Original · Unique Research · 2026-07-13
Editor's note: This is the original Chinese author's interview with Wu Di and its framing. This English rendition retains the full text in source order, including all 15 Q&A items. The interviewee's product claims, customer cases, and market views are his self-reports, attributed to him and not independently verified. Person, company, and product names are preserved as source attributions.
AI Industry Observation
"Take the ordinary person's thought of 'if only there were a little tool' and directly turn it into a small app that's truly usable, playable, and shareable."
Shanghai, an ordinary weekend. A 10-year-old primary schooler opened Lingzhu AI and said one sentence by voice: "I want to make a Pac-Man game; the beans are walled off and can't be reached, so let Pac-Man pass through walls, and let players draw their own map." A few days later, he posted the link to "Super Pac Adventure" in his class group. Classmates clicked and played — controlling Pac-Man with arrow keys, with a wall-penetrating skill, and opening a map editor to design their own maze levels.
This child can't write a line of code. He doesn't know what Python is, has never heard of React, and doesn't understand game engines. But he had a thought in his head — "if only I could play it this way" — and the thought became a thing.
This is what Lingzhu AI does. Its CMO Wu Di (吴荻) explains in one sentence: "Take the ordinary person's thought of 'if only there were a little tool' and directly turn it into a small app that's truly usable, playable, and shareable."
Sounds simple; doing it is unusual.
Not Helping Programmers Write Code, But Opening a Door for a Billion Non-Coders
You may ask: aren't there already tons of AI coding tools? Cursor, GitHub Copilot, v0... what's different about Lingzhu?
Ultimately, AI coding tools help programmers. However powerful Cursor is, you must first understand programming logic. However smart Copilot is, you must sit in front of VS Code. These tools' essence is "let people who can code code faster."
Wu Di puts it plainly: "AI coding, AI site-building, essentially still help 'people who already know how to develop' improve efficiency."
But Lingzhu faces another group. China has 9.4 million programmers, already well served. Lingzhu wants the billion-plus ordinary people — with ideas, scenarios, needs, but who can't code and don't want to learn.
"The true revolution of AI programming tools isn't making programmers ten times more efficient, but letting people who can't program have 'creation' ability for the first time."
Wu Di positions Lingzhu as a "zero-basis AI creation and interactive app platform." Zero-code generation is only the entry, not the end. What's the end? You make something, then someone uses it, plays it, modifies it — a flowing creation scene.
He gave an analogy that stuck with me: "Before short video, shooting and editing were professional skills. But now ordinary people expressing emotion, opinion, and life through video is instinct. I think AI app creation will go through a similar change."
In other words, Lingzhu isn't "lowering the programming barrier." It's trying to make "making an app" as natural as shooting a short video.
Three-Step Creation: Say It Out, Modify It, Throw It Out
Lingzhu compresses the creation flow into three steps: input-to-create, conversational modification, one-click publish and share.
Sounds flat at first. But once you truly understand user behavior, you find hidden precise insight into user psychology.
The Lingzhu team initially thought the hardest step was the first — "saying the requirement." But they found that the moment users see an input box, they naturally state their requirement. "Make me a bookkeeping tool," "I want to calculate my BMI," "make me a Pomodoro timer" — no teaching needed; everyday language suffices.
The hard part is the next two steps.
The second step is continuous modification. AI can't guess the perfect product in your head in one shot. You adjust colors, change button positions, add a leaderboard, change tone — "more snarky," "make this button bigger," "add a share function." These seemingly small changes are actually the user polishing a generic template into "my thing."
"For ordinary people, conversational modification matters more than one-shot generation. Because real creation rarely comes out perfectly in one pass."
What does this mean? The user isn't "submitting a requirement" — that's traditional software-outsourcing thinking. The user is "polishing a work." The psychological experience is completely different. One is client-to-vendor; the other is creator-to-work.
The third step is harder — publishing. Publishing has a psychological threshold. Most people make something and dare not post, fearing it's ugly, no one will use it, someone will say "what is this." So Lingzhu makes publishing as light as possible: one-click generate a link, click and use, no review, no server config.
From a growth view, whether the "generate-modify-share" loop runs through is Lingzhu's lifeline. Once a user completes the whole chain, they move from "trier" to "creator." The two differ by more than an order of magnitude in retention, activity, and willingness to spread.
Not Whichever Model Is Cheapest; Different Steps Use Different Models
Supporting this user experience is a not-easily-seen model strategy underneath.
After integrating DeepSeek V4, requirement-analysis efficiency and effect improved markedly. But Lingzhu's approach isn't "switch everything to DeepSeek," but split the creation flow into steps, matching each step with the most suitable model.
"We don't simply use whichever model is cheapest; we do model scheduling by scenario. We have our own benchmark, testing and comparing the final app scores in our Agent scenarios, to choose the more suitable model."
Translation: in the same creation flow, requirement understanding may use one model, UI generation another, code generation a third, debugging and optimization a fourth. Users perceive nothing, but behind is a whole engineered model orchestration.
The cleverness: it doesn't use AI as "one brain" but as an assembly line — each station staffed with the most suitable person. Better cost, better effect.
Start Something From Zero? Most People Can't Even Start
OK, now suppose Lingzhu made making an app simple enough. Next question: will users really create from zero?
Answer: no. At least the vast majority won't.
A blank input box has pressure. You stare at it; you may have a vague idea, but when you go to write — "what was I going to make?" Creative startup is itself a barrier.
But conversely, if you open the app plaza and see someone else's finished "BMI calculator," your reaction is completely different. You think: "can I change this to my version? Add body fat rate? Change the colors? Switch prompts to Chinese?"
Wu Di calls this mechanism "magic modding" (魔改). Secondary creation lowers not just the technical barrier but the psychological barrier and creative-startup barrier.
"Back when Musical.ly, TikTok's overseas predecessor, launched 'shoot the same style' and 'lip-sync,' that was actually magic modding, secondary creation, further pushing creation rights to a broader population. That's what gave short-video platforms their breakout moment."
The short-video breakout wasn't the first masterpiece, but the moment the "shoot the same style" button appeared — anyone could borrow someone else's idea to make their own version.
Lingzhu's logic is identical: one work → N people modify → N new works → more people modify. This isn't single spread; it's chain fission.
Magic Modding Is Deeper Than Secondary Creation: What You Modify, Others Can Still Use
What is short-video secondary creation? "I'll shoot one too." You shoot, others watch, done. It's a content-consumption chain.
AI-app secondary creation is different. "I'll make one too, but it can still be used by others." You modify someone's bookkeeping tool's categories, change the currency, add charts; this new tool itself is a usable product. The next person seeing it can keep magic-modding.
Works here aren't static content but a living, operable, re-creatable node.
Wu Di calls this a "social engine." Note, not "social feature" — adding likes and comments is a social feature. Engine means it's the power source of the whole platform. Each magic mod is a fission; each fission grows the creator base and app base.
"What users truly need isn't just a 'functionally correct' tool, but a tool 'more like me, more understanding of my scenario.'"
What does this mean? A generic bookkeeping tool may be worth 3 points to me. But my magic-modded version — adding a pet-expense category, switching to a pink theme, renaming "expense" to "money spent" — may be worth 8 points to me. And this 8-point tool may be worth 7 points to another cat owner. He can keep modifying.
The essence of magic modding is turning "creation" from a few people's patent into most people's instinct. Lingzhu's app plaza isn't a static shelf but a flowing creation scene. You can use, play, favorite, or magic-modify.
Wu Di defines Lingzhu as a "consumable + moddable" interactive app platform. Users not only consume what others made but, under the original author's authorization, magic-modify — change theme, gameplay, settings — and republish to the plaza.
"This is something Douyin can't do," he says.
Longer term, Wu Di hopes Lingzhu becomes an AI-native lightweight-app community. Not "the Douyin of the AI era" — one step more than Douyin. Not "a better AI coding tool" — it doesn't target programmers.
"We target not China's 9.4 million programmers, but the broader billion-plus ordinary users beyond them."
This is a big game.
Programming Rights Belong to Everyone: a Third Grader's 520 Gold Award
The Peng from Shanghai Pudong Vanke Experimental Elementary mentioned earlier later won gold at Lingzhu's 520 AI Creation Contest and gave an interview to People's Daily.
But more worth saying than the award is what he made — a Pac-Man game with a custom map editor. What does this mean? He didn't just make a playable game; he made a creation tool that lets others keep creating. Iterated over ten times, each time polished through conversation with AI.
You see, a 10-year-old proved: programming rights can belong to everyone. Not someday in the future, but now. What Lingzhu does isn't "teach kids programming" but directly pulls programming out of the process. Whatever you want, say it, and it exists.
Zero-Code Site Building Is No Dream: an Auto-Rescue Boss's 30 Minutes
Zhang Hongzhao (章红照) has done auto rescue in Guangdong for 20 years, running Guangtong Auto Rescue Co. He knows what his business needs — a website showing road rescue, auto repair, battery-swap services, so clients can find him.
How long did this take? Two to four weeks. How much? Several thousand to over ten thousand. Plus finding a site-building company, buying servers, communicating requirements, repeated revisions.
With Lingzhu? 30 minutes.
30 minutes, Zhang Hongzhao went from zero to a complete website. He doesn't need programming, doesn't need to know HTML, doesn't need domain resolution. He only needs to know what he wants — what goes on this button, what this page says.
Honestly, people like Zhang Hongzhao are everywhere in every county and street of China. They are coaches, shop owners, consultants, craftspeople. Their need isn't "learn programming" but "I have a need; I want it to become reality fast."
Knowledge Becomes Product Directly: a Urologist's "Sperm Game"
This story impressed me most.
A urologist in practice for years has piles of professional knowledge in his head — how sperm are produced, how they move, what affects them. This knowledge matters to the public, but no one wants to read a paper.
He used Lingzhu to make a level game called "The Romantic Decline of Sperm." Silver award at the 520 Lingzhu AI Creation Contest.
Taste the disruption here.
In the traditional flow, a medical-popularization product needs what people? Product managers research requirements first, designers make UI and characters, programmers write code, doctors provide content — four or five links, a months-long cycle, rounds of communication loss.
Now? An industry expert directly creates a digital product. No intermediary. The doctor has ten thousand "if only there were something to let people understand this" thoughts in his head; now the thought can directly become a work.
What does this mean? Experts in education, medicine, law, finance... every industry can turn their professional knowledge into an interactive digital product. Game is only the form; the essence is that the barrier to "knowledge productization" is torn down.
From One Idea to 5,000 Sets: a Sophomore's Board-Game Business
A Shanghai sophomore used Lingzhu to make a set of board-game cards. Made it for fun, put it in the plaza.
Then at a card expo a manufacturer spotted it, and 5,000 physical card sets are about to launch, expected revenue near 500,000 RMB.
Can you imagine? A sophomore, no team, no funding, no startup experience, just an idea + Lingzhu, and the work directly entered the commercial chain.
"This is the scariest part. After the creation barrier disappears, a good idea is only one card away from the market."
Before, you had an idea and first had to learn programming, find co-founders, make a prototype, gather resources. Now? Make it first; once made, someone uses it; once used, someone pays.
Four cases, four kinds of people. A 10-year-old child, a 40-something boss, a practicing doctor, a sophomore. Their only common point — none are programmers.
The App Plaza: Not a Shelf, a Flowing Creation Scene
OK, now the question. So many people made so many things; how do more people see them?
Lingzhu's answer is the "app plaza." But Wu Di repeatedly stresses: this isn't a static shelf.
What does this mean? A traditional App Store is essentially a shelf — you list, I pick, transaction done. Mini programs are closer to service and transaction scenes. Short-video platforms emphasize content flow, recommendation, and consumption.
Lingzhu wants to fuse these forms —
It has the App Store's "usability": what you get is a usable, playable, problem-solving tool, not a video you watch and forget.
It has mini programs' "lightness": no download, click and use, use and go.
It has content platforms' "flowingness" and "secondary creation" mechanism: you can use it, modify it, republish after modifying, and someone keeps modifying after you post.
So the plaza's core isn't "display" but "circulation." One work enters, magic-mods into ten, then a hundred. Good works are discovered in flow, amplified in modification.
Not Only Chasing Hits: Dozens of Users Can Still Be Huge Value
Platform builders all have a temptation — stare at traffic hits, chase hot spots, push the most hard-hitting thing.
"AI-app value can't only be judged by mass hits. Some apps may be used by only a few dozen people, but they're especially useful to some class, some store, some family, some industry expert — that's also value."
What does this mean? A classroom roll-call tool used by only 30 people may be worth 10 points to that teacher. A class game played by only 50 may be the happiest thing those 50 kids have all week.
So Lingzhu's plaza will do categories, rankings, recommendations, operations, vertical topics, modding chains, and look at user-behavior data. But the final goal isn't to manufacture a masterpiece on a leaderboard, but to let every person with a need find, use, and modify their own thing.
Small, beautiful works also need their own users. This logic differs from short video — short video is "the whole country watches the same hit"; AI apps are "everyone has their own tool."
One Step More Than Douyin: Consumable and Moddable
Lingzhu goes one step more than short-video platforms. Which step exactly?
Short video's chain is: consume → like → comment → done. Lingzhu's chain is: consume → magic-modify → re-create → republish → re-consume.
You see a great video on Douyin, think it's impressive, like, next one. You see a bookkeeping tool in Lingzhu's plaza, think it's good but want to change categories — you change, publish, your version is used by more people, and someone keeps modifying your version.
"Consumable + moddable" — this is Lingzhu's core differentiation. Not just use, but create after using. Not just watch, but get hands-on after watching.
What does this extra step add? It's the continuous devolution of creation rights. Short video devolved creation rights from professional teams to ordinary people, but creation ended once. Lingzhu turns creation rights into a daily ability — each interaction may be a re-creation.
Next Three Years: From "Playable" to "Usable" to "Value-Creating"
On the future, Wu Di gave three judgments.
First, generation quality will rise fast. Now people still discuss "can it generate"; future focus becomes "how accurate, how good, how like what I want." This inflection is arriving.
Second, ordinary-user scale will expand. AI app creation won't stay in tech circles; it will enter education, stores, enterprise training, personal creation, community activities, and many real scenarios. Today you see a 10-year-old making a game on Lingzhu; tomorrow you may see your store clerk using Lingzhu to make a customer-registration tool.
Third, the app plaza and commercialization ecosystem will gradually form. As more good works flow on the plaza, commercialization paths naturally appear — like that college student's board-game cards.
Three trends stack; the path is clear: AI-generated apps move from "playable" to "usable," then to "value-creating."
Not Eliminating Professionals, But Eliminating "I Can't"
Here you may ask: what about programmers? Designers? Product managers? What do they do?
Wu Di's answer is direct: "AI lowers the basic creation barrier, not eliminates professional ability."
Programmers shift from lots of repetitive implementation to complex systems, architecture, security, performance, high-quality engineering. Designers shift from drawing pages to aesthetic systems, experience judgment, brand expression. Product managers shift from writing requirement docs to deeper scene understanding, problem definition, mechanism design.
In other words, bottom-level brick-moving work keeps shrinking, but top-level design demand grows. Not that professionals aren't needed, but professionals can do more important things.
Those truly needing "liberation" are people who already had professional ability but were blocked by the programming barrier — doctors, teachers, consultants, craftspeople, students. The professional value in their heads finally has a direct expression channel.
But the bigger challenge isn't even technical.
If technology is no longer the problem, what slows universal app creation?
"Not technology itself, but user habits, trust, and platform mechanism. Most people still don't believe they can 'make an app.' The concept itself is outside their cognition."
They're used to the flow "have a need → find a professional → wait for result," never imagining "have a need → say it → handle it yourself."
This cognitive shift takes time. Like when short video first appeared, some thought "who'd watch what I shoot?" And then? It became instinct.
So much of Lingzhu's future work isn't tech hardening, but making users believe — you can do this too.
The Real Moat: Two Flywheels
Last question: what's Lingzhu's moat?
Models matter, but models aren't a moat — you lead today, someone switches to a new model tomorrow and catches up.
Wu Di says the real moat is a set of "cans": can it keep raising creative fidelity, can it turn ordinary people's vague expression into clear requirements, can it sediment a self-evolving Skill library through user creation and modification, can it form a modding ecosystem, can it make good works discoverable.
These add up to two flywheels.
Data flywheel: every app a user makes automatically sediments requirement docs, templates, Skills into the knowledge base. The next user wanting something similar, the system has "seen" it — steadier quality, faster generation, higher fidelity.
Community flywheel: the more people create and magic-modify in the plaza, the more good works are produced. The more good works, the more people come to use. The more people use, the more people start creating.
Two flywheels together; over time the platform's capability naturally strengthens. Wu Di said one line: "We are friends of large models, and friends of time."
Translation: Lingzhu doesn't compete with large models; it builds a layer on top. But time is on Lingzhu's side — the longer used, the more it understands users, the more users depend on it.
In Closing
After the interview I kept thinking about a question.
Over the past 30 years, the internet created a gigantic world, but the admission ticket was "code." 9.4 million people got the ticket; a billion-plus stood outside. They watched the wonderful inside, couldn't get in, and didn't expect to.
What Lingzhu does is tear down this door. Not sell the ticket cheaper, but make the ticket unnecessary.
When a 10-year-old child, an auto-rescue boss, a urologist, a sophomore, can all make something with just an idea, letting others use, play, even earn money —
The meaning goes far beyond "AI coding tools."
"What Lingzhu truly changes isn't how code is written, but returning app-creation rights from the few who know code to every ordinary person with ideas, needs, and willingness to express. The people outside finally come in."
Selected Interview Q&A
Q1: Without an official intro, how would you explain Lingzhu AI to someone who doesn't know it?
Wu Di: I'd say Lingzhu does something simple: take the ordinary person's thought of "if only there were a little tool" and directly turn it into a small app that's truly usable, playable, and shareable.
Before, your first reaction to an idea might be "I can't code, so never mind." But now you just say the idea out, and Lingzhu helps you understand the requirement, generate the app, keep modifying, and finally send it to others to use.
So Lingzhu doesn't make ordinary people learn programming; it lets ordinary people's ideas have a chance to become a real product for the first time.
Q2: Many understand Lingzhu AI as an "AI coding tool" or "AI site-building tool." Where's this understanding inaccurate?
Wu Di: It's not entirely wrong, but it only sees one thing Lingzhu can do, not the whole picture.
AI coding and AI site-building essentially still help "people who already know how to develop" improve efficiency. But Lingzhu faces not programmers, but ordinary people with ideas, scenarios, needs, who can't code.
We care less about "how fast code is written" than whether an ordinary person can complete the whole loop from idea, generation, modification, publishing, sharing, to being used and even second-created by others.
So I often say zero-code generation is only Lingzhu's entry, not its end. What Lingzhu truly wants is a zero-basis AI creation and interactive app platform.
Q3: Lingzhu's core slogan is "turn the idea in your head directly into a product." Do you think AI app creation's core barrier is coding ability, or the ability to express requirements clearly?
Wu Di: The past biggest barrier was of course code. Without coding, you couldn't land an idea.
But now the new barrier is: can you state your requirement clearly.
There's a misconception. Many think ordinary people don't understand products or requirements; actually they just haven't had product-manager-style expression training. For example, a boss wanting a website system, a teacher wanting a classroom game, a parent wanting a child's math-practice tool — their needs are very real, just expressed in everyday language.
What Lingzhu does is translate these everyday expressions into requirements AI can understand and execute. We don't ask users to become product managers; we help them make vague ideas clear and finally land them.
Q4: Internet product development used to be a few people's ability; will AI today make "making a little tool" an ordinary person's daily expression?
Wu Di: I think so.
Before short video, shooting and editing were professional skills. But now ordinary people expressing emotion, opinion, and life through video is very natural.
I think AI app creation will go through a similar change. In future, not everyone must make a complex app or start a business, but when everyone hits a concrete problem in life, study, or work, they can conveniently make a little tool.
For example, make a quiz game for the class, an event-signup page for the store, an info-registration form for clients, an exclusive practice app for a child. These used to be too small, too fragmented, too long-tail, so traditional development wouldn't serve them. But these needs are real.
Lingzhu's meaning is truly empowering ordinary people with AI, letting needs once not worth developing also be realized.
Q5: Lingzhu's signature is the "three-step creation": input-to-create, conversational modification, one-click publish and share. Which step did you find ordinary users hardest to cross?
Wu Di: At first we thought users' hardest step was the first, describing the requirement. But actually, many users, the moment they see the input box, naturally say "I want to make such-and-such."
The hard part is the second and third steps.
Second is continuous modification. After generating the first version, the user enters a new state: they realize "this can really be made," so they raise more detailed demands — colors, buttons, copy, flow, functional logic. Now they need to move from "saying an idea" to "polishing a work."
Third is publishing. Publishing has a psychological threshold. Playing by yourself is one thing; sending it to others is another. The user's willingness to publish shows they begin treating this as their work, not just an AI experience.
So from a growth view, we especially value whether users complete the "generate-modify-share" loop.
Q6: Generating an app demo is easy, but making it truly usable and shareable — where's the hard part?
Wu Di: A demo proves whether AI can understand you. A usable product proves whether it can enter real scenarios.
The hard part is mainly three places.
First, the function must be right. It can't just look like, but truly complete what the user wants.
Second, the experience must be smooth. Ordinary users won't tolerate complex operations; they judge with life experience: is this thing usable.
Third, users must have a sharing motive. Why send this thing to friends, classmates, clients? Because it's fun, useful, or expresses their own idea.
So Lingzhu must raise not only generation ability but creative fidelity, modification efficiency, publishing experience, and the plaza's distribution. A work only truly enters the platform ecosystem when used, given feedback, and magic-modified.
Q7: Is conversational modification more important than one-shot generation?
Wu Di: Yes, and I think for ordinary people conversational modification matters more than one-shot generation.
Because real creation rarely "comes out perfectly in one pass." Even professional product teams discuss, modify, and iterate. Ordinary users more so — they often only know what they truly want after seeing the first version.
Users say "change the color," "make this button bigger," "add a leaderboard," "snarkier tone" — these seemingly tiny modifications are actually them turning a generic tool into their personalized work.
Conversational modification's value is here: it makes users no longer just "people who submit requirements" but co-creators of the work.
Q8: You've integrated multiple domestic large models, including DeepSeek V4. What's the biggest change in user experience from model upgrades?
Wu Di: Model upgrades bring not a single metric change but overall better experience.
For ordinary users, they don't care which model you use behind, nor technical terms. Their most direct feeling is: does the system understand me better, is generation steadier, is waiting shorter, is modification more accurate.
After integrating DeepSeek V4, requirement-analysis efficiency and effect both improved clearly. More importantly, we don't simply "use whichever model is cheapest," but do model scheduling by scenario.
We're not cost-only; we have a benchmark to test and compare final app scores in our Agent scenarios, then choose the more suitable model. Requirement analysis, code generation, bug fixing, conversational modification, frontend — different steps may suit different models.
Q9: Lingzhu's public beta pushes "magic modding," with the idea "everything in Lingzhu can be magic-modded." Why do you think secondary creation matters more than starting from zero?
Wu Di: Because "starting from zero" is actually pressured for many ordinary people.
A blank input box makes people hesitate: what exactly do I make? How to describe it? Is this idea good enough?
But if they see an already-made app — a fitness tool, ledger, little game, quiz — they easily have a second thought: can I change this to my version?
That's magic modding's value. It turns creation from "I must think of something out of thin air" to "I see something, then have my own way to change it."
Secondary creation lowers not just the technical barrier but the psychological and creative-startup barriers.
Q10: Short-video platforms' breakout came largely from templates, challenges, and secondary creation. Will AI app creation see similar spread mechanisms?
Wu Di: I think so, and app-creation secondary creation may be one layer deeper than content secondary creation.
When Musical.ly launched "shoot the same style" and "lip-sync," that was actually magic modding, secondary creation, further devolving creation rights to a broader population. That gave short-video platforms their breakout moment.
But AI-app secondary creation differs from short-video. Short-video secondary creation is more "I'll shoot one too." AI-app secondary creation is "I'll make one too, but it can still be used by others."
For example, one person makes a "snarky bookkeeping roommate"; another magic-mods it into a "gentle finance coach"; a third into a "couples' joint bookkeeping tool." They share a skeleton but have completely different expression and scenarios.
So app creation will also see templates, challenges, same-style, modifications, leaderboards. The difference is works here aren't just for viewing; they can be clicked, operated, re-created.
Magic modding is our social engine. One work, N people modify it, producing N new works, attracting more people to modify. This isn't single spread; it's chain fission.
Q11: From a generic fitness tool magic-modded into a postnatal-recovery trainer, from a ledger magic-modded into a fun snarky bookkeeping roommate — what do these cases show?
Wu Di: They show what users truly need isn't just a "functionally correct" tool, but a tool "more like me, more understanding of my scenario."
Generic tools solve the greatest common divisor, but people's needs are often very specific. Postnatal recovery isn't the same as ordinary fitness; snarky bookkeeping isn't the same as ordinary bookkeeping.
These segmented needs were hard to satisfy before, because development cost is too high and market size may not be large. But AI app creation makes such long-tail needs realizable.
So I think users don't not need tools; they're not satisfied with one-size-fits-all tools. They want tools with their own tone, their own scenario, their own life traces.
Q12: Lingzhu supports one-click independent links and listing to the plaza. How do you see lightweight-app distribution in the AI era?
Wu Di: I think it's neither like a traditional App Store nor exactly a mini program, nor just a short-video content flow.
Traditional App Stores are heavier, with long dev cycles and high listing thresholds. Mini programs are closer to service and transaction scenes. Short-video platforms emphasize content flow, recommendation, and consumption.
AI-era lightweight-app distribution may fuse these forms: it has the App Store's "usability," the mini program's "lightness," and content platforms' "flowingness" and "secondary creation" mechanism.
Lingzhu's app plaza isn't a static shelf but should be a flowing creation scene. You can use, play, favorite, or magic-modify.
Q13: Some market voices call Lingzhu "the Douyin of the AI era." How do you see this?
Wu Di: As a metaphor it works, but it can't be simply equated.
Where it's accurate: Douyin lowered video creation and distribution barriers, making ordinary people content creators; Lingzhu hopes to lower app creation and distribution barriers, making ordinary people lightweight-app creators.
Similar points include lowered creation barriers, flowing content, secondary-creation mechanisms, expanded ordinary participation.
But they're essentially different. Short video is mainly viewing and consumption; works on Lingzhu are interactive, usable, moddable. It's not one-way content but interactive apps.
We do one step more than Douyin. We're a "consumable + moddable" interactive app platform. Users not only consume what others made but, under the original author's authorization, "magic-modify" — change theme, gameplay, settings — and republish to the plaza. This is something Douyin can't do.
It doesn't make you kill time; it makes you create things.
Q14: Lingzhu's users cover students, teachers, medical staff, small-business owners, designers. Who did you observe first truly using it at high frequency?
Wu Di: We see several representative groups.
Students and teachers were moved very early. Students naturally try and share; teachers have lots of teaching scenarios — classroom interaction, exercises, little games, knowledge cards.
Designers and lawyers are also interesting. They demand high expression and efficiency; they keep modifying and polishing, treating Lingzhu as part of their workflow.
Another group is small-business owners, store operators, OPC entrepreneurs. They may not understand tech, but their needs are very clear: I want a website, signup page, customer-registration tool, appointment system, display page. They aren't here to "play AI"; they're here to solve problems.
So the first moved aren't necessarily the people who know AI best, but the people with the most real needs.
Q15: For Lingzhu, what's the most important future moat?
Wu Di: I think it's not a single moat but a composite capability.
Models matter, but models aren't everything. The real moat is: can it keep raising creative fidelity, can it turn ordinary people's vague expression into clear requirements, can it sediment a self-evolving Skill library through user creation and modification, can it form a modding ecosystem, can it make good works discoverable, can it keep lightweight apps running stably and safely.
Lingzhu is a self-evolving platform — the more it's used, the more it understands how you think. So Lingzhu's moat isn't "we can call some model," but the more people use the platform, the more real needs sediment, and the better next generation is. This is a long-term data flywheel and community flywheel.
Every app a user makes automatically sediments requirement docs, templates, Skills into the knowledge base. The next user wanting something similar, the system has "seen" it, quality steadier. Every "not usable" feedback, every modification, goes into a checklist, auto-avoided next time.
That's the flywheel. We are friends of large models and friends of time. Latecomers can copy the product form, but not the data flywheel sedimented from tens of thousands of real creations already run.