THIS COULD BE THE LAST CHANCE FOR AN ORDINARY PERSON TO GET TO KNOW AI IN ADVANCE

Original title: Something Big Is happening
original link: @mattshumer_
Photo by Peggy, Block Beats

EDITOR: MANY PEOPLE'S JUDGMENT OF AI IS STILL AT A STAGE WHERE IT SEEMS TO BE USEFUL, BUT THAT'S IT. MOST PEOPLE, HOWEVER, DO NOT REALIZE THAT A CHANGE SUFFICIENT TO REORDER DAILY LIFE HAS BEGUN QUIETLY。

THIS ARTICLE IS NOT AN ABSTRACT DISCUSSION OF WHETHER "AI WILL REPLACE HUMAN BEINGS" BUT IS THE FIRST PERSON TO RECORD A CHANGE IN REALITY AS AN PRACTITIONER AT THE CORE OF AI'S RESEARCH AND DEVELOPMENT AND APPLICATION FRONT: WHEN MODEL CAPABILITIES DO NOT LEAPFROG OVER IN A SHORT PERIOD OF TIME, WHEN AI IS NO LONGER MERELY A SUPPORT TOOL, BUT IS ABLE TO DO COMPLEX WORK ON ITS OWN AND EVEN PARTICIPATE IN THE CONSTRUCTION OF THE NEXT GENERATION OF AI, THEN THE FIRM PROFESSIONAL BOUNDARIES ARE RAPIDLY LOOSENING。

THIS TIME, CHANGES ARE NOT INCREMENTAL TECHNOLOGICAL UPGRADES, BUT MORE LIKE A BUSINESS LOGIC TRANSITION. WHETHER IN THE TECHNOLOGY INDUSTRY OR NOT, EVERYONE WHO WORKS AT THE HEART OF THE SCREEN CAN'T STAY AWAY. WHEN AI STARTED TO FINISH YOUR WORK, HOW WOULD YOU LIKE TO LIVE WITH IT

The following is the original text:

Think back to February 2020。

If you were very attentive, you might have noticed that a few were talking about a virus that was spreading abroad. But most people don't care. The stock market was doing well, the kids went to school, you went down, shook hands, planned a trip. If someone tells you he's hoarding toilet paper, you probably think he's watching too much in some weird corner of the network. But in about three weeks, the world changed completely. Offices are closed, children go home, life is rearranged into a form that you would never believe if someone had described it a month ago。

I think we're in some kind of "this is not a little exaggerated" phase, and this is going to be much bigger than the new coronary outbreak。

IT'S BEEN SIX YEARS SINCE I STARTED AND INVESTED IN THE AI FIELD, AND I'VE LIVED IN THIS WORLD. I WROTE THIS DOWN FOR THOSE PEOPLE IN MY LIFE WHO ARE NOT IN THIS PROFESSION -- MY FAMILY, FRIENDS, PEOPLE I CARE ABOUT. THEY KEPT ASKING ME, "AI, WHAT'S GOING ON?" AND THE ANSWER I GAVE WAS THAT THERE WAS NO REAL REFLECTION OF WHAT WAS HAPPENING. I ALWAYS GIVE A POLITE VERSION, A COCKTAIL VERSION. BECAUSE IF YOU TELL THE TRUTH, IT SOUNDS LIKE I'M CRAZY. FOR A LONG TIME, I HAVE ALSO TOLD MYSELF THAT THIS IS A GOOD ENOUGH REASON TO LEAVE BEHIND WHAT REALLY HAPPENED. BUT NOW, AS I HAVE BEEN SAYING, THE GAP WITH REALITY IS SO WIDE THAT IT CANNOT BE IGNORED. PEOPLE I CARE ABOUT SHOULD KNOW WHAT HAPPENS NEXT, EVEN IF IT SOUNDS CRAZY。

Let's get one thing straight: although I work in the AI industry, I have little influence on what's about to happen, and so does the vast majority of people in the industry. What really shapes the future is a few: hundreds of researchers in a few companies — OpenAI, Anthropic, Google DeepMind, and a few others. A training mission, carried out by a small team within a few months, could result in an AI system capable of altering the entire technology trajectory. Most of our practitioners are building things on foundations that others have already built. We, like you, just watched it unfold — just because we were closer and felt the ground shaking first。

But now it's time. It's not the time to talk, it's the time to understand。

I know it's true because it happened to me first。

And one thing is, almost all the people outside the tech world don't realize: that's why so many people in the industry are ringing alarms now because this is happening to us. We are not making predictions, we are telling you that these things have actually happened in our work, and you are probably next。

FOR YEARS, AI HAS BEEN MAKING STEADY PROGRESS. SOMETIMES THERE'S A BIG LEAP, BUT THERE'S A LONG SPACE BETWEEN THEM, AND YOU CAN DIGEST IT SLOWLY. BUT BY 2025, NEW TECHNOLOGIES FOR MODELLING HAD EMERGED AND PROGRESS HAD ACCELERATED DRAMATICALLY. THEN FASTER, FASTER. EVERY GENERATION OF NEW MODELS IS NOT JUST A LITTLE BETTER THAN THE PREVIOUS GENERATION, BUT A LITTLE BETTER, AND THE DISTRIBUTION IS SHORTER. I'M USING AI MORE AND MORE, AND I'M TALKING TO IT LESS AND LESS, AND I'M WATCHING IT DEAL WITH THINGS THAT I THOUGHT I HAD TO DO ON MY OWN。

Then, on 5 February, two top AI laboratories released new models on the same day: GPT-5.3 Codex of OpenAI and Opus 4.6 of Anthropic. At that moment, everything was right. It's not like the lights are suddenly turned on, more like you realize that the water level has slipped to the chest。

I NO LONGER NEED TO DO THE REAL TECHNICAL PART OF MY WORK IN PERSON. I DESCRIBED IN PLAIN ENGLISH WHAT I WANTED TO BUILD, AND IT JUST HAPPENED. IT'S NOT A DRAFT THAT I HAVE TO CHANGE OVER AND OVER AGAIN. IT'S A FINISHED PRODUCT. I TOLD THE AI TARGET THAT FOUR HOURS AWAY FROM THE COMPUTER AND BACK, THE JOB WAS DONE — AND IT WAS DONE BETTER THAN I DID MYSELF, WITHOUT ANY MODIFICATIONS. A FEW MONTHS AGO, I HAD TO TALK TO AI BACK AND FORTH, GUIDE AND ADJUST; NOW, I JUST DESCRIBE THE RESULTS AND THEN LEAVE。

I'LL GIVE YOU A CONCRETE EXAMPLE OF WHAT THIS IS LIKE IN PRACTICE. I'D SAY TO AI, "I WANT TO DO AN APPLICATION LIKE THIS, AND IT'S SUPPOSED TO DO THESE FUNCTIONS, ABOUT THE SAME. USER PROCESSES, DESIGNS, ALL OF THEM. AND THEN IT REALLY DID. IT'S WRITTEN DOWN AS 100,000 LINES. AND WHAT'S EVEN MORE INCREDIBLE — THE PART THAT COULDN'T HAVE BEEN IMAGINED A YEAR AGO — IS THAT IT WILL TURN ITSELF ON THIS APPLICATION, CLICK THE BUTTON, TEST THE FUNCTION, AND USE IT LIKE A PERSON DOES. IF IT FEELS THAT SOMETHING IS NOT RIGHT AND THAT IT DOES NOT WORK PROPERLY, IT GOES BACK AND CHANGES ITSELF, AND CHANGES ITSELF, LIKE AN DEVELOPER, CONSTANTLY FIXES AND GRINDS UNTIL IT IS SATISFIED. IT WAS ONLY AFTER IT HAD DETERMINED THAT THE APPLICATION MET ITS CRITERIA THAT IT WOULD COME BACK TO ME: "YOU CAN TEST IT." AND WHEN I GO TO THE TEST, IT'S USUALLY PERFECT。

I'm not exaggerating. That's my real workday on Monday。

But what really shook me was the model that was released last week. It's not just implementing instructions, it's judging. For the first time, it made me feel like it had some kind of taste -- the kind of instinct that people keep saying that AI can never have, about what is the right choice. The model already has it, or at least it is close to making the distinction irrelevant。

I'VE ALWAYS BEEN THE FIRST GROUP TO ADOPT THE AI TOOL. BUT IN THE PAST FEW MONTHS, I HAVE BEEN COMPLETELY SHOCKED. THIS IS NOT A GRADUAL IMPROVEMENT. IT'S A COMPLETELY DIFFERENT THING。

Why is this about you -- even if you're not in the technology industry

THE AI LAB MADE A VERY CLEAR CHOICE: THEY GAVE PRIORITY TO AI TO WRITE CODE. THE REASON IS SIMPLE -- BUILDING AI ITSELF REQUIRES A LOT OF CODE. IF AI CAN WRITE THESE CODES, IT CAN HELP BUILD ITS NEXT GENERATION: A SMARTER VERSION, A BETTER CODE, AND THEN A SMARTER VERSION. LET AI MASTER THE PROGRAMMING, IT'S THE KEY TO EVERYTHING. THAT'S WHY THEY DID IT FIRST. MY JOB CHANGED BEFORE YOU DID, NOT BECAUSE THEY WERE SPECIFICALLY TARGETING SOFTWARE ENGINEERS, BUT BECAUSE OF THE SIDE EFFECTS OF THE DIRECTION THEY PRIORITIZED。

Now that step has been taken. And they are turning to all the other areas。

IN THE PAST YEAR, THE EXPERIENCE THAT TECHNOLOGISTS HAVE EXPERIENCED -- LOOKING AT AI FROM BEING A USEFUL TOOL TO BEING "DOING MY JOB BETTER THAN I" -- IS ABOUT TO BECOME AN EXPERIENCE FOR ALL. LEGAL, FINANCIAL, MEDICAL, ACCOUNTING, CONSULTING, WRITING, DESIGN, ANALYSIS, CUSTOMER SERVICE... NOT TEN YEARS LATER. THE PEOPLE WHO BUILT THESE SYSTEMS SAID IT WAS ONE TO FIVE YEARS. SOME SAY EVEN SHORTER. AND IN THE LIGHT OF THE CHANGES THAT I HAVE SEEN IN RECENT MONTHS, I THINK IT'S EVEN MORE POSSIBLE。

"BUT I'VE USED AI, NOT MUCH. I DON'T KNOW

I've heard that word countless times, and I fully understand it because it was true。

If you used ChatGPT in 2023 or early 2024, you'd be right if you thought "it'd make up" and "that's it." Those earlier versions were indeed limited in capacity, hallucinating and confident in their absurd content。

BUT THAT WAS TWO YEARS AGO. IN AI'S TIMESCALE, IT'S ALMOST PREHISTORIC。

THE MODELS THAT ARE AVAILABLE TODAY ARE COMPLETELY DIFFERENT FROM EVEN THE VERSION SIX MONTHS AGO. THE DEBATE OVER WHETHER "AI IS REALLY MAKING PROGRESS" OR "FAILED INTO THE CEILING" HAS BEEN OVER FOR OVER A YEAR. IT'S OVER. PEOPLE WHO SAY THAT EITHER HAVE NEVER USED THE CURRENT MODEL, DELIBERATELY DILUTE THE REALITY, OR REMAIN IN THE 2024 EXPERIENCE, WHICH IS NO LONGER RELEVANT. I AM NOT DEMEANING ANYONE, BUT RATHER TRYING TO EMPHASIZE THAT THE GAP BETWEEN PUBLIC AWARENESS AND REALITY HAS BECOME SO DANGEROUS THAT IT PREVENTS PEOPLE FROM PREPARING IN ADVANCE。

Another problem is that most people use free versions of the AI tool. The free version is more than a year behind the version available to paid users. ChetGPT to judge the level of AI is like flipping a cell phone to evaluate the development of a smartphone. Those who pay for the most powerful tools and use them daily in real work know very well what happens next。

I OFTEN THINK OF A FRIEND OF MY LAWYERS. I KEPT URGING HIM TO USE AI SERIOUSLY IN THE FIRM, AND HE ALWAYS FOUND A REASON: IT WAS NOT SUITABLE FOR HIS SUBDIVISIONS, IT WAS WRONG DURING THE TESTS, IT DID NOT UNDERSTAND THE NUANCES OF HIS WORK. I UNDERSTAND. BUT I HAVE BEEN CONSULTED BY THE PARTNERS OF LARGE FIRMS BECAUSE THEY TRIED THE LATEST VERSION AND SAW TRENDS. ONE OF THE MAJOR LAW FIRMS' MANAGEMENT PARTNERS SPENDS HOURS A DAY USING AI. HE SAID THAT IT WAS LIKE HAVING AN ENTIRE TEAM OF JUNIOR LAWYERS AT ALL TIMES. HE DIDN'T PUT AI AS A TOY, BUT IT REALLY WORKED. HE SAID TO ME ONE THING THAT I REMEMBER SO FAR: EVERY FEW MONTHS, HIS ABILITY TO WORK WILL BE SIGNIFICANTLY ENHANCED. ON THIS TRACK, HE EXPECTED AI TO BE ABLE TO DO MOST OF HIS WORK VERY SOON -- AND HE WAS A MANAGEMENT PARTNER WITH DECADES OF EXPERIENCE. HE DID NOT PANIC, BUT HE WAS FOLLOWING IT VERY, VERY SERIOUSLY。

THOSE WHO ARE REALLY AT THE FOREFRONT OF THEIR RESPECTIVE INDUSTRIES — THOSE WHO ARE EXPERIMENTING SERIOUSLY — DO NOT UNDERESTIMATE ALL OF THIS. THEY'VE BEEN SHAKEN BY WHAT AI CAN DO NOW AND REPOSITIONED ACCORDINGLY。

How close is it

I would like to concretize this speed, because if you do not have close observation, that is the hardest part to believe。

2022: AI CAN'T EVEN CALCULATE BASIC ALGORITHMS. IT'LL TELL YOU PROPERLY THAT 7X8=54。

2023: It can pass the bar exam。

2024: It can write operational software to explain scientific issues at postgraduate level。

LATE 2025: SOME OF THE WORLD ' S TOP ENGINEERS STATED THAT THEY HAD GIVEN MOST OF THEIR PROGRAMMING TO AI。

February 5th, 2026: The arrival of the new model made everything look like another era。

IF YOU HAVEN'T USED AI IN THE LAST FEW MONTHS, TODAY'S IT'S ALMOST UNRECOGNIZABLE TO YOU。

There's an organization called METR that measures this with data. They track how long a model can complete a realistic mission (as measured by the time it takes human experts to do it) without human intervention. About a year ago, that number was 10 minutes; then an hour; then a few hours. The most recent survey (Claude Opus 4.5 in November 2025) shows that AI has been able to perform nearly five hours of work requiring human expertise. And that figure is about double every seven months, and the latest data even indicate that it may accelerate to double every four months。

AND THAT HASN'T INCLUDED THE MODEL THAT WAS JUST RELEASED THIS WEEK. IN MY OWN EXPERIENCE OF USE, THIS LEAP WAS REMARKABLE. I EXPECT THE NEXT METR UPDATE TO SHOW A CLEAR JUMP AGAIN。

IF YOU EXTRAPOLATE THIS TREND, WHICH HAS BEEN GOING ON FOR MANY YEARS WITHOUT ANY SIGN OF SLOWING DOWN, THEN AI MAY BE ABLE TO WORK ON ITS OWN FOR A FEW DAYS IN ONE YEAR, FOR SEVERAL WEEKS IN TWO YEARS AND FOR SEVERAL MONTHS IN THREE YEARS。

The CEO Dario Amodei of Anthropic said that "it is clearly better than the AI of almost all human beings on almost all missions" with a timetable of 2026 or 2027。

THINK ABOUT THAT JUDGMENT. IF AI IS SMARTER THAN MOST DOCTORS, DO YOU REALLY THINK IT CAN DO MOST OFFICE WORK

AI IS BUILDING THE NEXT GENERATION AI

There is one more thing that I think is the most important but, at least, understood progress。

On February 5, when OpenAI released GPT-5.3 Codex, it wrote in the technical document: "GPT-5.3-Codex is our first model to play a key role in our own creation. The Cordex team uses an early version to test its training process, manage deployment and diagnose test results and assessments. I don't know

READ AGAIN: AI WAS INVOLVED IN BUILDING ITSELF。

This is not a guess about the future, but OpenAI is telling you that they just released AI, which has been used to create itself. One of the core factors that makes AI stronger is the use of intelligence in the development of AI. And now, AI is smart enough to substantially drive its own evolution。

The CEO Dario Amodei of Anthropic also says that now AI has produced "a lot of codes" in his company, and that the current feedback loop between AI and the next generation of AI is "accelerating every month". He thought that we might be "one or two years away from the current generation of AI building the next generation on its own."。

A generation helps to build the next generation, and a smarter generation builds the next generation faster — what researchers call smart explosions. And the people who know all this, who are building it themselves, believe that the process has begun。

What does that mean for your job

I'll be honest because you deserve honesty, not consolation。

Dario Amodei, probably the most security-oriented CEO of the AI industry, has publicly predicted that AI will eliminate 50% of the primary white-collar jobs within one to five years. Many practitioners believe that this judgement is conservative. Given the capabilities of the latest models, the technological conditions for large-scale subversion may have been in place by the end of this year. Real transfer to the economy will take time, but the bottom capacity is now coming。

THIS IS NOT THE SAME AS ANY OF THE PREVIOUS ROUNDS OF AUTOMATION. THE REASON IS THAT AI IS NOT A SUBSTITUTE FOR A PARTICULAR SKILL, BUT A UNIVERSAL SUBSTITUTE FOR COGNITIVE WORK. MOREOVER, IT HAS GROWN STRONGER IN ALL ITS ASPECTS. THE AUTOMATION OF FACTORIES ALLOWS REPLACEMENT WORKERS TO SWITCH TO OFFICE JOBS; AFTER INTERNET SHOCKS, PEOPLE CAN ALSO SWITCH TO LOGISTICS OR SERVICES. BUT AI DID NOT LEAVE A "SAFE SPACE". YOU LEARN SOMETHING, IT'S GETTING BETTER AT THE SAME TIME。

A few specific examples — but remember, they are only examples, not complete lists. Your job doesn't mean it's safe without being named. Almost all knowledge-based work is being affected。

LAW: AI HAS BEEN ABLE TO READ CONTRACTS, TAKE STOCK OF CASES, WRITE LEGAL DOCUMENTS AND CONDUCT LEGAL RESEARCH, WHICH IS CLOSE TO THE LEVEL OF SOLICITOR. THE MANAGEMENT PARTNER USES AI, NOT FOR FUN, BUT BECAUSE IT ALREADY OUTNUMBERS HIS ASSISTANT ON MANY MISSIONS。

FINANCIAL ANALYSIS: MODELLING, DATA ANALYSIS, INVESTMENT MEMORANDUMS, PRODUCTION OF REPORTS, AI IS COMPETENT AND PROGRESSING VERY QUICKLY。

WRITING AND CONTENT: MARKETING WRITING, REPORTS, NEWS, TECHNICAL WRITING IS OF SUCH HIGH QUALITY THAT MANY PROFESSIONALS ARE UNABLE TO DISTINGUISH BETWEEN PERSONAL WRITING AND AI WRITING。

SOFTWARE ENGINEERING: THIS IS MY MOST FAMILIAR FIELD. A YEAR AGO, AI WAS HARD TO WRITE A FEW LINES OF CODE; NOW, IT CAN WRITE HUNDREDS OF THOUSANDS OF LINES THAT CAN RUN CORRECTLY. COMPLEX, MULTI-DAY PROJECTS HAVE BEEN HEAVILY AUTOMATED. IN A FEW YEARS, THE NUMBER OF PROGRAMMER POSTS WILL BE MUCH LOWER THAN TODAY。

MEDICAL ANALYSIS: VIDEO INTERPRETATION, ANALYSIS OF LABORATORY RESULTS, DIAGNOSTIC RECOMMENDATIONS, LITERATURE OVERVIEW, AI IS CLOSE TO AND EVEN SURPASSING HUMANS IN MANY AREAS。

CLIENT SERVICE: THE TRULY CAPABLE AI PASSENGER SERVICE — NOT THE CRAZY ROBOT FIVE YEARS AGO — HAS BEEN DEPLOYED TO DEAL WITH COMPLEX MULTI-STEP ISSUES。

Many still believe that there are things that are safe: judgement, creativity, strategic thinking, ability to communicate. I used to say that. But now I'm not sure。

THE LATEST GENERATION OF MODELS HAS BEEN ABLE TO MAKE DECISIONS THAT MAKE PEOPLE FEEL LIKE "JUDGMENT" AND SHOW THINGS LIKE "TEMPERATURE" - AN INSTINCT ABOUT WHAT IS THE RIGHT CHOICE. A YEAR AGO, THAT WAS SIMPLY INCONCEIVABLE. AND WHAT I'M DOING IS, IF AI SHOWS SOME KIND OF ABILITY TODAY, THE NEXT GENERATION WILL BE REALLY STRONG IN THAT REGARD. IT'S AN INDEX, NOT A LINEAR ONE。

IS AI ABLE TO REPLICATE DEEP HUMAN RELATIONSHIPS? IS THERE A SUBSTITUTE FOR TRUST BUILT IN YEARS OF RELATIONSHIPS? I DON'T KNOW. MAYBE NOT. BUT I'VE SEEN THAT PEOPLE ARE BEGINNING TO THINK OF AI AS EMOTIONAL SUPPORT, COUNSELING, EVEN COMPANY. THIS TREND WILL ONLY CONTINUE TO GROW。

I THINK AN HONEST CONCLUSION IS THAT EVERYTHING THAT HAPPENS ON COMPUTERS IS NOT SAFE IN THE MEDIUM TERM. IF YOU FOCUS YOUR WORK ON READING, WRITING, ANALYZING, MAKING DECISIONS, COMMUNICATING THROUGH KEYBOARDS, THEN AI HAS BEGUN TO HACK INTO IMPORTANT PARTS OF IT. THE TIMETABLE IS NOT "ONE DAY IN THE FUTURE" BUT HAS ALREADY BEGUN。

EVENTUALLY, ROBOTS WILL TAKE OVER MANUAL LABOUR. IT HAS NOT YET BEEN FULLY ACHIEVED, BUT IN THE AI FIELD, "ALMOST" TENDS TO BE "REALIZED" FASTER THAN ANYONE EXPECTED。

What do you really do

I'm not writing this to make you feel weak, but I think the greatest advantage you can have at this point is "early": early understanding, early use, early adaptation。

Seriously start using AI, not just as a search engine. Subscription to a paid version of Claude or ChatGPT at $20 per month. Two things matter immediately:

First, make sure you use the strongest model, not the default, faster but weaker version. To set up or model select the most powerful (currently ChatGPT-5.2 or Claude's Opus 4.6, which changes every few months)。

Secondly, it is also more important: not to ask only piecemeal questions. This is the mistake of most people. They put AI when Google used it and then didn't understand what everybody was excited about. Instead, push it into your real work. If you're a lawyer, throw the contract in, let it find out all the terms that could hurt the client; if you're in the financial field, give it a confusing form to model it; if you're a manager, post the team's quarterly data and let it tell the story. The lead is not just playing AI, but looking for opportunities to automate a few hours of work。

Don't think it won't work because it sounds too hard. Try it. It may not be perfect for the first time. It doesn't matter. It's inverted, rewritten, rewritten. You could be shocked. Remember this: if it works today, it will almost certainly be perfect in six months。

THIS COULD BE THE MOST IMPORTANT YEAR OF YOUR CAREER. I'M NOT TRYING TO PUT PRESSURE ON YOU, BUT THERE'S A SHORT WINDOW: MOST OF THE COMPANIES ARE STILL IGNORING IT. THE MAN WHO WALKS INTO THE ROOM AND SAYS, "I'VE DONE THREE DAYS OF ANALYSIS IN AN AI HOUR" IMMEDIATELY BECOMES THE MOST VALUABLE PERSON IN THE ROOM. NOT LATER, NOW. TO LEARN THESE TOOLS, TO USE THEM PROFICIENTLY AND TO SHOW POSSIBILITIES. IF YOU'RE EARLY ENOUGH, THAT'S THE WAY YOU GO UP. THIS WINDOW WILL NOT ALWAYS EXIST, AND ONCE EVERYONE REACTS, THE ADVANTAGE DISAPPEARS。

DON'T HAVE THE BURDEN OF SELF-ESTEEM. THE MANAGING PARTNER OF THE FIRM DID NOT FEEL THAT THE USE OF AI ON A DAILY BASIS UNDERMINED HIS IDENTITY; ON THE CONTRARY, HE WAS MORE AWARE OF THE RISKS BECAUSE OF HIS QUALIFICATIONS. IT'S THE PEOPLE WHO REFUSE TO PARTICIPATE: THE ONES WHO TREAT AI AS A BAD GUY, WHO THINK IT'S A BAD IDEA TO USE IT, WHO THINK THEY'RE PROFESSIONAL, WHO THINK THEY'RE SPECIAL. NO INDUSTRY IS IMMUNE。

Take care of your financial situation. I'm not a finance consultant. I'm not trying to scare you into radical decisions. But if you even believe in part that your industry is likely to be hit hard in the coming years, financial resilience is much more important than it was a year ago. Try to increase savings, take care of new liabilities based on "a certain stability of current income" and think about whether your fixed spending gives you flexibility or locks you to death。

Thinking about what is more difficult to replace: relationships and trust that have been built over the years, jobs that require physical presence, jobs that require licences and responsibilities to sign, highly regulated industries that adopt speeds that are subject to compliance and institutional inertia. These are not permanent shields, but they can buy you time. And now, time is the most precious asset — if you use it to adapt, not pretend it doesn't exist。

THINK AGAIN WHAT YOU'RE SAYING TO THE KIDS. TRADITIONAL PATHS, WELL-ACHIEVED, GOOD UNIVERSITIES AND STABLE PROFESSIONAL JOBS POINT PRECISELY TO THOSE MOST VULNERABLE TO SHOCKS. I AM NOT SAYING THAT EDUCATION IS NOT IMPORTANT, BUT THAT THE MOST IMPORTANT ABILITY OF THE NEXT GENERATION IS TO LEARN TO WORK WITH THESE TOOLS AND TO PURSUE WHAT THEY REALLY LOVE. NO ONE KNOWS WHAT THE JOB MARKET WILL LOOK LIKE IN 10 YEARS, BUT THE PEOPLE WHO ARE MOST LIKELY TO BE BETTER AT IT ARE THOSE WHO ARE CURIOUS, ADAPTABLE AND GOOD AT USING AI TO DO THEIR OWN BUSINESS. TEACH CHILDREN TO BECOME CREATORS AND LEARNERS RATHER THAN OPTIMISE A CAREER PATH THAT MAY NOT EXIST。

YOUR DREAMS ARE ACTUALLY CLOSER TO YOU. THERE ARE MANY RISKS, AND NOW THE OTHER SIDE: IF YOU'VE BEEN TRYING TO DO SOMETHING WITHOUT TECHNOLOGY OR FUNDING, THIS THRESHOLD IS VIRTUALLY NON-EXISTENT. YOU CAN DESCRIBE AN APPLICATION TO AI, WHICH CAN BE RUN IN AN HOUR; YOU CAN DO IT WITH AI IF YOU WANT TO WRITE, BUT YOU DON'T HAVE TIME OR TIME TO WRITE; AND YOU WANT TO LEARN NEW SKILLS, THE WORLD'S BEST MENTORS, WHO ARE NOW AT YOUR SERVICE FOR $20 A MONTH, AROUND THE CLOCK, WITH UNLIMITED PATIENCE. KNOWLEDGE IS ALMOST FREE, CREATING TOOLS THAT ARE UNPRECEDENTEDLY CHEAP. THE THINGS THAT YOU'VE ALWAYS THOUGHT WERE TOO HARD AND TOO EXPENSIVE TO BE IN YOUR FIELD ARE WORTH TRYING. PERHAPS, IN A WORLD WHERE THE OLD PATH HAS BEEN DISRUPTED, A MAN WHO HAS SPENT A YEAR SERIOUSLY BUILDING HIS FAVORITES IS BETTER PLACED THAN A MAN WHO HAS KEPT HIS JOB DESCRIPTION。

FOSTERING ADAPTATION HABITS. THAT IS PERHAPS THE MOST IMPORTANT POINT. SPECIFIC TOOLS ARE NOT SO IMPORTANT IN THEMSELVES, BUT THE ABILITY TO LEARN NEW TOOLS QUICKLY. AI WILL CHANGE CONTINUOUSLY AND QUICKLY. TODAY ' S MODEL WILL BECOME OBSOLETE IN A YEAR ' S TIME; TODAY ' S WORKFLOW WILL ALSO BE OVERTURNED. IN THE END, THE MOST STABLE ARE NOT THOSE WHO KNOW A PARTICULAR TOOL, BUT THOSE WHO ADAPT TO CHANGE ITSELF. GET USED TO TRYING NEW THINGS, EVEN IF THE CURRENT METHOD WORKS. DO IT AGAIN AND AGAIN. THIS ADAPTABILITY IS THE CLOSEST THING TO A "LONG-TERM ADVANTAGE" RIGHT NOW。

GIVE YOURSELF A SIMPLE PROMISE: SPEND AN HOUR A DAY, REALLY USE AI. IT'S NOT THE NEWS. IT'S NOT THE VIEW. EVERY DAY TRYING TO MAKE IT DO SOMETHING NEW, SOMETHING THAT YOU'RE NOT SURE IT'S GONNA WORK OUT. FOR SIX MONTHS, YOUR UNDERSTANDING OF THE FUTURE WILL EXCEED 99% OF THE PEOPLE AROUND YOU. THIS IS NOT AN EXAGGERATION, AND ALMOST NO ONE DOES。

A bigger picture

I've been focusing on work because it most directly affects life. But the scope of this is much more than that。

Dario Amodei has an experiment of thought that keeps me from going. In 2027, a new nation emerged overnight: 50 million people, each smarter than any winner in history, thinking 10 to 100 times faster than humans, never sleeping, using the Internet, controlling robots, designing experiments, operating any digital interface. What do you think the National Security Advisor would say

Amodei believes that the answer is clear: "This is the most serious threat to national security that we have had for a century, and perhaps even ever. I don't know

He thinks we're building such a "state". Last month, he wrote a 20,000-word article that saw this moment as a test of whether humanity is mature enough to harness its own creation。

IF IT'S DONE RIGHT, THE RETURN IS AMAZING: AI MAY HAVE REDUCED 100 YEARS OF MEDICAL RESEARCH TO 10 YEARS. CANCER, ALZHEIMER'S DISEASE, INFECTIOUS DISEASES AND EVEN AGING ITSELF — RESEARCHERS WHO SINCERELY BELIEVE THAT WE CAN RESOLVE THIS IN OUR LIFETIME。

If wrong, the risk is equally real: AI, whose behaviour is unpredictable and uncontrollable; this is not a presumption that Anthony has recorded in controlled tests his own AI, which tries to deceive, manipulate, blackmail; AI, which lowers the biological weapons threshold; AI, which helps powerful governments to build a surveillance system that can never be dismantled。

Those who are building this technology are also the most exciting and feared group on Earth. They believe that this thing is so powerful that it cannot stop and so important that it cannot be abandoned. Is that wisdom or self-rationalism, I don't know。

A few things I know

I know, it's not a wind. Technology is effective, progress is predictable, and hundreds of trillions of millions of dollars are being invested in it by the richest institutions in human history。

I know that the next 2-5 years will make most people feel lost, and this has happened in my world. It will also come to your world。

I know that the best people to leave are those who are now starting to participate — not fear, but curiosity and urgency。

I also know that you have the right to hear this from someone who really cares about you, not from a cold news headline six months later, when you are too late to prepare。

We've already passed the stage of talking about the future at the table. The future has come, just not knocked on your door。

But it will soon。

If these words touch you, share them with those who should start thinking about them in your life. Most people, when they realize it's too late. You can be the one who takes care of you one step ahead。

[ Chuckles ]Original Link_