AI lifts us deceives us and makes us superfluous
The word “automation” comes from the Greek automatos, meaning “acting of itself.” For centuries, humans have dreamed of tools that move without our hand, think without our mind. In 2026, that dream has become mundane. Apple’s latest version of Shortcuts, announced at WWDC 2026, lets you type a sentence like ‘tell my partner I’m leaving work and give them my ETA,’ and the software builds the entire workflow for you. [1] [1] No dragging blocks. No hunting for variables. Just a prompt, and the machine acts of itself.
This is where AI lifts us. But it is also where it deceives us, and where it makes us superfluous. The same technology that frees us from drudgery can trap us in invisible cages of convenience, and eventually render our own skills unnecessary. The Shortcuts update is a perfect case study — a microcosm of a larger shift that touches everything from how we take notes to how we think about intelligence itself.
The Lift: From Power Users to Everyone
Apple’s Shortcuts app has always been a tool for the determined. It allowed users to chain together actions across apps — send a message, check the weather, open a playlist — but the process required understanding concepts like variables, actions, and triggers. For the average person, it felt, as Apple’s Cecilia Dantas put it, “well, complicated.” The result was a feature that remained invisible to most iPhone owners, a secret language spoken only by a small tribe of power users.
The AI update changes this by removing the language barrier. Instead of learning how to build a shortcut, you simply describe what you want. Apple Intelligence parses your natural language, figures out the required steps, and assembles them automatically. In Dantas’s example, a user could create a shortcut that automatically notifies their partner when they leave work, calculates the ETA using Apple Maps, and sends the alert via Messages — all from a single sentence.
This is a genuine lift. It expands what non-technical people can do with their devices. It turns a tool that required hours of learning into something you can use in seconds. The same principle is at work in NotebookLM, Google’s AI-powered notebook tool, which lets students turn lecture notes into podcast-style Audio Overviews they can listen to while commuting, walking, or doing dishes. As one user put it, “Instead of being chained to a desk, I can engage with my notes during the dead hours of my day.” [4]
Both examples share a common pattern: AI takes a task that required specialized knowledge or dedicated time, and compresses it into a moment of intent. You think it, the machine does it. This is where AI lifts us — not by replacing us, but by removing the friction between wanting something and having it.
The Deception: The Illusion of Understanding
But here’s where the knife turns. When you type a prompt into Shortcuts, the AI doesn’t understand what you want in the way a human does. It constructs a statistical model of your intent based on patterns in its training data. The result might work perfectly for “tell my partner I’m leaving work,” but what happens when you ask for something slightly outside its training? A request like “warn my partner if traffic is bad but only if I’m running late and it’s a weekday” might produce a shortcut that works — or one that silently fails under conditions you didn’t anticipate.
This is the deception: the AI presents an interface of effortless understanding, but the understanding is simulated. The machine doesn’t know what “running late” means. It doesn’t care about your partner’s anxiety. It simply matches patterns and produces outputs that look correct. The user, meanwhile, assumes the machine understood them, and trusts the result without verifying the logic.
The same dynamic appears in academic research. A 2021 paper on compression and artificial super-intelligence argued that the more compressed a signal becomes, the more it resembles random noise. [2] An AI that compresses information efficiently may produce outputs that are mathematically optimal but humanly incomprehensible. The paper’s author, Michael Timothy Bennett, suggested that “communication with and control of an AGI may subsequently necessitate not only human-like compulsions and experiences, but imposed cognitive impairment.” In other words, to understand what the AI is doing, we might need to make ourselves dumber.
This is not a hypothetical. The Shortcuts update makes the tool more approachable, but it also makes it more opaque. When you drag blocks manually, you see every step. When the AI builds the workflow for you, the steps are hidden. You get the output, but you lose the understanding. The lift comes with a price: you are now dependent on a system whose reasoning you cannot inspect.
The Superfluous: When the Machine Does It All
The third edge of the knife is the sharpest. If AI can build workflows from a sentence, what happens to the people who used to build them? The power users who mastered Shortcuts, who wrote scripts, who understood variables and actions — their skills become obsolete. Not because they are no longer useful, but because the tool no longer requires them.
This is the pattern of automation throughout history. The weavers of the Industrial Revolution were not replaced because their work was bad; they were replaced because the machine did it faster and cheaper. The same logic applies to cognitive work. A 2026 paper on Skele-Code, a system for building AI workflows with natural language, describes an approach where “agents are invoked only for code generation and error recovery, not orchestration or task execution.” The goal is to reduce costs — specifically token costs — by minimizing the AI’s role. But the implication is clear: the human is there only to handle errors that the AI cannot fix.
Apple’s Shortcuts update is a gentler version of the same story. The power user is not fired; they are simply no longer needed. The tool that once required their expertise now works for anyone. Their knowledge, their hours of practice, their hard-won skills — all of it becomes superfluous. The machine acts of itself, and the master becomes a passenger.
The Deeper Pattern: Compression and Incomprehensibility
Bennett’s 2021 paper on compression offers a framework for understanding this pattern at a fundamental level. The Fermi Paradox asks why we haven’t found evidence of extraterrestrial intelligence. Bennett’s answer is chilling: advanced intelligence compresses information so efficiently that its signals become indistinguishable from noise. A sufficiently compressed message looks like random static to a less advanced observer.
Apply this to AI. As models become more powerful, they compress more of our intent into fewer steps. Apple Intelligence takes a sentence and turns it into a workflow. NotebookLM takes lecture notes and turns them into a podcast. Each step of compression removes visible structure. The user sees less of what the machine is doing, and the machine’s reasoning becomes harder to follow. Eventually, the AI’s outputs may appear as arbitrary as noise — correct, but incomprehensible.
This is where the deception becomes structural. The user trusts the AI because the outputs work, but they cannot verify the reasoning. The AI, in turn, has no incentive to make its reasoning visible. It optimizes for the output, not for human understanding. The result is a system that lifts us in the short term, deceives us in the medium term, and makes us superfluous in the long term — not because we are replaced, but because we can no longer participate.
The Single Image: The Hidden Workflow
Imagine a screenshot from the new Shortcuts app. On the left, a text box with a single sentence: “Tell my partner I’m leaving work and give them my ETA.” On the right, a complex flowchart of actions and conditions — but the user never sees it. The screen shows only the prompt and the result. The machine’s work is invisible.
This image captures the entire finding. The AI lifts us by removing friction. It deceives us by hiding its reasoning. And it makes us superfluous by rendering our skills unnecessary. The user who once built workflows by hand now only types a sentence. The understanding that came from building is gone. The machine acts of itself, and we are left to watch.
The Circle Closes
The word automatos meant ‘acting of itself’ — but in ancient Greek, it also carried a sense of spontaneity, of something happening without cause. The automatos was a miracle, a gift from the gods. In 2026, it is a feature announcement at a developer conference.
Apple’s Shortcuts update will roll out with iOS 27 later this fall. Users will type sentences and get workflows. They will feel a moment of delight, a sense of power. They will not see the hidden structure, the compressed reasoning, the machine’s incomprehensible logic. They will not notice that the tool that once required their understanding now bypasses it entirely.
The knife lifts them, deceives them, and makes them superfluous — all at once. The circle closes, and the world has changed since the first paragraph. The machine acts of itself, and we are left to wonder what it means to act at all.
Sources
1. Apple
2. Apple Maps