From the Big Bang to a stick man and a phone number, humans have spent civilization trying to make the complex simple. AI may finally become extraordinarily good at it.
You hear about a company that has just raised $1 billion at a $10 billion valuation and is building what it hopes will become your personal AI agent. Naturally, you go looking for it.
You expect the usual technological plumage. Product demos. Founder interviews. Benchmark charts. A video of somebody accomplishing something breathtaking while apparently living in an apartment containing no electrical cords.
Instead, you find a white page.
A stick man.
A cryptic sentence.
And a box asking for your phone number.
Welcome to Instinct.
At first, the whole thing feels faintly ridiculous. A $10 billion company apparently operating its marketing department from a mud hut.
Then you start following the breadcrumbs. And the stick man begins to make sense.
Instinct was founded in 2025 by Noah Shinn, an early employee of Sierra, the enterprise AI-agent company.
Shinn’s background is unusually relevant. Before Instinct, his research included work on Reflexion, which explored agents learning from feedback, and τ-bench, designed to evaluate how well agents can use tools while interacting with humans in realistic situations.
Then came Sierra, where agents were being built to work on behalf of companies. Instinct turns the telescope around.
The agent works for you.
But the clue to the larger idea is sitting right on Instinct’s almost comically austere website: “The interface is simple: there are no new interfaces.”
You text it. You call it. It operates a virtual phone and computer.
Suddenly the empty website looks less like something they forgot to finish. It looks like a thesis.
Even the name is almost poetically appropriate.
Instinct is one of nature’s great compression mechanisms. Behind the sensation of fear lies an almost incomprehensibly complicated arrangement of sensory processing, memory, neural activity, hormones, muscles and evolutionary history.
The organism experiences: Run.
Hunger compresses another enormous biological apparatus into: Eat.
Protection. Desire. Pain. Curiosity.
Instinct is complexity made actionable.
Whether Shinn intended anything remotely this philosophical when choosing the name is beside the point. There is no reason to turn a young AI founder into Aristotle with a term sheet.
The name simply points toward something humans have been doing for a very long time. We compress.
Reality arrives without folders. We invented those.
Humans encounter an overwhelming world and immediately begin cutting it into manageable pieces. We name things. Then categorize the names. We create alphabets, taxonomies, maps, filing systems, equations, libraries, databases and search engines.
Eventually we invent software to manage the databases containing the categories we created to understand the things we named. Then we need software to manage that software.
Things occasionally get out of hand.
Understanding itself is partly an act of compression.
We keep drilling downward, hoping to discover something simpler underneath. Then we use those simpler pieces to reconstruct the whole.
Here philosophy becomes useful precisely because it isn’t physics.
Nature appears extraordinarily good at producing complexity from simpler beginnings. The early universe eventually produced particles, atoms, stars, planets, chemistry, cells, organisms, nervous systems and human beings.
Then one of those human beings invented the password-reset email.
Complexity had achieved consciousness.
Run our descriptions of the universe backward far enough and classical general relativity eventually takes us toward what physicists call a singularity, a point at which the theory itself ceases to provide an adequate description. It shouldn’t be casually described as a little object that exploded into the universe. Modern cosmology is subtler than the popular image of the Big Bang.
But philosophically, the symmetry is irresistible.
Behind us, singularity becomes complexity. Ahead of us, we imagine complexity collapsing toward singularity. And here we are in the middle, compulsively organizing the furniture.
The second law of thermodynamics doesn’t say that your garage must become cluttered. Your garage merely seems deeply committed to independent verification.
Entropy is a precise physical concept, and civilization is not literally defeating the second law. Creating local order requires energy and produces entropy elsewhere.
But philosophically, entropy makes a wonderful antagonist.
Things disperse. Structures deteriorate. Information accumulates. Organizations become bureaucracies. Software acquires features. Closets acquire objects nobody remembers purchasing.
Human beings respond by expending enormous amounts of energy trying to make things easier.
We will move heaven and Earth to avoid moving heaven and Earth.
Computers were supposed to simplify things. They did. Then things got complicated.
Machine instructions became command lines. Command lines became graphical interfaces. Graphical interfaces became windows. Windows became browsers. Browsers became search engines. Computers became smartphones. Smartphones became platforms. Platforms filled with apps.
Eventually we found ourselves using password managers to manage the passwords required to access the software designed to simplify our lives.
Each successful simplification created new capability. New capability created new complexity.
Hypertrophy.
The interface grew muscular. More tabs. More menus. More dashboards. More icons. More notifications. More software for managing the software managing the software.
And beneath it all, the human being still wanted something remarkably simple.
Imagine an immutable ring.
Now imagine an enormous, spinning, vibrating bundle of wires being forced through it.
The wires fight one another. They twist. They bunch. They throw off heat. They contain airlines, banks, calendars, databases, payment processors, loyalty programs, passwords, rules, preferences, APIs and several thousand edge cases nobody considered until somebody in Des Moines tried something unusual on a Tuesday.
The ring doesn’t get larger. Because the ring is human intention.
Get me to Chicago Tuesday. I need to be home Thursday night.
That’s it.
Those are procedures imposed between intention and result. For most of computing history, we solved this by teaching humans how to navigate the procedures.
AI agents suggest something radically different.
Make the machinery navigate itself.
This is why calling AI agents another interface misses what may be happening. They could become a compression layer.
Search engines compressed information retrieval. Smartphones compressed computing devices. Graphical interfaces compressed machine commands into visual metaphors. Natural language compressed commands into conversation.
Agents may compress something larger:
Procedure into intention.
Instead of learning how software works, you tell software what you want. The agent figures out the procedure.
That is not merely a better menu. It is an attempt to eliminate the need for the menu.
Screens aren’t going to disappear next Thursday. Neither are buttons, dashboards or apps.
But the interface as we have understood it for roughly forty years is under heavy bombardment.
The old interface answers: How do I operate this machine?
The emerging interface increasingly asks: What do you want?
The visible interface could actually become simpler while the invisible system underneath becomes vastly more complicated.
Greater complexity below. Greater simplicity above.
Which sounds wonderful. Until you notice what disappeared with the interface.
A traditional interface doesn’t merely create friction. It exposes decisions.
You select the flight. You see the price. You choose the hotel. You approve the purchase. You press BUY.
Now tell an agent: Take care of my trip.
The sentence is gloriously simple. The decision tree underneath it isn’t.
Which airline? Which airport? How much may it spend? May it upgrade you? Can it use your points? What information may it disclose? Can it negotiate? Can it accept something nonrefundable? Can it speak as you?
We have spent civilization trying to compress complexity into simplicity. AI may finally become extraordinarily good at it. The danger is forgetting that the complexity didn’t vanish. We merely stopped looking at it.
That is where interface design stops being merely design. It becomes governance.
Perhaps the interface doesn’t disappear at all. Perhaps it changes jobs.
Today’s interface primarily helps us operate the machine. Tomorrow’s interface may increasingly help us supervise it.
Trust becomes the interface.
The same capability that makes the system useful makes visibility more important, not less.
The interface approaches zero precisely as the authority approaches everything.
There’s still that phone-number box.
Instinct says there are no new interfaces, but technically there is one. A white rectangle. Type your phone number. Press the arrow.
An interface stripped nearly to its mathematical minimum.
But if human technological history really is a relentless attempt to compress the distance between intention and action, why assume the compression stops there?
Typing is friction. Talking is friction. Taking a phone out of your pocket is friction. Formulating an instruction is friction.
The next logical question becomes slightly uncomfortable.
What comes after language?
Brain-computer interfaces are no longer exclusively science fiction. Companies including Neuralink are developing systems intended to establish direct communication between neural activity and computers.
That doesn’t mean your grandchildren are going to order pizza telepathically. But follow the trajectory.
Machine code → keyboard → mouse → touchscreen → voice → natural language → agent → intention.
Not because somebody decided civilization desperately needed another gadget implanted in its skull. Because complexity keeps encountering the same pressure.
Simplify. Remove the intermediary. Shorten the distance. Make the machinery disappear.
Maybe that’s what we’re actually witnessing.
Not merely the rise of artificial intelligence. Not merely the death of the app. Not another round of Silicon Valley interface fashion.
Something older.
Nature produces complexity with astonishing enthusiasm. Human beings encounter it and begin organizing.
We name. Categorize. Measure. File. Abstract. Model. Compress.
Then we build machines to help us compress faster. Eventually those machines become complicated enough that we build intelligent machines to operate them.
And if those intelligent machines become sufficiently capable, the great tangled bundle of modern technological life begins squeezing through an extraordinarily small ring:
What do you want?
The wires remain. The databases remain. The APIs remain. The payment networks remain. The computational infrastructure grows larger. The data centers consume more energy. The machinery becomes almost incomprehensibly complicated.
But from the human side?
A white page.
A stick man.
A cryptic sentence.
A box asking for your phone number.
Instinct.
Its stated design philosophy is that users should communicate by text or phone rather than learn another elaborate software interface. The larger argument here is our interpretation of where that approach could lead.
No. It can hide or reorganize complexity from the user’s perspective. The underlying systems may become more complicated even as the visible interaction becomes simpler.
No. Entropy is a precise physical concept. Here it serves as a philosophical metaphor for humanity’s recurring effort to create local order and intelligibility from complexity.
No. Neural interfaces are presented as the logical extreme of the compression trend, not a prediction that implanted interfaces will replace phones or language.