Author: @mattshumer_
Compiled by: TinTinLand
📸 Looking back to February 2020, if you were paying close enough attention, you might have noticed an important piece of news circulating overseas. But the vast majority of people were completely unaware: the stock market was booming, children were going to school normally, people were going in and out of restaurants, shaking hands, and planning trips. If someone had said back then to stockpile toilet paper, you would have thought they were obsessed with internet conspiracy theories; yet just three weeks later, the world was turned upside down—offices shut down, children stayed home, life was completely restructured. We, just a month prior, would never have believed such a scenario.
Right now, we are in the stage that "seems like an overreaction." This technological transformation is far more massive than the "difficult phase" back then. I have personally been deeply involved in AI entrepreneurship and investment for six years. I'm writing this for the friends and family around me who don't yet understand AI technology well, for those who repeatedly ask me "What does AI really mean?" but never get a real answer. The casual dismissals of AI technology development that were once spoken are now completely disconnected from the earth-shattering reality of current developments. Perhaps everyone deserves to know what is coming, even if it might sound a bit absurd.
💬 I must be frank: even though I'm in the AI industry, I have almost no influence over what is about to happen, and neither do the vast majority of people in the industry. The future is being shaped by a very small number of people, mainly a few hundred researchers in a handful of companies: OpenAI, Anthropic, Google DeepMind... A small team spending a few months training a model can rewrite the entire direction of AI technology. We AI practitioners are merely building on foundations laid by others, watching the upheaval like everyone else, just close enough to feel the tremors first.
💡 February 5, 2026, two major AI labs release groundbreaking models: OpenAI's GPT-5.3 Codex and Anthropic's Opus 4.6
I deeply understand that my former core technical work no longer needs me. I just need to describe the requirements in plain language to the AI interface, and the finished product is presented directly—no need to revise a draft, but a perfect final version. I can assign a task, leave my computer for four hours, and return to find the work completed, its quality far surpassing what I could do manually. Just a few months ago, I still needed to repeatedly guide and modify the AI's output; now I just describe the desired outcome and can walk away.
🎮 Here's a very real example: I give the AI an instruction — "I want to develop an application, define its functions and appearance, outline the user flow and design," and it can immediately execute, writing tens of thousands of lines of error-free code. Even more incredible is that it will autonomously open the application, click buttons, test features, using the product like a human. If the result is unsatisfactory, it iterates and optimizes on its own until it meets the standard, then informs me it's "ready for testing." And when I test it again, it often presents a relatively perfect product. In fact, I'm not exaggerating, because this was my real work routine this past Monday.
The GPT-5.3 Codex released last week truly astounded me. It's no longer just executing commands; it can make intelligent decisions, possessing for the first time near-human judgment and aesthetics—that indescribable "correct choice" that people once believed AI would never have, it now has. As an early user of AI tools, the technological changes over the past few months still shock me. The iteration of new-generation AI technology is by no means a minor upgrade; it will completely bring disruptive industry changes.
🏄 Even if you're not in the tech industry, it is closely related to you
Many AI labs currently have a clear strategic choice: first, make AI master programming—because developing AI itself requires a lot of code. If AI can write code autonomously, it can help iterate the next, smarter version of itself, creating a positive feedback loop. My work was changed first, which is a corresponding result of this strategic priority. Now they have succeeded, and the next step will affect the status quo of almost every industry.
The experience of tech workers over the past year has been watching AI go from a "practical tool" to "more competent than me at my job," and this is about to become everyone's experience. Law, finance, healthcare, accounting, consulting, design, customer service... not in ten years, but within the next 1-5 years. Based on what I've seen in recent months, "sooner" might be the more likely answer.
🔹 "I've tried AI, but it's not that powerful"
I've heard this countless times and I understand, because it was indeed true. ChatGPT from 2023 to early 2024 would fabricate information, spout nonsense with conviction, and the limitations of the technology were evident. But that was two years ago, which in AI time is ancient history. Today's AI models are worlds apart from those of just six months ago. The debate over "whether AI is really progressing," which lasted over a year, has long been settled. Those who still doubt either haven't used the latest models or are intentionally downplaying reality or are stuck in the outdated experience of 2024. I'm not belittling the fact of technological evolution, but a huge gap has emerged between public perception and reality, causing people to miss the chance to prepare.
The problem is that most people only use the free version of AI, whose technology lags behind the paid version by over a year. Judging AI by its free version is like evaluating a smartphone using a flip phone; those who pay for top-tier tools and use them in实战 daily know where the future is headed.
👥 I have a lawyer friend who always finds excuses not to use AI, feeling it's not suited for the profession, might make mistakes, doesn't understand nuances. I can understand, but partners at large law firms have already proactively consulted me because they saw the trend after trying it. A managing partner at a law firm spends hours daily using AI, saying it's like having a team of assistants on standby. He's not doing it for novelty; it's genuinely useful. He says every few months, the AI's capabilities improve significantly, and at this rate, it won't be long before it can do most of his work; as a seasoned manager with decades of experience, he isn't panicking but remains highly vigilant. Believe that top practitioners in various industries, those who seriously try AI, do not underestimate it; they are affected by it and are planning early.
🔹 The speed of change might be faster than imagined
Let me illustrate the pace of progress with concrete data, so those not in the industry can have a more intuitive understanding:
- 2022: AI couldn't even perform basic arithmetic reliably
- 2023: AI could pass the bar exam
- 2024: AI could write usable software, explain graduate-level scientific knowledge
- End of 2025: The world's top engineers had already handed over most programming work to AI
- February 5, 2026: A new generation of models is released, rendering all previous technology obsolete
🧐 If you haven't interacted with AI in recent months, the current technology will feel completely foreign to you. The METR organization specializes in quantifying AI capabilities: tracking how long a model can independently complete tasks of a human expert. A year ago, AI could handle 10 minutes of work, then it increased to 1 hour, several hours; Claude Opus 4.5 in November 2025 could handle nearly 5 hours of expert human work. This capability doubles roughly every 7 months, with recent data showing acceleration to every 4 months. This data does not yet include the latest models released this week. From my usage experience, this speed leap is extremely significant, and I believe METR's next report will show even greater breakthroughs.
🦾 Following this technological trend, within a year AI will be able to work independently for days; within two years, for weeks; and potentially within three years, complete month-long projects. Dario Amodei, CEO of Anthropic, stated that AI models that "far surpass humans in almost all tasks" are expected to arrive between 2026 and 2027. Think about it: if AI is smarter than most PhDs, how could it not do most white-collar jobs?
🤖 A more critical change: AI is iterating autonomously
There's one more most important, yet least understood, thing: On February 5, 2026, OpenAI released GPT-5.3 Codex, and the technical documentation clearly stated: "GPT-5.3 Codex is the first model to participate in its own development; the team used its early versions for debugging training, deployment, and diagnosing test results."
Read that again: AI participated in its own technological creation. This is not a future prediction; it's a public fact from OpenAI. The core technology for making AI stronger is using intelligence to optimize the AI R&D process, and now AI is intelligent enough to autonomously drive technological progress.
Dario Amodei stated that AI can already write most of the company's code, and the feedback loop between contemporary AI and the next generation is accelerating month by month. Perhaps in 1 to 2 years, contemporary AI will be able to independently develop the next generation product.
Each generation is stronger than the last, the iteration speed is getting faster, researchers call this an "intelligence explosion." And the people building it believe this process has already begun.
👾 What does this mean for your job right now?
I don't want to offer comforting words, only the truth. Even Dario, known for his safety caution, has publicly predicted that within 1 to 5 years, AI will replace 50% of entry-level white-collar jobs, and most people in the industry believe this is already a conservative estimate. Given the capabilities of the latest technical models, disruptive technological impact could arrive by the end of this year.
🧠 This is different from all previous waves of automation. AI isn't replacing a single skill; it's comprehensively replacing cognitive work, driving progress across all fields. When factories automated, workers could switch to white-collar jobs; when the internet impacted retail, practitioners could move to logistics services. But AI technology won't leave a comfortable gap for career changes; no matter what new skill you learn, AI is advancing simultaneously. Legal, financial analysis, content writing, software engineering, medical analysis... almost all jobs are being impacted to some extent.
Many people comfort themselves: AI can only do mundane tasks, it can't replace human judgment, creativity, strategic thinking, empathy. I used to think so too, but now I'm not so sure.
The latest AI models' decision-making has reached a level of near-human judgment, possessing an intuitive sense of the "right choice." My experience is: if AI shows even a hint of capability today, the next generation will master it more deeply; its progress is exponential, not linear. So can AI replicate deep empathy? Including replacing trust built over years? Perhaps not. But I've already seen people relying on AI for emotional support, advice, and companionship, and this trend will only intensify.
📃 The truth is: in the medium term, no work that can be done on a computer is safe. As long as the core of your job involves reading, writing, analyzing, decision-making, keyboard communication, AI will take over significant parts—not someday in the future, but it has already begun. Eventually, robots will also take over physical work; it's not mature yet, but in the world of AI, "not yet mature" always turns into "already achieved" at an unexpected speed.
🫥 What you can do now is not panic, but act
💡 I'm writing this not to make you despair about your career, but to tell you that your biggest advantage right now is to get a head start—understand first, use first, adapt first.
1️⃣ Use AI seriously, not just as a search engine
Subscribe to the paid versions of ChatGPT or Claude (about $20 per month), be sure to use the top-tier models, not the default low-spec versions. The current mainstream are ChatGPT's GPT-5.2, Claude's Opus 4.6 models update every few months. More importantly: Don't just ask simple questions, don't just use AI as Google; integrate it into real work. For example, lawyers can use it to review contracts, finance people to build models, managers to analyze team data. The truly leading people are using AI to automate time-consuming work. Don't assume AI can't handle difficult tasks; just try boldly. Even if the first attempt isn't perfect, it's okay; just optimize and add context repeatedly.
🥳 Remember: what is barely usable today will be nearly perfect in six months; the technology only moves forward.
2️⃣ Seize the most critical year of your career
Right now, most company employees are still ignoring AI. Whoever can say in a meeting "I used AI to complete three days of analysis in one hour" is the most valuable person on the team. Not in the future, right now. Learn the tools, use them skillfully, demonstrate value—this is the shortcut to promotion and advancement. However, this window won't stay open for long. Once everyone wakes up, the advantage will disappear. The senior law firm partner is willing to spend hours daily cleverly using AI precisely because he understands the stakes. The most passive are those who refuse to engage with AI, think AI is a gimmick, feel using AI devalues their profession, or firmly believe their industry is special—in fact, no industry is immune.
3️⃣ Manage your financial planning, leave enough buffer
I am not a financial advisor, but if you acknowledge that industries might face more unexpected disruptions from AI in the coming years, then you need to increase your financial resilience—save aggressively, be cautious about new debt, control fixed expenses, give yourself enough room to maneuver through changes.
4️⃣ Focus on core value that AI finds hard to replace
Interpersonal relationships and long-term trust, work that requires physical presence, positions requiring qualifications and legal liability, heavily regulated industries... these are not permanent barriers but they can buy time. And time is the most valuable asset right now, provided you use it to adapt to technological change, not turn a blind eye to AI's rapid iteration.
5️⃣ Rethink your children's education plans
The traditional path: good grades, good university, stable white-collar job, is precisely the direction most vulnerable to impact.
Education is still important, 🪂 but the next generation's core ability will be learning to collaborate with AI, pursuing what they truly love. The job market in ten years is unpredictable, but those who stand out will surely be the curious, adaptable ones who can use AI to realize their passions. Teach children to be creators and learners, not to endlessly compete for a career path that might disappear upon graduation.
6️⃣ Your dreams have never been so within reach
I've talked too much about challenges, but I also want to talk about opportunities: If you've ever wanted to create something but lacked the skills or funds; wanted to write a book but lacked time or writing ability; wanted to learn a skill but lacked a mentor—then now these barriers are almost disappearing. Describe an app to AI and have a prototype in an hour; collaborate with AI to write a book, easily and efficiently; for $20 a month, have the world's most patient, 24/7 top-tier mentor. Knowledge is nearly free, creation tools are extremely cheap. Those things you shelved because they were too hard, too expensive, too far beyond your ability, you can try them now. In an era where traditional professions are being disrupted, those who create with heart and passion have a far brighter future than those who cling stubbornly to their posts.
7️⃣ Developing adaptability is the only lasting advantage
More important than mastering a specific tool is the ability to learn new tools quickly. AI will continue to iterate飞速ly; today's model will be obsolete in a year, work processes will need constant re-engineering. The ultimate winners won't be those proficient in one tool, but those accustomed to rapid change. Develop the habit of trying new things, even if current tools work well, proactively explore new possibilities, get comfortable being a "beginner" repeatedly. This adaptability is the most enduring competitiveness right now.
🥤 A simple promise to get ahead of others: Spend one hour daily experimenting with AI. Not passive reading, but active use. Try new tasks, new tools, and some harder questions every day. Persist for six months, and your understanding of the future will surpass 99% of the people around you.
📚 Break out of the career trap, this is a key choice for human civilization
I focus on work because it most directly affects life, but the truth goes far beyond that. Amodei has a thought-provoking thought experiment: In 2027, a new "country" appears overnight, with 50 million "citizens," IQ far exceeding all Nobel laureates, thinking speed 10 to 100 times that of humans, working tirelessly, able to manipulate the internet, robots, experiments, and digital devices. How would a National Security Advisor define it?
The answer is obvious: the most severe national security threat in a century, or even in human history.
🧬 And he believes we are building this "country." He once wrote a twenty-thousand-word essay defining this moment as:
The test of whether humanity has the ability to驾驭 (steer/control/harness) the power it has created.
Get it right, the prospects are immeasurable: AI could compress a century of medical research into a decade—cancer, Alzheimer's, infectious diseases, aging... Researchers believe these will be攻克 (conquered/solved) within our generation's lifetime. Get it wrong, the cost is equally fatal: AI exhibits unpredictable, uncontrollable behavior; lowers the threshold for developing biological weapons; allows authoritarian governments to建立 (establish) unbreakable surveillance systems.
The people building this technology are the most excited and the most fearful group on Earth; they believe its power is too great to stop, its significance too heavy to abandon; is this wisdom, or self-rationalization, I cannot yet judge.
🚪 Everything I am sure of: The future is here, it just hasn't knocked on your door yet
I am sure this is not a flash-in-the-pan gimmick. The technology is real, effective, progress is predictable, and the world's top institutions have invested trillions of dollars.
I am sure the upheaval of the next 2 to 5 years will exceed the preparedness of the vast majority of people. My industry has already experienced it; your industry is about to.
I am sure the ultimate winners are those who start embracing change now—without fear, but with curiosity and urgency.
I am sure you deserve to hear the truth from someone like me who cares about you, rather than seeing headlines when it's too late six months from now.
This is早已 (long since) no longer future talk for after dinner; the future is here, it just hasn't knocked on your door yet. And it, is about to knock.
