas into storytelling

Laper's core design treats characters, locations, and scenes as independent objects (files), each with Git-style version history — a "filesystem for storytelling." It applies software engineering's idea of diff-tracked version control directly to a problem every writer faces: as a script accumulates edits, character details and planted setups drift out of sync. That Zhao — not a trained engineer — arrived at this structural idea while pushing AI coding tools to their limit is itself a sign of the times: structural thinking no longer requires a CS background to become working software.

A suggestion rejected, then reconsidered

Wang Huiwen's advice — build an AI screenwriting tool that leans on your own screenwriting background — landed badly at first: by Zhao's own account, he was annoyed enough to delete Wang's contact on the spot. Yet within days he was searching for "AI screenwriter." The lesson is roundabout: expert advice you reject in the moment can still work on you later. What mattered wasn't accepting the suggestion outright, but reinterpreting it against his own specific expertise after rejecting it — the fork that led him to something only he was positioned to build.

Turning the invite code into an ad

Beta invite codes were distributed through a deliberately inconvenient mechanic: forward a designated tweet to WeChat Moments, keep it visible for at least five minutes, then screenshot it. The friction is the point — it forces a stretch of visibility in front of friends that a plain sign-up form never would, tying every redeemed code to real-world exposure. Layered on top was a referral-points system (up to 60 days of free membership) for posting about the experience on X, Xiaohongshu, and TikTok — spreading Laper's name across multiple platforms at once.

The story is the build itself

While hits like Cal AI compete on translating existing powerful AI into a product experience, Laper's most striking story isn't its in-product AI features — it's that the entire build process was handed to Claude Code. By Zhao's own account, a live, production codebase exceeding 100,000 lines was written with zero manual coding. That's a directly reproducible signal for this site's readers: AI development has reached the point where someone without a formal engineering background can ship a real, collaborative SaaS product.

Caveats on the numbers

The reported $40,708 MRR and 1,165 paid subscriptions come from TrustMRR, an independent Stripe-connected tracker — but one Zhao himself connected, so it isn't a fully independent audit. Some aggregator listings show an earlier, lower snapshot from the same period ($30,200 MRR, 898 subscriptions), consistent with rapid recent growth but also a reminder that the numbers move. Whether there are formal co-founders, who makes up the five-person team, and where the company is actually based (officially registered in Hong Kong, though Zhao's own activity looks mainland-China-centered) all remain unconfirmed across independent sources.

FAQ

Where is Laper from, and who built it?
Laper is an AI screenwriting tool built by a team that started as a solo effort led by Zhao Chunxiang (赵纯想, birth name Zhao Xiangyu), a former screenwriter at Han Han's film studio Tingdong Pictures. The operating company is reportedly registered in Hong Kong, though Zhao's own posts suggest activity centered in mainland China.
How much does Laper cost?
Beyond the free Junior plan, there are four paid tiers from Senior ($20/mo) to Legend ($400/mo), with roughly a 20% discount for annual billing. See the pricing section in the article for details.
Who wrote Laper's code?
By Zhao's own account, the 100,000+ line codebase was written with zero manual coding — 100% by Claude Code.

Cross-case “growth playbook” report (coming soon)

We're building a paid report that aggregates every case in this database: which acquisition tactics worked, in which categories, and how well — insights you can't see from a single story. Get notified first when it launches.

Related cases