Even if Meta lays off 90% of its staff, apps like Instagram and Facebook will continue to operate normally.
Eva is a senior engineer at Meta, not on the layoff list, has strong performance, and is proactively adopting AI tools.
But he said, "No one is safe; it's all dangerous—it's just a matter of time."
This is a story about how performance is evaluated, how promotions occur, how management operates, and even how hard work itself is defined—those caught in it, from Zuckerberg down to the newest junior engineer, cannot say when the storm will end.
The layoffs are real, but the reasons are false.
Meta has laid off approximately 25,000 people since 2022.
In November 2022, 11,000 employees were laid off, followed by another 10,000 in 2023, which Zuckerberg called the year of efficiency. In January 2025, Zuckerberg announced in an internal memo the elimination of the bottom 5% of performers, approximately 3,600 people. In March 2026, an additional 700 employees were laid off. According to Reuters, another 8,000 employees are set to be cut by late May, representing about 10% of the global workforce of nearly 79,000, with a second round planned for the second half of the year.
Layoffs are genuinely happening, but not necessarily because AI has taken these people's jobs.
Eva believes that most people laid off at this stage would have left regardless of AI. "A few years ago, the entire CS industry hired far more people than it actually needed—industry prosperity, capital overheating, stock prices rising steadily, and many companies hired large numbers of employees. After Musk acquired Twitter and laid off most of the staff, the app still worked fine—back then, there was no AI at all."
In 2026, Meta’s capital expenditure guidance is $115 billion to $135 billion, nearly double that of 2025, all directed toward data centers, GPUs, and AI infrastructure. The savings from workforce reductions have been redirected toward computing power.
At this stage, AI serves as a respectable facade, allowing companies to claim that efficiency has improved and fewer employees are needed.
Small companies are nimble and agile, but as they grow into large enterprises, decision-making slows down, and they find themselves unable to compete with emerging unicorns and startups—so they begin to streamline operations, flatten structures, and focus on core products. AI is merely accelerating a cycle that was already underway.
When AI usage is included in performance evaluations
However, the involvement of AI has still changed some of the rules surrounding layoffs.
Meta’s original performance evaluation method was quite unique among major Silicon Valley tech companies. Managers did not assign direct scores; instead, they compiled a performance rating document based on your self-assessment, peer feedback, and their own observations.
Then, proceed to a Calibration Meeting, where roughly a dozen peers are grouped together; each manager takes turns presenting their team members' performance and explaining why each individual deserves a particular rating. The group discusses collectively, and finally assigns ratings to everyone.
This process is lengthy and cumbersome, but its value lies in incorporating multiple perspectives and peer-level comparisons, making it difficult for any single manager’s preferences to determine the outcome. Eva considers this relatively fair.
In early 2026, the Calibration Meeting was canceled. Eva explained, “The company reverted to biannual performance reviews, as AI enables managers to use AI-assisted tools for self-evaluations, reducing the need for extensive collaboration and accelerating the process.”
Meanwhile, Meta has launched Checkpoint, an AI performance tracking system that automatically aggregates employee work data from internal systems such as Google Workspace to generate contribution summaries for managers. For software engineers, Checkpoint tracks over 200 data dimensions, including the percentage of AI-generated code, while monitoring metrics such as error rates and the number of associated bugs.
Meta’s Chief Human Resources Officer, Janelle Gale, stated in an internal memo at the end of 2025 that AI collaboration skills will become a core criterion for performance evaluations in 2026.
Additionally, for every segment of code written by Meta engineers, the system automatically assigns a percentage indicating the proportion of the code assisted by AI, and this data has become part of the performance evaluation criteria.
Each team sets its own minimum threshold based on its circumstances—for example, requiring 50% or 90% of the code to be generated by AI. You must meet this threshold, but performance evaluations will still assess how much actual value your work delivers. “The company’s idea is that you start using it first, and we’ll see how well you do later,” Eva said.
Incorporating AI usage into performance metrics acts as a form of mandatory promotion—rewarding those who use it frequently is not the goal, but penalizing those who don’t use it is.
This approach is not unique to Meta.
NVIDIA CEO Jensen Huang publicly stated at the GTC conference in March 2026 that, in the future, every engineer at the company will need an annual token budget, with an additional half of their base salary allocated specifically for AI consumption. He even said that if an engineer earning $500,000 per year spends less than $250,000 on AI annually, he would be “deeply concerned.”
Jensen Huang is selling tokens—why wouldn’t a merchant promote their own product? But Meta once also reached the extreme of this quantitative frenzy.
An employee spontaneously created an internal leaderboard called "Claudeonomics," named after Anthropic's Claude model, tracking AI token consumption among 85,000 employees. Within 30 days, the entire company consumed over 60 trillion tokens.
The leaderboard features badge tiers ranging from Bronze to Emerald, and the top 250 participants earn titles such as Token Legend and Cache Wizard. The top-ranked employee consumed 281 billion tokens within 30 days; some employees manipulated the rankings by running AI agents in idle mode for hours, consuming tokens without performing any actual tasks. Measuring productivity by token consumption is like evaluating a truck driver by fuel usage—just because the engine is running doesn’t mean deliveries are being made.
Eva didn’t feel any pressure from the leaderboard within her team: “Anyway, we don’t have any direct connection to this leaderboard—we just keep doing our work and glanced at it for fun.” Her manager didn’t use it as a talking point, but even after the leaderboard website was taken down, the underlying logic remained. The percentage of AI-generated code was still being tracked, and the minimum threshold still existed.
And as everyone is pushed to use AI, and everyone’s output increases, the performance standards themselves will rise along with it. “If 60% of people are performing better, the standard will inevitably go up. But it’s hard to say how much of that improvement comes from AI versus how much comes from working late and burning out.”
The wind of intense competition has reached Silicon Valley
Eva’s top boss is also under pressure: “All the other senior leaders are pushing their teams relentlessly—if he doesn’t succeed in doing the same, his position is at risk.”
According to The Wall Street Journal, Meta has newly established an AI engineering department with a manager-to-engineer ratio of 1:50—one manager overseeing 50 engineers, double the traditional Silicon Valley upper limit of 1:25.
Gallup data shows that the average number of direct reports per manager in the U.S. rose from 10.9 in 2024 to 12.1 in 2025, but Meta’s 50:1 ratio is still more than four times the industry average.
Eva personally experienced this change. In a typical large company, a manager oversees dozens of people, helping with career planning, having one-on-one conversations, and understanding your needs.
1:50 means a team that previously had five managers now only needs one, leaving four without jobs.
No one knows how this new department will operate, although outside voices believe the change will end in tragedy.
Other departments are still maintaining their original management rhythm, and managers will still have one-on-one career planning discussions with you. However, everyone expects this state won’t last much longer—some teams have already eliminated junior managers and now have only the next level of leadership directly oversee everyone.
Management itself is also questioning whether their work has become meaningless. “Everyone is in the same situation, facing the question of whether their role is still necessary. This applies to leaders as well—their days haven’t become any easier.”
AI is indeed helping managers improve efficiency by automatically summarizing what their team members have recently coded, posted, or attended in meetings, and generating regular reports. Previously, managers had to spend time tracking this down themselves; now, they simply review the AI’s summaries.
On the other hand, increased efficiency makes management cheaper, and cheap things never lack alternatives.
The pressure of intense competition trickles down, and those at the bottom—entry-level positions—bear the most direct impact.
As a senior engineer, Eva used to assign minor bugs to junior engineers during project planning. Now, if the issue is small, he simply opens an AI window and resolves it in just a few minutes. “No need to communicate with junior engineers—I take care of it myself in no time.”
Large projects still require human effort, but the mundane tasks that once made up the bulk of junior engineers' workloads are now being effortlessly handled by AI at the fingertips of senior engineers.
Eva speaks quickly: “If you can do as early as possible—being both an engineering manager, a product manager, an engineer, and a designer—all in one person, so you can build a feature or even an entire team by yourself, your chances of being laid off might be slightly lower.”
Regarding how many people will ultimately remain, Eva smiled and said, “At this moment, even if Meta retains only half of its staff, it can still keep running. If AI continues to develop at the pace advertised, eventually only about 10% of programmers might be left to review what the AI has produced and align product decisions—while the remaining 90% could be laid off. Even in that scenario, Meta would still be able to keep going.”
No one is safe, not even Zuckerberg.
No one feels safe.
Senior leaders feel pressure because other senior leaders are competing fiercely; managers feel pressure because their span of control may increase from 1:15 to 1:50; senior engineers feel pressure as standards continue to rise; junior engineers feel pressure because their tasks are increasingly being absorbed by AI tools used by senior engineers.
Even Zuckerberg himself is experiencing anxiety.
The uncertainty of the AI era is real; every new feature released by Claude Code could put a company out of business, Figma’s stock price surged dramatically after the Claude Design announcement, and the entire SaaS industry is being dismantled piece by piece.
Social networks may seem to have barriers, but those barriers are never as thick as they appear. Eva believes the transition from QQ to WeChat took just one or two years.
Zuckerberg, while concerned about the company’s future, is aggressively cutting jobs. To Eva, this is a management strategy: “He wants to keep the most driven and the most intelligent employees. What’s the best way to do that? He found that offering money isn’t the most effective approach—layoffs work better.”
Creating a sense of insecurity drives output more than issuing bonuses.
But this strategy comes at a cost. Top engineers won’t tolerate this pressure indefinitely—they’ll leave for places that value their contributions more. Layoffs may drive away underperformers, but they may also push out those with the most options.
Eva chose to stay because the situation is more practical—although Silicon Valley has become more competitive lately, it’s still not as intense as back home.
However, behind these individual choices, the industry's overall trend can no longer be ignored: "AI will replace most jobs, and the internet industry can no longer return to its former glory of earning substantial income without being overly busy."
If you can't beat them, join them.
AI has reshaped how existing employees work and has also transformed the entry points for new hires.
Meta’s engineering interviews traditionally consist of three parts: Coding, Behavioral Questions, and System Design. Coding involves solving an algorithmic problem, such as sorting a dataset, testing your choice of algorithm and your considerations for performance and cost. Behavioral questions are subjective, focusing on how you handle feedback and conflict. System Design, typically reserved for senior-level candidates, involves architectural design challenges.
In October 2025, Meta introduced an AI coding segment into its interview process. What was previously two rounds of pure coding has now become one round of traditional coding and one round of AI coding. Candidates are given a multi-file, complex project in a CoderPad environment, with an AI chat window on the right side that allows them to switch between multiple AI models during the interview, including the GPT series, Claude series, Gemini, and Llama. Within 60 minutes, you must understand a codebase you’ve never seen before, break down the problem, and use AI to implement features or fix bugs.
It’s not about whether you can write code or craft prompts—it’s about your judgment in collaborating with AI. The AI’s output may be correct, incorrect, or partially correct; what matters is how you interact with the AI to achieve satisfactory results, and whether you can determine if the generated code is optimal. The interviewer will observe every prompt and interaction in real time.
Eva believes this closely resembles a real work environment, observing whether candidates can use the latest tools to solve complex problems in a short time.
The new entry standards mean that anyone entering this field from day one is expected to possess the ability to collaborate with AI. A candidate who went through this round of interviews summarized in their reflection that AI did not make the interview easier; instead, it raised the bar. With AI assistance, interviewers expect you to solve more complex problems within the same time frame.
Faced with this situation, Eva chose the strategy of joining forces rather than fighting against it.
If this is the prevailing trend, you can't change it—resisting AI is pointless.
Eva’s daily workflow has completely changed—she now opens multiple AI windows simultaneously to have them handle different tasks in parallel. “You only have one brain and can only do one thing at a time. But the advantage of AI is that you can run ten of them to handle different tasks for you.”
It took about a month to go from trying it out to getting comfortable with it.
He uses AI across nearly every aspect of his work: writing documents and brainstorming during project planning, comparing solutions, writing SQL queries to estimate potential impacts, and coding. After completing a feature, he also uses it to write summaries and post social media updates to increase visibility.
Be among the first to master AI, and you might become one of the last to be laid off. But how fast layoffs will happen, or whether you’ll truly avoid them, no one knows—so make the best of it.
Beyond this self-reassurance, AI offers vastly different value to people at different levels.
For senior engineers who have accumulated sufficient experience and can identify issues and grasp direction, AI is a tangible lever—what used to be a daunting two-week analysis can now be started immediately. But for those early in their careers, AI removes precisely the part of thinking and trial-and-error they most need to develop.
Efficiency has improved, but learning opportunities have disappeared.
Eva is unwilling to classify herself as either optimistic or pessimistic: “You can’t change this big trend—just like the workers laid off in Northeast China back then; you can only accept it. Some opened restaurants, others headed south to start businesses. Who knows? Life is too long—it’s pointless to overthink it.”
So far in this game, the only certainty is that no one is a winner.