When anyone can Vibe Code to build an app, AI will replace the people who don’t know what “good” is.

The app “Food Saver Hunter,” which claims to be developed with the assistance of AI and integrates information on near-expiry products from convenience stores in Taiwan, has recently drawn attention on social media. The product focuses on real-time queries for nearby near-expiry items and arrival notifications, with developers stating they completed the development in just two weeks using AI tools. However, shortly after its launch, it faced criticism, not only for API authorization issues but also for serious cybersecurity vulnerabilities, potentially exposing users’ precise GPS coordinates at home.

This highlights the risks of “vibe coding”: when developers lack engineering training, they often cannot discern what constitutes “good code” and find it even harder to identify hidden security issues. Thus, when AI produces seemingly usable results, there is a tendency to accept them uncritically and push them straight to market. Despite the developers making fixes after external warnings, subsequent tests showed only partial repairs, and core risks remained unaddressed. In other words, the problem is not just “writing incorrect code,” but rather that the overall system design lacked a basic understanding of cybersecurity from the outset.

Rather than saying AI will replace engineers, it is more accurate to describe it as an accelerator. It can handle low-level and repetitive tasks, but the truly crucial aspect remains those engineers who know what it means to “write well,” what requires rework, and which risks are unacceptable. The real impact of AI is to enhance efficiency, not replace judgment. Just as an untrained AI cannot produce good writing, a person with poor taste in music will only create songs reflective of that taste.

When AI can increase efficiency fivefold, will you use it to reduce manpower by five times, or to create five times the output?

Vibe Coding Application “Food Saver Hunter” Exposes Cybersecurity Vulnerabilities

According to a report by Crypto City, the app “Food Saver Hunter,” which claims to be developed with the assistance of AI and integrates information on near-expiry products from convenience stores in Taiwan, has sparked interest on social media. The product focuses on real-time queries for nearby near-expiry items and arrival notifications, with developers stressing that it took about two weeks to complete development using AI tools. However, shortly after its launch, it faced skepticism from engineers and netizens, not only involving API authorization disputes but also exposing serious cybersecurity vulnerabilities that could even lead to users’ precise GPS coordinates being leaked.

According to Zeabur engineers, simply using the app with the location feature enabled means that related coordinate data will be written to the database and exposed directly on the public network without proper protection. Although the developers made repairs after external warnings, subsequent tests showed only partial fixes, and core risks remained unaddressed. In other words, the problem is not just “writing incorrect code,” but rather that the overall system design lacked a basic understanding of cybersecurity from the outset.

Can AI replace human engineers? You must first understand what constitutes good code.

This incident has once again drawn attention to a recently popular development model: the so-called “vibe coding,” where individuals without a technical background rely on AI tools to quickly generate product prototypes or even launch them directly. The advantage of this model lies in its speed, but the risks are equally apparent: when developers lack engineering training, they often cannot determine what constitutes “good code,” making it even harder to identify hidden security issues. Thus, when AI produces seemingly usable results, there is a tendency to accept them uncritically and push them straight to market.

From this perspective, the issues with “Food Saver Hunter” are not simply due to AI errors, but rather the inability of people to judge whether AI is correct. When the system has vulnerabilities, only superficial issues are patched, rather than revisiting the structural layer for a reassessment, resulting in ongoing risks. This echoes the key observation from previous reports: individuals who have been trained and possess judgment skills, when paired with AI, can achieve tremendous results; however, without that capability, AI may instead amplify errors.

(AI can score 80 points, those who cannot reach 100 are destined to be eliminated! McKinsey and Harvard alumni suggest newcomers do it this way.)

Will you use AI to reduce manpower by five times, or to enhance productivity by five times?

In reality, AI’s role in programming is more akin to that of an “accelerator” rather than a “replacement.” It can quickly complete repetitive, structured, low-level tasks, significantly boosting development efficiency, but it cannot replace the judgment of quality and the overall understanding of systems. The truly valuable engineers are still those who know what it means to “write well,” what situations require refactoring, and which risks cannot be ignored.

This is similar to the management roles within companies. The value of a manager is never merely in execution, but in determining which direction is correct, which outcomes are unacceptable, and which issues need reworking. To some extent, companies borrow the “vision” of these individuals for decision-making. AI can assist in output, but it cannot replace this kind of judgment ability.

Therefore, even as claims of “AI will replace engineers” continue to emerge in the market, given the current development trajectory, the more realistic impact remains in saving manpower and enhancing efficiency, rather than complete replacement. In a core team, senior engineers capable of judging quality and risk will become even more critical.

The question thus returns to everyone: when AI can increase efficiency fivefold, will you choose to use it to reduce manpower by five times, or to create five times the output?

This article “When Anyone Can Vibe Code to Write Apps, AI Will Replace Those Who Don’t Know What Is Good” first appeared in Chain News ABMedia.