Foresight has published an article titled The 5 Most Important Habits for the Future in the AI Era, written by SanTi Li, Nahida and Yinyue. The authors argue that as AI capabilities continue to compound, the scarcest resources are no longer information itself, but inspiration, attention, judgment and the ability to think independently.
In the piece, the writers say that once AI can search, analyze, write, code and even propose solutions within seconds, the bigger risk is not merely failing to use AI. It is becoming so dependent on it that people surrender their own cognitive lead and judgment. Their core argument is that the capabilities that matter most are not the ones that make humans behave more like machines, but the ones that let people use machines without losing what is distinctly human.
1. Disconnect on purpose and return to the physical world
The article starts from the body. It describes the body as the base layer of the human cognitive system and notes that AI responds in milliseconds. A person can open several windows at once, talk to multiple models and push large amounts of work through AI systems. Productivity may appear to jump, but screens and servers do not get tired. Human bodies and brains do.
The authors write that once productivity gains stop being an edge held by a few individuals and become a broad technological update across a group, people can easily slip into a state defined by constant information stimulation, prolonged sitting and repeated task switching. Over time, that setup drains attention and cognitive resources, producing a condition that looks efficient on the surface while quietly overloading the mind and leaving the body still.
The article recommends setting up “digital fasting zones,” including fixed periods away from screens such as the first hour after waking up and the last hour before sleep, or a dedicated block of time with no internet and no smart devices at all. The point, the authors say, is to pull attention back into the physical world and give the brain room to reset.
It also urges readers to use bodily sensation to counter virtual immersion. Regular aerobic exercise and time in natural settings such as basketball courts, football fields, forests and parks are presented as practical ways to restore attention. The article says the nonlinear visual input of natural environments, along with dopamine release tied to movement, can help replenish attention and trigger forms of unconscious creativity, including what it describes as “shower inspiration.”
The authors draw a distinction between “high stimulation” and real rest. Short-video scrolling, social media feeds, intense rhythm games and purposeless chatting with AI are all categorized as digital stimulation, not recovery. According to the article, real restoration comes from movement, deep sleep, sunlight, nature, breathing and real human interaction. Its conclusion is blunt: the body is not a burden on productivity. It is part of cognition itself, and stronger AI makes protecting that physical foundation more important, not less.
2. Build rules for working with AI instead of only learning prompts
The second habit is about structure. The authors say people should not stop at learning prompts, but should build what they call a Human-AI Operating System of their own. In that framework, AI is treated as a digital double or top-tier aide rather than a tool used only once in a while to look up answers.
The article says AI must not take over first principles, especially because hallucinations and confident but incorrect outputs remain common enough to matter. Different models can produce the same kind of misplaced confidence. In the authors’ view, the stronger the models get, the less room there is for humans to drop verification and final decision-making.
What will separate people in the future, the piece argues, is not who has memorized more temporary prompt tricks. It is who can design and lock in a stable system for human-machine collaboration. The article maps that division of labor in explicit terms:
- AI handles information gathering, data organization, first drafts, framework building, code writing, formatting, repetitive work and running through multiple options.
- Humans handle timely feedback, strategic direction, value judgment, risk ownership, final decisions, aesthetic taste, ethical boundaries and understanding of the real world.
The authors also recommend creating tailored system prompts by turning work standards and core principles into a written specification and embedding that into the model setup. One example in the article says AI should distinguish facts from inferences first, clearly mark uncertain information, avoid inventing facts just to complete a formal structure and provide a verification path for important conclusions.
From there, the piece calls for constant iteration of collaboration SOPs. After difficult tasks, users should review where communication costs were highest, why the model misunderstood the task and which rules should have been stated in advance. Over time, that process is meant to harden into a stable standard operating procedure. The authors compare AI to a “digital puppy,” saying clearer rules make it easier for the system to form consistent working habits.
3. Do not grind against AI when your thinking is no longer clear
The third habit addresses a trap the authors see as common in AI use: the machine can keep going forever, but humans should not push themselves without limits. One of AI’s biggest strengths, they write, is also one of its biggest traps. It can always continue. Users can revise a prompt again, ask for another rewrite, switch models, try a new method, run another analysis and keep looping until hours have passed and the original question is no longer clear.
The article frames this as a cognitive hazard specific to the AI era. Machines do not have a fatigue mechanism. People do. When someone is mentally stuck or emotionally anxious, it becomes easy to fall into a loop of prompt tweaking in pursuit of an immediate perfect answer. In that state, repeated back-and-forth with AI drains cognitive energy fast and also makes the model’s processing direction more chaotic, turning the exchange into a tool that burns tokens and money in place.
To illustrate the point, the authors recount an experience with a small algorithm bug that kept codex and claude, described in the article as two top-tier models, revising for 2 days without fixing the problem. After stopping, the writer returned the next day with a clearer head, thought of a completely different approach and had AI solve it through another path in 3 minutes.
Based on that example, the article suggests watching for what it calls signals of “diminishing cognitive returns.” If a user has revised a prompt 3 times in a row without getting a satisfying answer, or has started feeling obvious irritation toward the output, that may mean the line of thinking has entered a blind spot and cognitive energy is largely spent.
The proposed response is a hard stop. When that “overheated” state appears, the article says, step away from the desk, stop interacting with AI and do something like drinking water, walking or taking a shower. That pause gives the brain a buffer so implicit thinking can keep working in the background. The authors also say it helps to move away from obsessive detail work and back toward redefining the problem: What is the real issue to solve? Was a false premise built into the task? Is the problem worth continuing at all? The article adds that just as stock investing can produce sunk-cost psychology, token spending in the AI era may produce something similar.
4. Interrupt AI early and keep control of the interaction
The fourth habit is about steering. The article argues that people should not let AI’s output rhythm take over their own thinking rhythm. Because of how large language models generate text, they can produce answers that look logically coherent while still containing bias or hallucination. If a user passively waits for a full 1,000-word or 2,000-word response before reading, the authors say that user’s own thinking can be pulled off course.
The skill they highlight is iterative steering. When the output drifts away from the core intention, starts piling up empty phrasing or is already headed in the wrong direction from the first sentence, the instruction is to hit “stop generating” right away and avoid spending time on information that does not help. The article says people now have to watch AI reasoning while it unfolds and judge whether the logic has gone wrong, which itself costs energy. In that sense, the authors compare AI to autonomous driving: useful, but not something people can fully rely on without supervision.
Once the output is interrupted, the next step is immediate correction. The article gives an example of adding a constraint that says the model has misunderstood the request and that the user needs not an industry overview but 3 concrete implementation steps.
It also recommends forcing a step-by-step mode instead of accepting giant all-in-one responses. Rather than asking for a complete plan spanning thousands of words, users can ask the model to provide 3 core assumptions first, confirm those assumptions manually and only then expand the first section. The goal is to keep the steering wheel of the interaction in human hands throughout the process.
5. Use inspiration to find opportunities and taste to decide on answers
The fifth habit deals with scarcity. As AI keeps improving, the article says, the execution cost of writing articles, coding, building PPTs and analyzing data is moving closer and closer to zero. It gives one example: if a future top-tier model at the level of “Claude Fable5” were priced like the recent “deepseek flash,” overall costs would fall sharply. Once output becomes extremely cheap, the scarce questions become whether an idea should be published, what is worth doing, why it is worth doing and whose problem should be solved.
AI is good at solving problems, the authors write, but humans still have to choose the problem. The article presents inspiration as a competitive edge and argues for a shift from laborer to creative director. AI may hold a vast knowledge base, but it cannot generate cross-domain intuition, lived perception or flashes of odd, immediate imagination in the way humans do. The recommendation is to capture those flashes in notes or voice memos, then use them as seeds for AI to test broadly, add detail and build fast prototypes.
The piece also treats questioning ability as a form of productivity. Deep questions, it says, come from insight into business essentials and an understanding of human needs. The authors encourage asking “why” and “what if” more often so that high-quality questions can shape the direction of AI outputs.
On top of that sits taste. The article says AI usually produces an average answer, one that often carries banality, moralizing or logical gaps. People therefore need enough aesthetic judgment to identify and remove those parts, then refine and rebuild the material so the final result carries an individual style and something the authors call a human soul.
The closing argument: stronger AI makes human judgment more necessary
In its closing section, the article says the real danger in the AI era may not be that machines become smarter than people, but that people gradually give up independent thinking. If the public follows AI instructions blindly and one system affects tens of millions or even hundreds of millions of people, the authors warn that even a very small rate of systematic error could be amplified at scale and influence large numbers of decisions and perceptions.
The article singles out the generation it calls those born in the 2020s. It says one reason humans can still control AI today is that engineers at the level of inventors are still alive, and those people understand the lowest-level code and the core bugs themselves. If AI systems later move toward programming and optimizing themselves, the authors say, the number of such engineers could keep shrinking.
It then widens the lens to civilization and infrastructure. As civilizations advance, systems become more complex. As systems become more complex, they rely more heavily on infrastructure and key nodes. The article contrasts older forms of knowledge storage such as stone, bamboo slips and paper, which could preserve information for hundreds or thousands of years, with today’s reliance on servers, cloud computing and networked storage. The gain is efficiency. The tradeoff is that part of civilization’s memory is now entrusted to digital infrastructure.
The authors describe that tension as a side of technological progress worth watching closely: greater capability can mean deeper dependency. They compare it to a kind of AI-era “taiji,” where gaining one power may involve losing another ability. One example in the article imagines a large-scale failure of network infrastructure. In that case, knowledge may still exist, but the ability to reach it could suddenly disappear. The authors compare that feeling to the way many people instantly feel penniless when they lose access to electricity and their phones.
The article closes with a final question. If AI can search, analyze, write and code for you, do you still have the ability to observe life, the courage to define the problem and the taste to judge what is good? In the authors’ view, the core competitive barriers of the future will no longer be how much knowledge someone stores, but physical perception, independent thought, sharp judgment, cross-domain inspiration and the ultimate decision-making power to define the problem. AI may keep producing answers faster, but the question of what kind of world people want should, for a long time yet, remain a human one.
The article ends with a disclaimer stating that markets carry risk and investment requires caution. It says the piece does not constitute investment advice, and users should consider whether any opinion, view or conclusion in the article fits their specific circumstances. Responsibility for investing on that basis lies with the individual.