HomeArticle

The former king in the app store market is on the verge of death. OpenClaw may eliminate 80% of mobile apps.

新智元2026-02-03 19:28
80% of apps will completely disappear.

Just now, Peter Steinberger, the father of OpenClaw (formerly Clawdbot), made a high - profile appearance and revealed the crazy story behind this "AI freak" that has shaken Silicon Valley. He said bluntly: 80% of the apps on our phones are going to be finished! Even more astonishing is that he sharply criticized current agents as "electronic garbage", and the real big move is actually...

Peter Steinberger, the founder of OpenClaw (also known as Clawdbot, which is well - known to everyone), dropped a bombshell in the latest in - depth interview:

"80% of the apps on your phone are actually dead, you just don't know it yet."

This is not alarmist talk.

When your AI friend can help you book a flight, fix bugs, and even tease you mercilessly when you sneak a late - night KFC, do you still need those fancy apps?

Peter Steinberger, this geek who was jolted out of his "retired life", is staging a real - life version of "Black Mirror" for the world with his "audacious" OpenClaw.

And the starting point of all this is actually just because he wanted to "play on the computer while lying in bed"?

OpenClaw is not an assistant, but a "weird friend" in your computer

Many people think that OpenClaw is just another ChatGPT clone?

Completely wrong!

Peter pointed out sharply in the interview: "ChatGPT is dancing in shackles, while OpenClaw is a monster breaking free from its chains."

Imagine having a new friend who is extremely smart, has unlimited resources, and lives right in your computer.

It can not only chat with you, but also directly control your file system, your terminal, and your browser.

As long as you give it permission, it can do anything you can do.

Peter shared a plot that seems straight out of a science - fiction movie:

While on vacation in Morocco, someone sent him a screenshot of a bug on Twitter. He was too lazy to turn on his computer, so he just sent the screenshot to OpenClaw.

What happened then?

OpenClaw read the screenshot, understood where the bug was, automatically checked the Git repository, fixed the code, submitted a commit, and even went to Twitter to reply to that fan: "Hey, it's fixed!"

Throughout the process, Peter just moved his fingers.

This is not programming, it's magic!

Moreover, OpenClaw has learned to "act on its own initiative".

Once, Peter casually sent it a voice message (he hadn't even written the code for voice processing at that time!).

As a result, OpenClaw rummaged through the file headers on its own, found that it was an audio format, immediately looked for FFmpeg on the computer for transcoding, found that there was no Whisper.cpp, and then went to find OpenAI's API Key...

Finally, it calmly replied to Peter.

Peter was stunned at that time: "I didn't teach you how to do this!"

This is the terrifying part of OpenClaw: extremely resourceful.

80% of apps will completely disappear

Peter has connected OpenClaw to everything on his computer - email, calendar, files, lights, music, camera... even the smart door lock.

"It can control the door of my house and even lock me out," he said. "This has reached the level of HAL in '2001: A Space Odyssey'."

Most of the connections to these devices were made by OpenClaw itself - by automatically discovering APIs and keys in the system, configuring them autonomously, and even automatically debugging. If there is no ready - made API, it can search for information, use the browser to simulate clicks, and complete the operation step by step.

For example: Peter asked OpenClaw to help him complete the check - in operation on the airline's website. OpenClaw automatically found his passport photo, extracted the information, filled in and submitted the form, solved the captcha, and even clicked off the "I'm not a robot" verification -

Because it controls a real browser, anti - crawler systems can't recognize it at all.

You may think that AI can only chat and write code, but in fact, as long as you equip it with the "skills", it can do anything for you. Just like the all - powerful robot you fantasized about as a child.

Users' imaginations are greater than developers'

Peter admits that he is not the best at "playing" with OpenClaw.

Many users in the community have turned OpenClaw into a "family member", "roommate", "assistant", and even helped "friends who can't write code" complete automated operations.

Users have a wide variety of ways to use it:

Join family groups and send daily reminders;

Create GitHub issues to track project progress;

Automatically synchronize points of interest on Google Places;

Collect content on Twitter and automatically generate to - do lists;

Monitor health data and record whether you have enough sleep;

Come with a 1Password vault to manage your (and your own) passwords.

Some users have asked OpenClaw to keep accounts, cook, arrange fitness plans, help buy groceries, and even monitor their children's homework.

OpenClaw remembers what you like to eat and whether you exercise every day. It can remind you to eat less and exercise. Even if you send a photo of your dinner, it can automatically recognize the food, estimate the calories, and write a "teasing reminder": "If you keep eating like this, you won't even be able to see your belly button, let alone six - pack abs."

OpenClaw is gradually swallowing up your app ecosystem

Peter believes that: "80% of the apps on your phone may be replaced by AI assistants in the future."

The reason is simple ──

Why use a dedicated app to record sleep when you can let OpenClaw directly read the device data?

Why use a fitness app to punch in when you can let OpenClaw remind you "don't stay up late" when you don't sleep at night?

Why use a shopping app to compare prices when you can just send a message to OpenClaw saying "Help me find a cost - effective hair dryer"?

"The interface of OpenClaw is just chatting. It's naturally much more comfortable than those apps with lots of clicks. More importantly, it knows who you are and your preferences. You don't have to set everything up again every time."

The strongest thing about OpenClaw is not what it can do, but that it can remember you. Every interaction will make it understand you better and do better.

Minimalist AI programming

Peter created "OpenClaw", but the AI programming workflow is surprisingly simple, completely going against all the "best practices" of AI programming:

No need for a planning mode. The planning mode is just a stop - gap measure for the old model. He directly gives the AI the prompt "Let's discuss", and then starts a conversation.

Don't use MCP. "Most MCPs should use the CLI. The intelligent agent will try the CLI, get the help menu, and then everything will go smoothly."

Don't use an orchestrator or sub - agents. He directly uses multiple terminal windows.

Compared with Claude Opus, the programming intelligent agent Codex can handle larger code libraries and make fewer errors.

Peter said: "Don't waste time on RAG, sub - intelligent agents, Agents 2.0, or other things that are mostly just for show. Just talk to it directly."

Peter said bluntly, "After OpenClaw came along, a large number of 'non - technical people' around me started writing code."

For example, his friend, an "ordinary person" who had never even used GitHub, actually started submitting pull requests (code change requests) for the OpenClaw project.

He couldn't even type terminal commands before, but now he not only writes code with OpenClaw but also submits it. OpenClaw makes him transform the world like he's holding a magic wand.

Why can OpenClaw turn "newbies" into hackers?

It's very simple. It "hides" all the complicated things from you - you don't need to know what dependencies are, what a compiler is, or what an API call is.

As long as you say "I want to do this", OpenClaw will be like an all - powerful partner and help you complete it step by step.

Peter said that this has even become a "new type of programming method": you no longer need to know specific syntax or type code, but "talk to the AI", and it will help you write, debug, and optimize the entire project.

Of course, what you need is "ideas", not skills.

In the past, knowing code was the threshold. Now, the most valuable things are your imagination, your product intuition, and your insight.

He added, "I used to specialize in iOS development, but when I switched to web development, the frustration was too strong. I even had to look up how to slice an array. But with OpenClaw, as long as I tell it the function I want, it will automatically set up the project."

Has the singularity arrived?

Peter believes that although current AIs are smart, they are "spikily smart". They have no taste.

If you don't add human intuition, aesthetics, and vision to the loop (Human - in - the - loop) and just let the AI run aimlessly, it's like letting a group of monkeys type on a typewriter. They won't be able to write Shakespeare.

"Vibe Coding" (atmospheric programming) is the right solution.

Don't try to write perfect code from the start. Instead, have a chat with the AI.

"Hey, I want to do this. What do you think?"

"Give me a few solutions."

"This one won't work. Try that one."

In this process, your ideas will become clearer and the AI's output will become more accurate. This is the future workflow of human - machine symbiosis.

Peter finally mentioned that his business partner, a former lawyer who doesn't understand technology at all, is now actually submitting pull requests (code merge requests) to him!

This was absolutely unimaginable in the past.

AI has given ordinary people "superpowers". Language is no longer a barrier, and logic and creativity are the key.

Although OpenClaw still has a bit of "weirdness" and "uncontrollability", this interaction mode of "treating AI as a friend rather than a tool" is undoubtedly opening the door to a new world.

As Peter said: "It's like you can train a new brain from scratch."

In addition to requiring technical skills and patience, OpenClaw is not suitable for running on public computers because it may leak personal information. There is also a risk of "prompt injection" -

Hackers may send emails or files that deceive the AI model, causing it to blurt out confidential information.

ZeroLeaks gave it a score of 2 (out of 100);

The information extraction success rate is as high as 84%;

The success rate of the prompt injection attack is 91%.

The system prompt was leaked in the first round of the conversation.

"This is a trade - off with security," early user Peguine admitted.

Even so, he still hopes to deploy OpenClaw in a bolder way - using it to manage his family's small and micro - enterprise. He is teaching his father, who runs a tea company, to use OpenClaw to process invoices, track inventory, and contact customers. "The beauty of it is that it's a general - purpose system," Peguine said. "I think it can run a business."

We are in the rift between the collapse of the old era and the establishment of the new era.

Are you ready to welcome your "weird friend living