Uncertainty

Viewpoint: AI product forms are immature and highly uncertain; using amoeba-style small teams (each with front-end and data members) to experiment in parallel makes it easier to produce results.

Logic chain: Small teams have shorter decision chains and less organizational friction, enabling rapid experimentation; running many small teams in parallel increases the probability of success.

Failure condition: If the business model requires heavy assets or strong collaboration, small teams lack sufficient resources.

Related areas: Management & teams, organizational design, innovation mechanisms

Viewpoint: In economics and finance, while LLMs can compute explicit models such as option pricing, pricing risk-taking and the market intuition required to place bets under uncertainty become ever scarcer.

Logic chain: Models rely on statistical correlations and historical data. When structural breaks occur—such as the difference between a liquidity crisis and the collapse of fundamentals—only human causal intuition and the courage to bear consequences can drive decisions. This kind of intuition is difficult to algorithmize.

Failure conditions: If markets become fully rule-based, free of black swan events, or if all risks can be perfectly hedged through complete contracts, then such intuition would have no application.

Related fields: Decision-making & cognition, business models

Claim: A prompt functions by carving out a subset within an LLM's trillion-scale conditional probability distribution, reducing uncertainty and enabling ordered output—analogous to Maxwell's demon in thermodynamics.

Logical chain: From the perspective of Shannon entropy, a prompt acts as a conditioning operation that constrains the direction of generation within a vast probability space. This requires humans to expend cognitive energy in crafting the prompt, thereby filtering meaningful sequences out of seemingly chaotic token probabilities.

Failure conditions: If the prompt is poorly designed and fails to effectively restrict the probability space, or if the model temperature is too high and the entropy reduction fails, the output will still drift toward randomness.

Related field: AI applications

Opinion: Planning is not about eliminating uncertainty; rather, it is about increasing the probability of winning under the premise of acknowledging randomness, by setting boundaries, preparing resources, and maintaining flexibility.

Logic chain: The future is unpredictable, but risk exposure can be managed. Good planning is like "adjusting your stance" when rolling dice: it includes spreading your bets, setting stop-losses, and building feedback mechanisms for learning.

Failure conditions: When the environment is completely chaotic and no identifiable patterns exist, no planning can improve the odds of winning. In that case, survival and remaining flexible take first priority.

Related fields: Decision-making and cognition; management and teams.

Viewpoint: Planning is not about predicting the future; it's about improving the odds of the dice roll—constructing a better probability structure within uncertainty.

Logic chain: The future is inherently random, and completely eliminating uncertainty is unrealistic. Good planning uses scenario planning, flexible resources, and rapid feedback loops to keep decisions oriented amid change.

Failure conditions: In an environment of absolute chaos or completely inaccessible information, any planning can become an illusion; when execution breaks down, even the best plan is useless.

Related fields: Decision-making and cognition, management and teams, entrepreneurship.

Viewpoint — Life holds countless possible directions, but we can never truly know which of them will come to pass. The story quietly fades to an ending, suggesting that possibilities exist yet remain unknowable.

Logical chain — The two characters in the story never develop a deep connection, symbolizing that the chance encounters between people in reality do not necessarily lead to outcomes. Possibilities linger like ghosts, but due to incomplete information and the one-way arrow of time, they cannot be foreseen.

Conditions for failure — If one could attain an omniscient perspective or possess the power of prophecy, the fog of possibility would be dispelled; but in reality, this is unattainable.

Related fields — Decision psychology, cognitive philosophy, literary narrative.

Viewpoint: Coworker-style AI, as represented by Crayfish, can autonomously complete complex tasks, but its output is highly uncertain and consumes a large number of tokens, making it unsuitable for serious work and severely wasting users' attention.

Logical chain: Such products may complete a task today, yet fail tomorrow for unknown reasons; errors are diverse and hard to prevent. In the face of frequent failures and corrections, users' attention is constantly interrupted, and overall efficiency declines. For work that demands deterministic output, the risk is unacceptable.

Invalidation conditions: Uncertainty is acceptable in creative, game-like, or exploratory tasks with high error tolerance. When the technical team encapsulates uncertainty internally (e.g., via a management platform) to provide stable services, this form can be adopted on a limited basis.

Related fields: AI applications (agents and automation), workplace development (tool selection and efficiency).

Viewpoint: In everyday AI use, it is better to fall back to weaker tools or methods with higher certainty, ensuring output is directly usable, and never accept uncertain results that require repeated adjustment—because protecting your own attention is far more valuable than unleashing AI's full potential.

Logical Chain: Attention is a scarcer resource than AI compute → If AI output is uncertain, you must debug, retry, and correct, consuming large amounts of attention → Conversely, as long as you limit the AI's task scope and use high-certainty tools, you can generate usable results in one pass, requiring only waiting and acceptance → The whole process does not interrupt the flow of thought, minimizing attention waste.

Failure Condition: When the task itself demands extremely high creativity, complex reasoning, or maximum output, over-pursuing certainty can limit results. In such cases, you may briefly use more expensive, higher-capability models (e.g., Claude Opus) in specific scenarios, but only invest in them at the few necessary moments.

Related Areas: AI tool usage strategy, personal productivity, attention management.

[View] The core of entrepreneurship is to keep believing in the direction you have chosen when faced with uncertainty, unpredictable prospects, and widespread doubt at the level of common sense—and ultimately prove that truth is in the hands of the few. Leading brands need a small vanguard to tear open a rift in the era, and then subsequent forces can scale that breakthrough into an industry.

[Logic chain] Common sense often represents the view of the majority, but real innovation and breakthroughs tend to defy common sense. If founders choose to follow, they will not be questioned; but to lead, they must endure loneliness and the courage to be disliked. Steve Jobs was ousted from the company he founded, yet he did not change his vision—an example showing that great achievements are often accompanied by doubt from the team, customers, or even the whole world. In the end, only after truth has been validated by the few can it be amplified through large-scale strategies.

[Failure conditions] If the direction the founder insists on is itself wrong, or if the direction is right but the timing is too early and conditions are not yet in place, or if the founder cannot gain recognition and support before resources run out, then this persistence will lead to failure. Moreover, persistence is not rigidity; it requires fine-tuning based on practical feedback, or it can turn into stubbornness.

[Related fields] Entrepreneurship, entrepreneurial spirit, innovation, decision-making and cognition

Viewpoint: When a company or individual voluntarily abandons a proven profit model (the comfort zone) and steps into a completely unknown new field, they will inevitably experience extreme uncertainty, self-doubt, and external skepticism. This is the price that must be paid to become a leading brand. Safe strategies are only suitable for scaling mature businesses, not for pioneering innovation.

Logic Chain: When things are going well, imagination can overshadow capability, and certainty about the future brings pleasure. But once you proactively choose a new direction, certainty is lost—even if the old business remains tempting, there is no turning back. At that point, the team, customers, and even you yourself will harbor doubts. Only by withstanding this pressure and ultimately delivering results can you become an industry pioneer. Followers can adopt safe strategies (imitation, chasing revenue over brand) to scale quickly, but they cannot become leaders.

Failure Conditions: If the new field is chosen incorrectly, or if the company exhausts its resources and the team falls apart during the struggle, this ordeal may not bear fruit. Additionally, if the founder is merely stubborn and fails to adapt flexibly based on market feedback, they may die from persistence rather than breakthrough.

Related Fields: Entrepreneurship, Strategy, Brand Management, Innovation

Viewpoint: High-priced products should be positioned around "helping clients make critical decisions — which way the gun points, left or right" — rather than competing on execution capability.

Logic chain: Executives and owners typically have their own teams with formidable execution; what they lack is not execution but judgment. A subtle difference in judgment — such as a strategic direction off by five centimeters — can swing costs by hundreds of millions. Your value lies in providing perspective (seeing how the leading players play), correcting course in real time, and reducing uncertainty. Once the anchor client endorses your judgment map, that "eye-dotting" touch on the dragon's head sets the long execution budget behind it in motion — two ounces of force moving a thousand pounds.

Failure condition: When clients can access a broader perspective themselves, when new technology sharply reduces uncertainty, or when execution itself becomes the core barrier, this high-price logic weakens.

Related fields: Business models, decision-making, and cognition

Viewpoint: Entrepreneurship is different from being an employee: there is no teacher to highlight key points or provide definitive answers. Success is a probabilistic outcome after repeated failures, and uncertainty must be accepted.

Logic chain: The education system conditions people to expect certainty, but in the business environment there are only probabilities. Entrepreneurs often need to sustain substantial losses before they understand the logic of making money; failure is the norm. Pursuing 100% certainty leads to inaction or giving up too early.

Failure conditions: If an individual is extremely well-capitalized, they can afford large-scale trial and error with a lower probability of success; however, most ordinary people still need to accept probability and control the cost of each trial.

Related domains: Entrepreneurship, decision-making, and cognition.

【Viewpoint】You must dare to step out of your comfort zone and embrace uncertainty.

【Logic chain】Staying inside your comfort zone limits the possibilities you can explore. Only by stepping outward can you encounter interesting opportunities; even if a new attempt fails, the process itself can become an experience worth being proud of. The key is to manage your energy while advancing multiple workstreams at the same time.

【Failure conditions】If you lack basic risk tolerance and energy management skills, blindly jumping out of your comfort zone may damage your primary work or lead to a series of setbacks; or if the uncertainty in your environment is excessively negative (for example, there is no safety net), caution is needed.

【Related fields】Personal growth, entrepreneurship, side hustles.

Viewpoint: The more “SOUL”-type personalized content is incorporated into an AI model, the greater the uncertainty of its output.

Logic chain: Personalized elements increase the randomness and unpredictability of model behavior, which even relatively advanced models such as Kimi2.5 cannot avoid.

Invalidation condition: If the model architecture can achieve an effective balance between personalization and determinism.

Related field: AI applications.

Viewpoint: Individuals need to retain the unknown, uncertainty, and creative space in their work in order to feel fulfilled, while certainty-based corporate operations suppress this need, producing misalignment and distress.

Logic chain: Managers derive deep motivation from creative challenges, but organizations, for the sake of efficiency and control, strive to eliminate uncertainty and rely on non-creative processes. When executives themselves favor creativity while organizational mechanisms resist the unknown, both sides experience discomfort.

Failure conditions: In R&D organizations or early-stage ventures, creativity is already central to the work itself, so this conflict is weaker.

Related domains: Career development, management, and teams.

Viewpoint: The virality of content is highly uncertain. Even for the same creator, audience preferences are hard to pin down, and past performance cannot accurately predict which piece of content will become a hit.

Logic chain: Observed video views can exceed 600,000, yet the patterns of what users like or dislike are entirely uncontrollable. This indicates that traffic distribution and user preferences on content platforms involve a great deal of randomness: carefully produced content may not receive high traffic, while casually posted content may unexpectedly go viral.

Conditions under which this fails: When creators build detailed user personas, master the platform's algorithm rules, and conduct rigorous A/B testing, content predictability improves. If platform algorithms become transparent or shift toward strong editorial recommendations, traffic uncertainty also decreases.

Related fields: Content creation, marketing, and traffic.

Viewpoint: In repetitive creative work, deliberately leaving room for the uncertainty created by “inaccuracy” is an effective way to sustain interest and surprise.

Logic chain: Long-term mechanical repetition of the same type of creative work erodes enthusiasm → deliberately introducing inaccuracy breaks fixed expectations → uncertainty produces a sense of random beauty and freshness → each act of creation carries the surprise of exploration → creative drive and enthusiasm are maintained.

Failure conditions: When the work has strict precision requirements, such as technical drawings or compliance documents, inaccuracy leads to quality defects.

Related fields: Content creation, personal growth.

Viewpoint: For some highly uncertain markets, accepting their unpredictability and building adaptive strategies accordingly is wiser than spending heavily in pursuit of false certainty.

Logic chain: Complex systems are nonlinear and irreducible → traditional forecasting models tend to fail → consulting reports may offer only psychological comfort, or may simply conclude “unpredictable” → the large sums invested produce no real output. Rather than pursuing false precision, it is better to channel resources into improving an organization’s agility, robustness, and learning capacity.

Failure conditions: When identifiable and stable patterns exist in a market—such as seasonal demand for some consumer goods—forecasting remains useful. This judgment applies more to areas that are extremely uncertain and heavily affected by technological or policy shocks.

Related domains: Decision-making and cognition; business models.

Viewpoint: AI features may appear to be implemented in demos, but there is substantial uncertainty behind them. As long as this uncertainty exists, they are almost unusable in commercial scenarios that demand high intensity and high stability.

Logic chain: An AI demo looks usable → in real-world use, uncertainty is high and error rates are high (for example, Siri has remained limited to a small number of built-in scenarios for years, and speech recognition occasionally errs) → commercial scenarios require high reliability → uncertainty creates unacceptable risk → deployment cannot be achieved.

Where this does not apply: When an application scenario has high fault tolerance (such as creative assistance or non-critical customer service) and does not impose strict stability requirements, AI uncertainty can be accepted, and its application value can still be realized.

Related fields: AI products, voice assistants, commercial deployment, AI application reliability.

Opinion: In AI painting, the uncertainty that comes from using pure prompts and low-fidelity remix brings more creative joy and surprise than precise image-to-image generation.

Logic chain: Precise image-to-image generation emphasizes restoration and control, so the results are predictable; by contrast, pure prompts plus low-fidelity generation introduce uncontrollable factors and produce unexpected outputs → this unexpectedness brings the pleasure of exploration and discovery and stimulates the creator’s sense of surprise.

Failure conditions: When creators have clear commercial delivery needs and need to precisely reproduce a specific image, uncertainty becomes a negative factor, reducing efficiency and quality.

Related fields: AI applications, content creation