Plenty of people are already writing the technical timetable — what improves when, which capability lands in which month, what the curve of compute does next. What has not been written is the painful half: what a person goes through while it happens.
What we want to predict is people.
Three years from now, what are people like? Ten years from now, what is daily life like? Brain, thinking, education, marriage, family, relationships, mobility, human-machine interaction — as AI accelerates and then reaches toward AGI, how do these change? Here are our current answers. Overturn them.
A note on names. In English, "AI 2027" usually means a well-known month-by-month scenario report about where AI development goes. That is a different project from ours, and we are not writing a correction to it — it is serious work, and roughly half of it is about the human situation. We are simply working on the other axis: not what the technology does, but what it does to the person using it.
Technical forecasts now come with dates attached: how clumsy agents will be, how far coding automation goes, where compute concentrates, who overtakes whom and when. That is a prediction of the machine.
"What scares me is not that AI becomes powerful. It's that people go through something painful while it does — and that we get far less preparation time than we did for the steam engine, for electricity, for the industrial era. It is a right-angle turn."
Jerry Jin · phone call, Oct 13 2026
The steam engine gave humanity three generations of runway. Electricity gave two. The internet gave one. This time there is none. The curve is a right angle: you may be working normally one quarter, and the next quarter the social meaning of your job has disappeared. Not unemployment — the premise that "this needed a human to do it" has been pulled out from under it.
So we move the lens off the machine and onto the person, and ask a few slow questions:
In 2030, how does a 25-year-old prove they are useful?
In 2030, what should parents teach their children that will not harm them?
In 2040, what does the word "human" still refer to?
Everybody compares AI to the industrial revolution, the steam engine, electricity, the internet. We think that analogy is wildly wrong — not a matter of degree, but a category error.
An asteroid about 10 kilometres across hit the Earth. Within a hundred kilometres of the impact point, living things were vaporised instantly.
But that is not how the dinosaurs died. After the impact, dust was thrown into the atmosphere and blotted out the sun — for two full years.
"The rules of survival changed. The old rule was bigger, stronger, dominant wins. After that, because there was no sun for two years, the high-consumption animals died first — the photosynthetic plants were finished; while the scavengers, the burrowers, the low-consumption ones survived."
Jerry Jin · phone call, Oct 13 2026
Notice what the dinosaurs actually died of: not the impact — they failed to adapt after the rules changed. The T-rex did nothing wrong. It was just doing the old rules to perfection.
And every ancestor of ours was one of those small, low-consumption, light-tolerant mammals. That is the moment we actually begin.
AI is not "a faster steam engine." The difference is this: the steam engine changed energy; AI changes judgement — and "who does the judging" has always been the ecological niche that our species occupied alone.
So the real question is not "when does AI reach AGI," but:
"When a big rule changes substantially, that's a reshuffle. AI will reshuffle entire industries. So how do we build the new rules? What are the first principles underneath? They will not be the same as the old ones."
Jerry Jin · phone call, Oct 13 2026
The danger of comparing it to the industrial revolution is this: you think you still have time, that yesterday's optimal answer just needs some optimisation. No. The perfect answer under the old rules can be a dead end under the new ones.
The table below is the most important thing this project has produced so far. It is incomplete, and we expect at least a third of it to turn out wrong — but saying it out loud is what makes it possible for others to overturn.
| Dimension | Old rules (the last few thousand years) | New rules (taking shape now) |
|---|---|---|
| Who survives | Big, strong, holds a lot of resources | Low consumption, fast to turn — the higher your fixed costs, the more dangerous you are |
| Core competitiveness | Knows a lot, remembers well, computes fast | Asks precisely, judges precisely — knowing which source to believe |
| What a person is worth | Priced by degree, seniority, title | Priced by verifiable process data |
| What is scarce | Information, channels, capital | Attention, authenticity, willingness to bear consequences |
| Organisational form | Hierarchy, economies of scale | Individual + agent swarm — size is no longer an advantage |
| The definition of a person | Carbon-based, the only species that can judge | "Intelligence" includes carbon and silicon — the boundary is being negotiated |
| Species form | Natural person: AI is just a web page I open | AI person: natural person + resident personal AI infrastructure (see §05) |
| Generational direction | One generation teaches the next | Two-way, sometimes reversed — children's intuition for the new rules is sharper |
We do not project month by month — that is someone else's method, and their strength. We mark time by changes a person can feel.
Every line below carries our confidence level: high = we think this is probably right; mid = the direction is right, the timing and the form are uncertain; low = we are not sure ourselves, but we think it has to be asked.
NowThe hand is freed (keyboard), the mouth is freed (voice), and next the mouth is skipped too — straight from brainwaves. The hardware threshold for this has already fallen to consumer level.
2030Memory externalisation completes. Intelligence is no longer measured by "how much you remember" but by "can you retrieve it, can you verify it." The side effect: ownership of memory becomes a new battleground in law and ethics — is your external memory yours, or your provider's?
2040"What I thought of" and "what it generated" become indistinguishable in many settings. The first officially named cognitive atrophy appears: not dementia, but judgement degenerating from long-term outsourcing.
NowThe main action of thinking is shifting from "solving" to "asking." Most people have not switched yet — they still use the grammar of questions to demand answers.
2030Critical thinking is redefined: no longer "can I derive this independently" but "can I design a test that makes it fail." The object of suspicion moves from the source of information to the reasoning process.
2040The smallest unit of thought shifts from the sentence to the intention. And the scarcest capability becomes not typing anything — in an environment where you can always ask, resisting the question and thinking it through yourself is the new threshold.
NowEducation systems still reward "memorising and reproducing," which is exactly the part that gets replaced first. This mismatch is the biggest current waste.
2030The degree accelerates as a capability signal and is replaced by verifiable work and process data. "Where did you study" gives way to "what did you do, and how."
2040The core subject of education is no longer knowledge but how to live alongside an omniscient object and still keep a self. That becomes formal basic training, like literacy.
NowHuman-machine emotional relationships have already moved from science fiction to a real option, and society has no language to describe them.
2030The first people "living long-term with an AI" start going public. Marriage's economic function of mutual support is stripped away, and what is left has one word in it: willingness. That makes marriage simultaneously purer and more unstable.
2040Marriage degrades from an institution into a personal preference; procreation and marriage fully decouple. The object of loyalty is no longer "this person" but "the definition of this relationship" — and that definition has to be renegotiated continuously by both sides.
"I told my son to play fewer games and do more homework… and in the end I pushed him too far and he said: What's the point of all this talk — we're the last generation of humans anyway."
Jerry Jin · phone call, Oct 13 2026
NowChildren's intuition for the new rules is often sharper than their parents'. The direction of knowledge flow between generations has already begun to reverse.
2030The core function of the family shifts from "raising and supporting" to emotional anchor — providing "the unchanging part" in an environment where everything else can change.
2040Family forms become plural as the norm: person-person, person-AI, multi-person multi-agent units. "My grandchild might be an AI" moves from a joke to an option that has to be taken seriously.
NowSocial currency shifts from "I know something you don't" to "I can prove I am real."
2030Authenticity becomes the scarce good. "There is a real person on the other side" has to be proven, and the proof becomes new social infrastructure. Connections get fewer, more expensive, more deliberate.
2040Loneliness and connection rise together — the defining condition of this generation. The extension of the word "companion" formally comes to include non-human intelligences.
NowMobility is shifting from "I drive" to "I am being routed."
2030The car becomes fully a third space. The value of time in transit is re-priced — it may be the last stretch of time that cannot be done by AI for you and that you have to spend yourself.
2040Physical movement itself becomes optional. Most mobility demand is dissolved by local production and remote presence; "showing up in person" becomes a gesture of goodwill rather than a necessity.
NowStage one of three: use AI as a tool.
2030Stage two: as a partner. Everyone has their own agent swarm, which negotiates, socialises and executes on your behalf. Your output is no longer equal to your labour — it equals the compute you command × the quality of your judgement.
2040Stage three: being represented. A person's main work becomes setting goals and boundaries for your own agents — which is, precisely, the part of being human: deciding what you want, and what you do not.
In 2026 Edward Dai wrote an article about giving yourself a personal AI server. He was talking about infrastructure. What we took from it was a different thing — a species.
Those who go all-in on AI will be the new species of the next ten years, to 2040. Let us give it a name: AI person.
Jerry Jin · October 2026
There is a line in Edward's article that we think marks the boundary between the old human and the AI person precisely:
"It turns AI from an application you open into infrastructure you own."
Edward Dai · "Give yourself a personal AI server" (2026, Chinese)
Notice the difference between the two words: opening is an action, owning is a state. What you open can be replaced, closed, or repriced. What you own grows onto you, accumulates, becomes part of you.
Edward's other judgement matters just as much — and it is not about pushing the cloud away:
"The cloud handles the most powerful brains; the local device handles the one that knows you best. Together, those two are the complete form of the next generation of personal AI."
Edward Dai · same article
Every new species comes with a new organ. When the smartphone arrived, the "digital person" grew a connected organ it carried everywhere. The AI person grows something else: a personal AI hub running 24/7 — local model, local knowledge base, local agents, remote access. The laptop, the phone and the tablet are just its clients.
Edward's definition: your data + your knowledge base + your model + your agents. Those four together are what a real personal AI is.
| Dimension | Old human (an AI user) | AI person |
|---|---|---|
| Where the AI lives | A web page in a browser | A personal node running 24/7 |
| How you use it | I open it when I need it | It is always there |
| Memory | In your head; you forget, it distorts | Externalised, retrievable, portable |
| Data sovereignty | Held by the provider | The data is yours (settled locally) |
| Who acts | You do it yourself | An agent swarm negotiates and executes for you |
| Source of capability | Whatever I learned, I can do | My knowledge base + my model + my agents |
| When the network drops | Disabled | Still running locally |
This section is deliberately left open. If you have answers, write them:
One. In 2040, what is a person without personal AI infrastructure? A new kind of digital
poverty, or a deliberate choice that deserves respect (the way some people refuse smartphones today)?
Two. "Your data + your knowledge base + your model + your agents" — do those four together
already constitute a new subject? And if so, legally, whose is it?
Three. When your agent makes a decision for you, does the responsibility belong to you, to
the agent, or to whoever wrote the model?
We do not have the answers. Our answers may be written today and overturned tomorrow — but the moment one person finishes reading and says "you are wrong," we have succeeded.
This project started from two people's judgements: a pile of phone recordings, a few essays, a few conversations. As more people from more fields join, those early materials will inevitably thin out, get overwritten, get corrected.
That is not a loss. That is the design. We do not want to build a monument. We want to build a foundation. If one day you find that Jerry's and Edward's voices are getting quieter on the page — that means it worked.
We have spent a lot of years doing the same thing: working out what things will look like a few years from now. We are good at it and we genuinely like it — so we turned it into a public project.
Founder of MirrorAI · founder of Human 2030
Repeat entrepreneur and angel investor; in 2015 he was through a nine-figure corporate exit and came out the other side.
Now based in Europe, a regular at European AI summits, tracking Silicon Valley and Central European technology, and researching at depth how the AI frontier changes human society.
He writes and talks about what he sees on a Chinese-language WeChat channel and video account, 靳伟说 (Jerry Jin), with 100,000+ readers and viewers. That channel is in Chinese; this page is the English door into the same work.
In English you can find him on X @jerrykisme and on YouTube @jinweishuo, or write to jinweitalk@gmail.com.
In recent years his work has centred on the "AI person": how a person comes to understand themselves, become more capable and more free through continuous merging with AI. That led to MirrorAI, and to Human 2030 — which puts his judgements about the human future into public discussion, challenge and revision.
Technical background, long working on the front line of AI and engineering. Writes on the Chinese platform 剑胆琴新. Website token2x.com, X @edwardai37598.
AI Is Me Identity and vision — AI is not just my tool, it is my digital extension.
Hand Everything to AI A principle for action — anything that can be standardised,
digitised or delegated should be handed to AI first.
Hand AI to AI The first step, and the base of the whole system — let AI take over AI
itself first.
His questions are the harder ones: what is actually the difference between a person and AI? Can a person be distilled?
Division of labour: Jerry, as lead founder, poses the questions, sets the direction and pushes the argument out to the world; Edward, as co-founder and technical advisor, owns the technical judgements so that the thing actually runs.
They work together but stay independent — each has his own field, audience and voice. Edward runs his own AI Labs, which is his own territory; only his role here is noted. human2030 is not everything either of us is. It is a shared question, not a merger.