In December, OpenAI announced a huge improvement in the code for o3.

Some CS graduates feel that their major has been wasted:

CS graduates who find the right job may still be happy, after all, o3 is only open to security researchers, has not entered the market, and the cost performance may be relatively low.

But this time the Wolf really came!

Winter in the US workplace

Before OpenAI announced its new AI model o3, cloud computing giant Salesforce announced that it will no longer hire software engineers in 2025.

Salesforce employees are already feeling the chill in 2024 - after all, the company has been known to lay off workers. In January of that year, Salesforce laid off 700 employees - roughly 1% of its 70,000 global workforce.

It's heartbreaking for many of those affected, but it's still a drop in the bucket compared to the year before.

In January 2023, the cloud computing giant laid off 7,000 people - about 10 percent of the entire company. At the time, CEO Marc Benioff wrote a letter to employees saying it was due to "excessive hiring" during the coronavirus pandemic.

The Saleforce layoffs are just a microcosm of the layoffs in the United States last year.

After mass Layoffs in 2022 and 2023, more than 150,000 jobs were cut globally in 2024, involving 545 companies, according to Layer.fyi.

Large companies such as Tesla, Amazon, Google, Snap, and Microsoft made massive layoffs in 2024.

Some smaller startups have also experienced layoffs, and some have shut down altogether.

Layoffs will continue in 2025, with major U.S. companies that have already identified layoffs including Microsoft, BlackRock and Ally.

25 years of US technology layoffs, Microsoft began

Since 2022, the companies with the most layoffs are Intel and Tesla. The two companies cut 15,000 and 14,000 jobs, respectively, last year, making them the highest number of employers in recent years.

While companies' reasons for cutting jobs vary, these cost-cutting moves are closely tied to the backdrop of technological change.

According to a recent report by the World Economic Forum (WEF), some 41% of companies worldwide are expected to reduce their workforce in the next five years due to the rise of AI.

Previously, Dropbox, Google, IBM and other companies have announced AI-related layoffs.

The past year has passed, and at the beginning of the New Year, an even bigger round of tech layoffs is surging.

Microsoft

First of all, Microsoft.

BI exclusively reports that a Microsoft spokesperson has confirmed that the company will soon start to lay off employees in 2025, the specific number of layoffs is not disclosed.

This time, however, Microsoft is shifting its focus to underperforming employees.

There are said to be job cuts across the company, particularly in the security department.

"Microsoft is focused on high performers. We are always helping people learn and grow, and if they continue to underperform, we will take appropriate action," a spokesperson said.

For those who were sent away because of poor performance, Microsoft said it would also recruit new employees to fill the vacant positions.

Therefore, the total number of employees in the company will not change much. At the end of June, Microsoft had 228,000 full-time employees.

Like its rivals, Microsoft is now taking a harder line on performance management, people familiar with the matter said.

For the past few months, the company's managers have been evaluating employees, even those as high as level 80, one of the highest ranks in the company.

In fact, as early as the end of last year, anonymous social platform Blind has been circulating a variety of gossip.

Some netizens said that Microsoft is about to carry out large-scale layoffs on January 15, 2025, and this proportion or as high as 5%.

Aqua Security

Aqua Security is a company focused on cloud-native security protection platform solutions.

As part of a strategic restructuring, the company plans to lay off dozens of people in global markets to improve profitability.

Ally

A spokesperson for Ally, a digital finance company, confirmed that the company is laying off about 500 employees (out of 11,000 employees) and has been sending out emails to employees on Tuesday.

"As we continue to resize the company, we have made the difficult decision to selectively reduce headcount in some areas while continuing to hire in other areas of the business," the spokesperson said.

Ally will provide severance pay, job placement support, and offer employees the opportunity to apply for open positions within the company.

Altruist

Los Angeles-based Fintech company Altruist plans to cut 37 jobs, affecting about 10% of its workforce.

In addition, according to Trueup, 2,977 people have been affected by layoffs at 16 tech companies. Among them, it listed the number of Microsoft layoffs at 2,280.

2,977, and that's just the beginning of the 2025 layoffs.

AI agents cause 'unemployment wave'?

In the next two to four years, Internet users predict that AI agents could lead to a Depression-level wave of unemployment.

How the hell did this happen?

One in five white-collar workers has a job so simple that "powerful AI" could quickly replace them:

Step 1: The company installs monitoring tools such as keylogger on employees' computers.

Step 2: The AI observes and learns from the employee's work behavior.

Step 3: AI replaces humans (note that AI can already operate computers, clicking and typing with ease).

"But what about those special cases?" you may ask.

For some jobs, AI may not be able to do 100 percent of the tasks, but it can do 80 percent, which means companies can fire 80 percent of their employees.

Typically, technology adoption takes decades because it takes a long time to reshape supply chains.

But there's no need here - the AI agent simply learns by looking at an employee's screen.

Maybe we can find new jobs for the people we've replaced, but maybe we can't?

Because unlike in previous economic revolutions, other jobs they can do may also be simple enough to be replaced by AI agents.

Note that AI continues to get exponentially smarter.

Historically, horse-drawn carriages have been replaced by cars, and this time human jobs may be replaced by AI.

Saleforce recruits and promotes AI products

Salesforce CEO and founder Marc Benioff revealed that Saleforce decided to freeze the hiring of software engineers in 2025 because of the significant productivity gains brought by AI technology.

And because of AI agents, Salesforce will also reduce after-sales engineer jobs.

When asked if Salesforce will have more or fewer employees five years from now, he said the company "could have more."

But he further explained: "We're not hiring any more software engineers next year because we've increased productivity by more than 30 percent this year with Agentforce and other AI technologies for engineering teams - engineering development is incredibly fast."

"In addition, next year we will reduce the number of support engineers because we have agents. We will add more sales people next year because we need to explain to people in detail the value we can achieve with AI. So in the short term, we may add 1,000 to 2,000 sales people."

Salesforce is already applying AI agents to customer support.

Experienced workers can provide clear and detailed instructions and make effective use of agents.

But novices may make poor choices, and AI agents become technical debt instead.

Can AI replace programmers?

Saleforce's approach seems to say that agents can't replace salespeople and customer service, but they can replace software engineers.

According to the World Economic Forum's latest survey, 86 percent of respondents believe that the macro trends driving business transformation include AI and information processing technologies.

But the truth is not simple.

Lazy AI's CEO doesn't think current AI can replace software engineers, and gives six reasons why.

1) Intelligent style programming = false propaganda

50% of AI-generated code is worthless.

The reason: Every code change has a 50% chance of introducing some kind of bug or unexpected behavior, which can be caused by the prompt word.

If you make five changes to a large Language model (LLM) at once, when it's done, you can introduce multiple bugs without even knowing how to trigger them.

2) Universal coding tools are lies

When a company claims to have built an AI product with no framework, no application type, and just send a prompt to let the AI do all the work, you should be skeptical.

Try it, and you'll see that the AI sometimes writes Python in JS files, confusing dependencies and mixing up comment syntax.

Most companies can't solve this kind of problem.

3) Authentication is your responsibility

In the AI-coding tool, run the following prompt: "Create a Google login page and have it jump to my profile page."

Broken OAuth processes, redirected URI errors, links to documents, and time-consuming API keys in Google Cloud Console...

Thousands of users stormed out.

4) LLM can't see your database

Almost every project needs a database. However, when building a project, the database is always changing.

The following cycle is common:

• Change the database, and the AI doesn't know what to do.
• Change code, database error.
• Code migration is required, and the AI has written a buggy migration script.
• And repeat.

5) Pay now, never succeed

You may have experienced the following:

When you ask the AI to build new features, it removes parts of the project. The AI copied parts of the code. When you explicitly ask it to do X, it does Y. The AI is stuck fixing the bugs it introduced.

It's not enough for an AI model to "get better" when you're paying for it.

One misplaced parenthesis can ruin the entire page. Whenever a new AI model is released, it takes weeks to test it with real-world applications.

6) The project itself is difficult.

As for building complex projects with AI, it's wishful thinking.