Anthropic CEO Dario Amodei: Jobs that are purely about writing code will be the first to be replaced by AI; humans should hold on to the 5% of judgment

If anyone in today's world can see clearly where humanity is headed over the next decade, Dario Amodei is certainly one of them.

He once gave a very in-depth interview. In that conversation, he candidly shared his predictions for the future—

Over the next decade, which industries will be completely upended?

What major should a 25-year-old choose to study today?

Is programming really a dead end? How can ordinary people keep their jobs in this great reshuffling?

"Pure coding" jobs will be the first to die

During the interview, the host asked a question that fills everyone with immense anxiety:

If I'm 25 today and want to achieve financial success in the next decade, what should I study? Is programming still useful? Which industries still have upside?

Dario's answer was brutal.

He said bluntly: "I think coding will be the first to disappear—or rather, the first to be completely replaced by AI models."

For the past two decades, "switching to coding" has been almost the fastest shortcut for ordinary people to move up the social ladder.

Countless young people flooded into CS programs, just to land a high-paying "code monkey" job at a big tech company.

But Dario points out mercilessly that the era of pure code-typing is coming to an end! Inside Anthropic, Claude Code is already evolving exponentially.

But note: Dario did not sentence the software industry to death.

He drew an extremely crucial distinction: Coding ≠ Software Engineering.

Writing specific lines of code will be done by AI, but "software engineering"—that is, acting as an architect, thinking about how to design a system, how to meet users' real needs, and how to manage a team made up of multiple AI agents—will persist for much longer.

"This is an extremely powerful 'comparative advantage.'" Dario explained.

Dario offered a highly disruptive mathematical formula: 5% + 95% = 20x productivity.

Humanity's comparative advantage has not disappeared; it has simply shifted dramatically. In the future workflow, AI will handle 95% of the execution work. Humans only need to hold fast to the most core 5%.

What is that 5%?

It is judgment. It is taste. It is "knowing where we are going."

This means your productivity will be amplified 20-fold! The future doesn't need a room full of junior programmers—it only needs one "super individual" who understands business and human nature, leading a group of tireless AIs, to take down a traditional hundred-person team.

Dario used a metaphor in the interview: a tsunami.

It's like a tsunami bearing down on us—already clearly visible on the horizon, close at hand—yet people are frantically looking for excuses, comforting themselves: 'That's not a tsunami, it's just a refraction of light.'

Why is Dario so certain about AI's evolution? Because he was one of the first to discover and firmly believe in the "Scaling Law."

For the Scaling Law, Dario used a highly vivid metaphor: it's like a chemical reaction.

If you want to make fire, you need different ingredients. If you put "data" and "compute" in the right proportions into the "model scale" reaction furnace, the final product of this chemical reaction is "intelligence."

Five years ago, computers couldn't even draw a decent picture of a cat; today, you can send Claude a video of a monkey doing tricks and ask it "how many times did the ball change hands among the monkeys," and it will answer you precisely.

This is the "sense of urgency" Dario feels. People are always accustomed to predicting the future with linear thinking, believing "this change is too big, it can't happen," but technology is racing forward at exponential speed.

When the tsunami comes, if you're still standing on the beach picking up seashells, the ending is doomed to destruction.

Where is the safe harbor? In the next decade, ordinary people have only these two fallbacks

Dario offered a "future survival guide." He believes the moats of the future are deeply rooted in the following three areas:

Fallback one: Extremely "human-centered" work

The high-paying jobs of the future will belong to those roles that can connect people, understand people, and mobilize people. Emotional intelligence, empathy, and the ability to build complex social relationship networks will become the most expensive scarce goods.

Fallback two: Going deep into the "physical world"

Dario strongly advises young people to pay attention to fields where the "physical world" intersects with "traditional engineering," such as semiconductor manufacturing, new energy infrastructure, and high-end manufacturing.

Core foundational capabilities: street smarts and critical thinking

Dario believes it is critical thinking and street smarts.

"You don't want to be blinded by illusions, you don't want to hold false beliefs, and you certainly don't want to be scammed."

During this interview, which lasted several hours, the host threw Dario a very tricky question—

"If I gave you 100 dollars right now, and you could only buy one stock or bet on one industry (other than your own AI track), what would you buy?"

Dario laughed. He didn't recommend Microsoft, nor Nvidia.

He gave an answer that was surprising yet entirely reasonable: biotechnology.

Don't forget, before founding a $40 billion AI empire, Dario was originally a top PhD in physics and biophysics, and did a postdoc at Stanford Medical School.

Dario asserted: "Biotechnology is about to experience an AI-driven renaissance. We are on the verge of curing a large number of desperate diseases."

He specifically pointed to one sub-sector: peptide therapeutics and cell-based therapies (such as CAR-T).

Why? Because traditional small-molecule chemical drugs are designed as if by luck; whereas new-generation biotechnologies such as peptides and mRNA have "programmable" characteristics. They are like computer code in the biological world, capable of digital computation and continuous optimization through AI.

If you are looking for a long-term investment trend, or if your child is hesitating about whether to study biomedicine, Dario's judgment is worth considering.

Stop building "wrapper apps" and learn to "play the piano"

Dario admitted frankly: if we use AI carelessly, humans will indeed become dumber.

But the choice is in human hands. Just as after the invention of the calculator, humans did not lose their mathematical ability; instead, they used calculators to compute the parameters of the moon landing trajectory. Dario himself still often does mental arithmetic in his head, because "this allows mathematics to deeply integrate with my thought process."

For ordinary people who want to start a business or become a super-individual in the AI era, Dario offered the following advice—

1. Absolutely do not build "wrapper products."

If you simply call the Claude or ChatGPT API, make a pretty UI, add some prompts, and sell it for money. Dario warned: "This kind of thing has no moat whatsoever. Anyone can devour your revenue. You don't even need to wait for the AI giants to act; you will die off first on your own."

2. Build extremely deep vertical barriers.

Anthropic will never touch highly specialized vertical fields, such as biomedicine crossed with AI, or the highly regulated financial services industry. The dirty, hard work and vertical know-how that the giants look down on or cannot be bothered to do are the soil for ordinary entrepreneurs to survive.

3. Treat "prompt engineering" as piano practice.

Many people complain that AI is not easy to use because they simply don't know how to use it. Dario quoted a famous saying: "Writing prompts is like playing the piano. You can't expect to sit down on the piano bench and immediately play Beethoven."

To enable non-technical people to use AI well, Anthropic developed Claude Co-work, which requires no coding. But no matter how simple the tool is, you need to spend time "hands-on practicing."

First principles are your only weapon for foreseeing the future

At the end of this in-depth interview, Dario said the following:

Over the past 10 years, I have repeatedly experienced this: when you extrapolate the future based on simple curves, or reason about it with simple logic, you will reach some counterintuitive conclusions. Many people will say: 'That's crazy, it can't happen.'

But in the end, it really did happen. You can predict the future almost 'for free,' with just a few empirical observations plus first-principles thinking."

What humans can do is understand the human heart, connect with the physical world, control AI, and be among the 5% who can steer the ship.

This article comes from the WeChat public account "Xin Zhi Yuan" (ID: AI_era), author: ASI Revelation, published by 36Kr with authorization.