AI makes the strong stronger, the weak weaker

After the New Year, I caught up with a few friends about recent developments, and inevitably we discussed the impact and changes AI is bringing to work. Here are some notes:
1/ A Case Study
Friend W is a senior software engineer. Recently, he integrated a one-year AI workflow into two of his friends' companies. The result was that these two tech teams reduced their staff by two-thirds, and the remaining one-third saw overall productivity increase rather than decline. The facts are clear: AI has indeed caused some people to lose their jobs.
In the programming field, entry-level programmers are finding it harder to get jobs. Meanwhile, programmers with some experience who can effectively leverage various AI tools have seen their productivity greatly unleashed. They can do the work of ten, no problem.
2/ Should We Do Something?
W has also developed many small tools, all of which he directly sketches as prototypes and hands over to Lovable. In the past, it was necessary to hire designers, spend money, and communicate; now, all of that is eliminated, reducing friction and enabling rapid development of small tool products without obstacles.
As this situation becomes more common, more people will have a thought: I need to do something, or it might be a pity.
This is one of the reasons why vibe coding is endlessly praised.
Of course, many of these "products" will be practically worthless "junk" and self-congratulatory.
The era of "everyone is a product manager" is over; now it’s "everyone is a programmer," at least those who can use AI tools to write code.
3/ Two Challenges
For people without technical experience, the increasingly complex engineering and technical issues in vibe coding will become a challenge. Especially as user numbers grow, problems like "code chaos," engineering management, and security will surface during product iteration and upgrades.
Additionally, like traditional internet products, creating the product is only Day 1. The real focus is on subsequent marketing, user (Agents) acquisition, and commercialization.
A true "one-person company" is difficult, but as productivity is massively unleashed at each stage, more small companies will become unicorns.
Such companies existed even before the AI era, facing similar challenges. For example, Craigslist, the pioneer of classified ads, operated with only one person for a long time; Instagram was acquired for $1 billion with only 12 people.
4/ What Is Scarce
The core value of vibe coding is that when there is a clear need and idea, it can greatly unleash human productivity. At least, quickly creating a demo is no longer a bottleneck.
But it cannot solve the question of what kind of products, especially those with commercial value, are worth building.
Understanding needs and continuously refining and operating the product after launch remain the core.
And throughout this process, the "decision quality" made by humans is still critical.
Cursor’s design lead said that even the best engineers can only manage four Agents simultaneously; human-AI collaboration remains a bottleneck.
TL;DR - The strong get stronger, the weak get weaker.
Finally, here’s a chart: each dot represents 32 million people × 2,500 dots = Earth’s 8.1 billion people. Gray indicates 6.8 billion people who have not used AI; green shows 1.3 billion who have used large models for free chat; yellow represents 15-35 million paid large model users; red is a tiny group of deep users.
The true AI wave has not yet begun.