As "AI programming" becomes increasingly accessible, a new wave is emerging: "micro-apps" that anyone can create and "super programmers" who can single-handedly match the output of an entire team.

For the general public, the right to software development is being decentralized, and programming is becoming as easy as sending a WeChat message. For industry practitioners, the mediocre middle layer is disappearing. Only those "super individuals" who can harness AI and be as powerful as an entire army can stay at the forefront of the trend.

Silicon Valley is frantically recruiting a new breed of people called "Cracked Engineers."

In the Bay Area under the shadow of the AI gold rush, the traditional recruitment logic is failing. J.X. Mo, the co-founder of the robotics startup Gradient, has directly canceled the company's intern recruitment program. The reason is simple and harsh - no one is "Cracked" enough.

In his view, in the era of AI-assisted programming, if a newcomer is not "Cracked" enough, not as skilled, in top form, and tireless as a gaming expert, they don't deserve an employee ID.

Today, as AI lowers the programming threshold infinitely, the global software industry is experiencing an unprecedented polarization.

On the one hand, the programming threshold is being lowered infinitely. Thanks to tools like Claude and ChatGPT, ordinary people with no coding knowledge are starting to build highly personalized "Micro Apps" through natural language. Software is changing from a "commodity" that needs to be purchased to a "tool" that can be self-made.

On the other hand, the competition among professional engineers is becoming more intense. In Silicon Valley, startups are frantically recruiting the aforementioned "Cracked Engineers." These young people use AI to maximize their personal output and try to replace an entire traditional development team single-handedly.

01 Micro Apps for Everyone: From "Subscribing to SaaS" to "Self-Made Tools"

For a long time, if you needed software to solve a specific problem, you would usually choose to download it from the App Store or subscribe to a SaaS service. But now, a new consumption habit is emerging: Build your own and use it only when needed.

Software built by non-professional developers is called "Micro apps" or "Fleeting apps." They have extremely distinct characteristics: They are highly specialized for specific scenarios, solve immediate pain points, and often have no intention of commercial promotion.

Legand L. Burge III, a computer science professor at Howard University, compares them to flash trends on social media, except this time the protagonist is the software itself - "When the need disappears, the software disappears with it."

Rebecca Yu's story is very representative.

To solve the "choice paralysis" problem when dining with friends, without any technical background, she used Claude and ChatGPT and "created" a web application called Where2Eat in just seven days. This application can recommend restaurants based on the common interests of her and her friends.

In the past, this usually required hiring a professional full-stack engineer or using a complex low-code platform.

Rebecca Yu said:

"Once I learned how to efficiently ask questions (Prompt) to AI and solve problems, the building process became much easier."

Filling the Gap between Excel and SaaS

Christina Melas-Kyriazi, a partner at Bain Capital Ventures, keenly pointed out that Micro apps are filling the huge market gap between "Excel spreadsheets" and "full-featured SaaS products." Just as Shopify made opening an online store easy back then, AI is making software development as casual as creating a spreadsheet.

People are starting to customize software for extremely specific and even "trivial" needs. Similar cases are exploding in Silicon Valley:

Medical Records: Software engineer James Waugh developed a simple log recorder for a friend with heart palpitations to show heart data to doctors.

Daily Chores: Media strategist Hollie Krause didn't like the app recommended by her doctor, so she wrote an allergy tracking app herself. She described the speed of development: "By the time my husband went out to buy dinner and came back, I had finished writing it."

One-Time Entertainment: Founder Jordi Amat developed a web game for a family holiday gathering. As soon as the holiday ended, the app fulfilled its historical mission and was shut down.

Bad Habit Tracking: An artist even developed a "bad habit tracker" to record how much shisha he smoked and how much alcohol he drank on weekends.

The Business Paradox of "Use and Leave"

Darrell Etherington, the vice president of SBS Comms, predicts that in the future, people will stop subscribing to monthly-fee tool applications and instead "be self-sufficient" using tools like Claude Code, Replit, or Bolt according to specific needs.

However, this wave is not without resistance. Although the development threshold for web applications has dropped to zero, mobile apps still face the obstacle of the "Apple tax" - the $99 annual developer account fee is too expensive for a "one-time app."

However, the capital market has smelled the opportunity. Startups like Anything (which received $11 million in financing) and VibeCode (which received $9.4 million in seed-round financing) are committed to solving the last-mile problem of "Vibe Coding" on mobile devices.

Of course, these self-developed software products have natural defects in quality, security, and maintenance, and they are doomed to be unable to be sold on a large scale. But for the creators, they don't need to serve the public; they only need to serve themselves - this is a subversion of the original supply-demand relationship in the software industry.

02 "Super Programmers" Who Can Replace a Team: "Cracked Engineers" in the AI Era

If micro apps are the result of the lowered programming threshold, then "Cracked Engineers" are the epitome of the fierce competition among professional practitioners.

The term "Cracked" comes from gaming slang, describing those players with amazing skills and in an "explosive" state.

In today's Silicon Valley, it is used to define the ideal image of software engineers in the AI era: Young (usually in their twenties), extremely eager for success, with a keen sense of technology, and able to achieve amazing output using AI tools.

J.X. Mo, the co-founder of the robotics startup Gradient, recently made a harsh decision: to cancel the intern recruitment. After interviewing the applicants, he found it "not worth the time" because no one was "Cracked" enough.

In the context of the AI gold rush, startups pursue ultimate human efficiency. Jonathan Siddharth, the CEO of Turing, believes that with the help of AI, a small and elite team can potentially achieve $100 million in annual revenue. Founders no longer need mediocre code workers who follow the routine. They need special forces.

These "Cracked Engineers" are fundamentally different from two types of people we are familiar with:

They are not "Vibe Coders": Vibe Coders often lack a solid foundation in underlying technology and are just "prompt operators" of AI. In contrast, Cracked Engineers have profound technical expertise. They use AI to greatly improve efficiency but also have the ability to review and correct AI's incorrect code. They are the riders who control AI, not the passengers.

They are not traditional "10x Engineers": The "10x engineers" favored by the previous generation of the tech circle are usually over 30 years old, work in large companies like Google, follow procedures, and may even be reserved about AI programming tools. Cracked Engineers are younger, more rebellious, not interested in the political struggles in large companies, and believe that "work is everything."

"One person can replace a team" is becoming a reality.

Ron Arel, the CEO of Intology, points out that a few people who are extremely focused and good at using Claude Code can now produce more than a 15-person team without AI assistance in the past.

Adam Gleave, the co-founder of Far.AI, reveals that one of his employees, with the help of AI, completed the prototype of a large model fine-tuning software in just a few weeks, which would have taken an open-source community a year to complete.

This high output is often accompanied by an extreme work intensity. These engineers generally accept and even advocate the "9-9-6" work mode (9 am to 9 pm, six days a week) because they know that in the AI competition, being one step behind means being eliminated.

James Hawkins, the co-founder of PostHog, describes them like this:

"They don't care about office politics, don't care about dressing up, and are even slovenly. Their work results speak for themselves."

However, this crazy pursuit of "super programmers" also hides potential problems.

Deedy Das, a partner at Menlo Ventures, observes that some young engineers are deliberately showing anti-social tendencies, using obscure language, or giving up all their hobbies for work in order to seem "Cracked." He warns:

"The most effective technology leaders are usually good at communication. This is not a single-player game."

Recruitment expert Kelsey Bishop is even more straightforward: Many founders try to cover up the flaws in their business models by recruiting a "Cracked Engineer." "They treat it as a band-aid, but it can't solve the fundamental problem."

03 Conclusion

As AI programming becomes easier, the middle ground is disappearing.

For ordinary users, the right to software development is being decentralized, and everyone can be the "product manager" of their own lives. For the professional field, the threshold is being raised infinitely. Only those "super individuals" who can harness AI and combine the human body with algorithms can survive in the fierce gold rush.

This is an era of gold and also a cruel era.

This article is from the WeChat official account "Hard AI," and the author focuses on technology R & D. It is published by 36Kr with authorization.