Silicon Valley CEOs have recently reached a romantic consensus: as long as there is enough computing power and the models are strong enough, humans are a redundant variable.

Standing in earnings calls, industry summits, and media interviews, they repeat the same narrative over and over: AI reshapes everything, efficiency liberates organizations, and the human baggage of the old era can be cleared with one click. From Altman talking about AI involvement in parenting, to Mayer imagining couples' AIs coordinating household affairs with each other, to Zuckerberg asserting that the public will gradually accept deep AI involvement in life, the undertone of this rhetoric is highly consistent. Technology itself can solve all organizational, process, and cost problems; humans are merely redundancy waiting to be optimized.

But the capital market used October's market action to pour a basin of cold water on this.

I. In October, the Market First Cooled AI Fever

In the October just past, an abnormal signal appeared in the macro market. Jeffrey Rosenberg, a senior portfolio manager at BlackRock, pointed out that the strong correlation between crude oil prices and US Treasury yields, which had lasted for a full year at as high as 80%, suddenly broke.

Before that, the trading logic was simple and clear: oil prices push up inflation -> the Fed maintains high rates -> Treasury yields rise, with geopolitical conflict as the main pricing theme. But in October the wind changed. Rosenberg's judgment was striking enough: in the market's eyes, geopolitical risk has given way to AI risk.

The collective pullback of the Mag 7 and Google's market value evaporating by nearly 500 billion in a single week became the landmark event of this round of expectation reversal. Capital once assumed that as long as money kept being spent on computing power and large models kept being stacked, AI would continuously deliver growth; in the third-quarter US GDP data, investment in information processing equipment surged 37%, with a large amount of capital expenditure riding the AI story upward.

But after the earnings season began, the market started pricing AI in tiers. Some companies delivered verifiable returns: Santander Bank's AI projects landed in the first three quarters, creating 194 million euros in business value; Novartis and Standard Chartered successively said that AI had penetrated R&D, marketing, and customer operations, raising operating leverage; Meta's AI advertising increment and Google Cloud's doubled revenue were all solid positive cases.

For another group, however, only narrative remained. The capital expenditure of hyperscale cloud vendors has outrun their own free cash flow. For the first time in nearly a decade, Microsoft issued 25 billion dollars in bonds in the market to increase AI investment. The bond market was willing to buy in the short term, with oversubscription, but CDS credit spreads continued to rise. Investors began to hedge against a reality: with an estimated nearly 4 trillion in AI-related financing demand over the next four years, the bond market's capacity to absorb it is not unlimited.

Capital no longer blindly believes that "AI = perpetual growth." It is slowly understanding one thing: there are two kinds of AI. One is embedded in business and collaborates with people to create incremental value; the other is merely wrapping paper used by management to cut costs, lay people off, and beautify quarterly reports. The latter is the "AI risk" the market is pricing.

II. Behind Glamorous AI Agents, Shadow Labor Is Hidden

In major launch events and promotional materials, AI agents are already omnipotent: helping you call to negotiate bills, book on-site services, compare prices for purchases, and arrange home renovations. OpenAI and Meta have successively entered the personal intelligent agent track, depicting a future in which machines handle affairs autonomously and liberate humanity.

But peel away the promotional shell, and a different set of operating rules exists inside the industry. Dazzle AI, the AI agent company founded by Marissa Mayer, does not shy away from this model externally: the large model is responsible for retrieving information and providing a candidate list. Once a task enters the real-world implementation stage - calling to communicate, booking construction, confirming service details - common-sense errors easily occur. They once encountered a typical failure: the AI agent confused window glass with dining table glass in a roof renovation project and tried to install tabletop glass into a window. When encountering such bottlenecks, the system does not loop endlessly and consume massive tokens; instead, it directly transfers the order to contracted outsourced personnel, and a real person takes over to complete the communication.

Agent startups such as Duckbill and Wajo AI simply turn "human-machine handoff" into a standardized API product. They connect with a group of background-checked contract workers specifically to handle tasks AI cannot manage: phone communication, offline appointments, and website human verification. Founder Megan Joyce said bluntly that this logic is much like the early Uber model: AI handles order display, and human labor handles backstop delivery.

Meta's Muse agent also once sparked internal testing controversy: it publicly promoted autonomous AI calls, while during internal testing a large number of calls were made by real people at call centers. Meta later explained that this was only early product testing, but this detail has already punctured the industry consensus: at this stage, the vast majority of AI that can "handle real-world affairs" is essentially a front-end interaction skin, with real-person contractors as the underlying foundation.

The black humor lies here: many CEOs take cases of "autonomous AI agents" to report to the board, arguing that positions can be replaced and teams can be optimized, yet deliberately omit changes in the cost structure - formal employees are laid off, but labor does not disappear; stable, well-paid full-time positions simply become outsourced shadow labor that can be shut down at any time and has no social insurance protection. Labor costs do not disappear; they are transferred and hidden, turning from conspicuous salary items in financial reports into scattered purchases of external services.

III. The Engineer Who Cut 200 Million in Costs to 2 Million

These past two days, a screenshot has circulated repeatedly on Maimai and various workplace communities, triggering empathy from hundreds of thousands of workers.

The story has many versions: at a major internet company, a 20-person technical team maintained a high-consumption system, and the servers burned 200 million yuan a year. This state lasted for a full five years. One engineer in the team, hoping to solve the problem, worked overtime for months to refactor the underlying architecture. After launch, annual costs were directly compressed to 2 million. He had expected a bonus, promotion, and project commendation, but what finally came was a layoff notice for the entire team, with his own name also on the list.

The story circulating online has no confirmed company entity, so strictly speaking it belongs to a workplace fable spread in communities. But the reason it struck a chord is not the exaggerated numbers, but that countless practitioners have seen isomorphic realities: workers calculate contributions, while bosses only calculate headcount.

There is a real case overseas with almost the same core. A US e-commerce developer, outside his main job, used his spare time to develop automation scripts on his own, saving the company thousands of labor hours each year. He did not receive a raise or performance reward; instead, management demanded all code permissions. Soon after, HR interviewed and hired a newcomer at a lower hourly rate; after obtaining the code, the company quickly revoked the developer's permissions, unilaterally terminated the labor contract, offered only two weeks' salary as severance, and attached a strict non-compete agreement.

There is also a repeatedly mentioned ironic loop: an algorithm engineer personally built the internal system used by the company for personnel inventory, workforce efficiency evaluation, and assisted layoffs. This algorithm identified low-efficiency positions and output optimization lists. After running for a period of time, the list produced by the algorithm included its creator. He was optimized out, took off his work uniform, and went to run instant delivery, where he continued to be managed, assessed, and rushed by another platform algorithm.

This exactly corresponds to the "empty city strategy" interpretation circulating in the workplace: Sima Yi had long seen through the empty city as a trap, but he would not enter the city and kill Zhuge Liang. If the opponent disappears, the value of the Sima family in counterbalancing the Cao family also disappears. Translated into corporate organizations: as long as the problem exists, the corresponding budget, headcount, and positions have a basis for continued existence. Once the problem is thoroughly solved once and for all, the team that existed attached to this problem will immediately become redundant costs on the report that need to be cleared.

It is not that employees having excess capability is a crime; it is that this short-term financial-report-oriented assessment naturally punishes those who "solve problems once and for all."

IV. AI Is Becoming the Best Excuse for "Clearing People Out"

This round of layoffs is fundamentally different from previous rounds. In earlier layoffs, the reasons were business contraction, track adjustment, and financing downturn; now, the unified rhetoric of many management teams is: AI can take over the work, so the positions are no longer necessary.

The emergence of AI has provided the most fashionable narrative cloak for "cost reduction and efficiency improvement." Management of listed companies faces quarterly financial report pressure, and labor is the most easily adjustable variable cost. Cut the formal team, hand the work to large-model APIs and outsourced labor, and the current-period workforce efficiency metrics immediately look better, allowing them to hand Wall Street a satisfactory answer. Meanwhile, the hidden assets accumulated by the team over the long term - cross-business trust relationships, boundary experience gained from stepping into pits, and risk backstop capability for complex abnormal scenarios - cannot be entered into the income statement, so naturally they are not within the scope of short-term decision-making.

Here it is necessary to distinguish between two completely different approaches to AI implementation: The first is human-machine collaboration: AI takes on standardized, repetitive, low-judgment assembly-line tasks, freeing employees' energy for solution negotiation, client communication, risk decisions, and innovative R&D. The implementations at Santander and Novartis follow this route; AI is an amplifier for employees. The second is the replacement fantasy: the CEO presupposes that AI can directly replace positions, treats AI as a tool to replace employees, lays people off first, and then hopes the model will take over the original business. A large number of failure cases come from this category.

Many companies step into the same pit: after hastily laying off key business staff, they discover that the model can only handle standard inputs. Once confronted with abnormal requests, multi-party wrangling, special business agreements, or legacy gray processes, AI gets stuck across the board. Business cannot stop, so the solution is to quietly purchase outsourcing, part-time labor, and shadow labor to fill the gap. When total costs are finally calculated, there is no significant decline; stable full-time employment has simply been replaced by a flexible labor pool with extremely weak risk resistance. The organization's fault-tolerance cushion has been pulled out.

The demand origin of many AI products comes from the life pain points of the CEO circle itself - booking flights, arranging dinners, handling chores, and planning family schedules. Marissa Mayer imagines couples' AIs coordinating household affairs with each other, Altman openly discusses AI-assisted parenting, and some imagine AI replacing humans in date screening. These are niche pain points of a small top-level group, not the real scenarios of the vast majority of business positions. Executives are addicted to the utopia that "all interactions can be optimized and automated," underestimating the irreplaceability of people in real business as communication hubs and subjects of risk judgment.

V. When Working Seriously Becomes a High-Risk Activity

The more fatal damage than a single layoff is the irreversible turn in organizational culture. When the people who "thoroughly solve problems" are ultimately optimized out; when the reward for delivering perfect results is team disbandment, a set of self-protection survival rules will spontaneously grow inside the organization. It is not the choice of any one employee, but a collective adaptation forced by the system's reward and punishment mechanism.

Everyone begins to learn to split results: a project that could be closed in one go is split into research, pilot, iteration, and phase-two optimization, delivered and reported in stages. Projects are always "in progress," never "completed." Documentation only goes as far as "basic work can be handed over"; core trade-off logic, hidden pitfalls, and boundary judgments are not archived and delivered all at once. Cross-department collaboration tries to make oneself a process node rather than an isolated screw, so that upstream and downstream business links are all bound to oneself. Many veteran employees no longer fully disclose their experience, newcomers can only access surface-level work, and the chain of core experience transfer slowly breaks. The short-term workforce report looks good, but in the long term, the probability of business gaps, sudden failures, and rework accidents continues to rise, which will backfire on the company at higher cost in the future.

The ten workplace self-protection rules currently making the rounds online have gone viral precisely because they amount to an open letter of grievance from employees to management. This self-protection logic repeatedly drives home a few basics: don't delete the database, don't leave malicious backdoors, and keep every operation above board; keep the pace controllable, leave an evidence trail, keep outside options open, understand your non-compete agreement, and don't treat the company as your home. This isn't about encouraging opportunism or slacking off—it's the bare minimum of self-protection for ordinary people in an environment where "being serious means losing out."

We've also seen plenty of cautionary tales. The Weimob operations staffer who wiped the database and the algorithm engineer who deleted training models both ended up with criminal convictions. Destroying the system is not a bargaining chip—it's a one-way ticket to judicial proceedings. That road absolutely does not work. The real root of the conflict has never been that employees want to protect themselves, but that organizations have failed to design mechanisms that carry the value of those who contributed. The people who solve problems bear all the risk, while the results of eliminating problems instead become grounds for layoffs; verbal praise is everywhere, while long-term incentives, job security, and value backstops are all absent.

VI. Capital Has Woken Up; Some CEOs Are Still Dreaming

The October market pullback was essentially a recalibration of expectations. Capital has seen the real incremental value AI brings—and has also seen through the bubble of pure narrative. Microsoft, Google, and Meta continue to issue bonds and ramp up capital investment, showing that the long-term AI track has not been disproven; but widening CDS spreads and the shift of funds from AI targets to long-end Treasuries for hedging signal that the market no longer buys the simple story that "adopt AI and growth is guaranteed."

What AI can do—and where its boundaries lie—is actually already clear: it processes text, summarizes information, generates first drafts, standardizes statistics, and screens in batches, with efficiency far beyond human capability. But it cannot navigate complex interest games, cannot bear business responsibility, cannot make trade-offs amid ambiguous information, and cannot rebuild trust in a demoralized team. Management diseases, organizational diseases, distorted short-term assessment, departmental silos, and misaligned authority and responsibility—these underlying corporate chronic diseases cannot be cured by computing power and models.

A company with chaotic management, wavering goals, and departments blaming each other will not suddenly become efficient and orderly just because it purchases a large model. It will only displace the original contradictions: people are laid off, the business can't be sustained, and outsourcing fills the gap; outsourcing proves unstable, service quality declines, and customers churn. In the end, management may even reach an absurd conclusion: AI isn't strong enough—so keep pouring on computing power and keep squeezing manpower.

This is the most ironic loop of the present moment. Some CEOs are addicted to a minimalist fantasy: relying on AI to bypass troublesome organizational governance, lengthy team run-in, and complex talent development—to shed employees and rely solely on computing power to change the enterprise, and even change the world. They underestimate two things: first, at this stage AI cannot do without human backstopping; second, the foundation of an enterprise is always collaborative people, not model parameters.

AI is a great efficiency tool, but not a miracle cure for layoffs. Healthy digital transformation is about letting AI lift people up; pathological AI fervor is about using AI as an excuse to abandon people. Capital has already started recalculating the numbers, and workers saw through this script long ago. The remaining suspense is left to those CEOs still lost in fantasy: when shadow headcount keeps growing, everyone on the team is just protecting themselves, and hidden business risks keep piling up—what will they use to catch the backlash when the pretty financial reports propped up by layoffs and AI stories come crashing down?