As the company owner, I'm experiencing the transformative power of AI firsthand. Besides buying Claude Max for everyone to use, we now regularly dedicate time in our operations meetings to allow everyone to share their recent experiences and challenges in using AI to assist in their work. The PM and I use it the most and provide the most feedback. Fortunately, our team members are very willing to learn, and I believe we'll soon bridge the cognitive gaps among team members. Anyone seriously using coding agents will definitely notice. A year ago, the agent was practically a completely different species from today's. In this year, coding agents have gone from "usable" to "essential," and adoption is accelerating. Before, I'd hesitate to write one myself; now, my first instinct is to dump it on the agent. Why is the programming world progressing so rapidly? Because code itself is a highly structured and logically rigorous language. There's also a massive amount of Stack Overflow Q&A and GitHub open-source projects online—all public, standardized, and verifiable training materials. AI has incredibly favorable conditions to grow from these resources. But other fields are not like this. The details of a lawyer's cases, the accumulated clinical judgments of a doctor, the valuation models built by an investment banker—this domain knowledge has long been locked within companies, never systematically digitized, let alone made public. Without sufficient nourishment, AI naturally cannot grow its strength. Silicon Valley has understood this, which is why a clear trend has emerged in recent years: AI companies are aggressively acquiring "domain knowledge companies." Harvey AI, a legal AI company, reached a valuation of $8 billion within three years of its founding, serving more than half of the top 100 law firms in the US, with annual recurring revenue exceeding $150 million. At the same time, Filevine acquired Pincites and Parrot, and opened a new office next to its Anthropic headquarters in San Francisco, directly poaching AI talent. The same logic applies to the medical field. Stryker acquired Care.ai to create smart wards, and Commure bought Augmedix's AI medical transcription technology for $139 million. Abridge, focusing on converting doctor-patient conversations into medical records, raised $550 million in a single funding round in 2025. These cases all illustrate one fact: capital is using mergers and acquisitions to directly buy up the experience curve accumulated by humans over a decade. Whoever masters domain knowledge can train truly useful vertical AI. What does this mean for white-collar workers? It means the time lag is rapidly shrinking. Previously, AI companies lacked domain data, so AI progress in vertical industries was slow. Now this gap is being filled. Fields previously considered to require highly specialized judgment, such as law, healthcare, finance, and accounting, are seeing their competitive advantages eroded layer by layer. AI replacing entry-level positions is already underway. Six months ago, I thought this would be three to five years from now; now, given the pace, structural changes within two years are not surprising. New job openings will decrease, and more and more companies will fail to fill vacancies or even lay off employees. Those who are not adept at using AI will be at the forefront of the tsunami. Many white-collar workers will ultimately only have one thing left to grasp: humans have licenses, AI does not. Just like legal documents and notaries, the workflow is highly mechanical, but the law requires a human signature to be valid. In the future, lawyers, accountants, and doctors may also follow this model, with AI handling most of the substantive work, and humans responsible for final review and liability. Autonomous driving follows the same path. The technology is already mature, but the real obstacle has always been liability attribution. However, with lower accident rates than humans and the efficiency and cost advantages being quantified, many US states have begun to allow commercial operation of driverless cars, shifting responsibility to operators and manufacturers. AI replacing white-collar workers will inevitably lead to the same outcome sooner or later. In a recent interview, Elon Musk mentioned that within three years, Optimus's surgical capabilities will surpass those of top doctors. Training a cardiac surgeon takes more than a decade; once one machine learns, hundreds of thousands can be updated simultaneously, without fatigue, tremors, or impaired judgment after consecutive surgeries. Humans spend ten years training an expert; AI only needs one training session to replicate an entire population. --- What can you, as an ordinary person, do? First, learn to use AI. You need to be more proficient than others, not just copy and paste to chat with chatbots, but learn the latest cutting-edge knowledge. For example, try Claude Code's latest Cowork, and proactively explore tools beyond ChatGPT. Second, don't just be an executor; be someone who "helps the team implement AI." Knowing how to use AI is basic, but being able to teach others how to use it and help the company implement it is what creates a competitive advantage. Pure executors are most easily replaced, while those who drive change will always have a place. Third, find a Plan B outside of your job. While you still have the resources, start developing income streams outside of your main job. Side hustles, investments, a second skill—anything is fine; don't put all your eggs in one basket.

I think the advice given is quite pertinent. Those who don't use AI will be left behind. Being a super-individual isn't something everyone can do. But having a side hustle to earn some extra money is something everyone can try.

The Twitter plugin my boss made with AI is absolutely amazing!

Serving everyone