Yeng-yu was born in the temple of Zhengfei.
Nvidia is public.
As the Silicon Valley is so big, it didn't take long for Musk to complete the next section of the project.
The quantum bits are known.XAI's entire multimodal team, about 70 people, was recently sacked.
What's wrong with multimodalls?
However, almost at the same time, xAI's official recruitment immediately began with a new position as Model Engineer, Multimodal Understanding, based in Palo Alto, California.
JD shows that the position covers comprehension and production of images, video, audio, and text, ranging from data, tokenizer, pre-training, post-training alignment to infrastructure and product experience, with an annual salary range of $160K to $250K USD ($1.15 million to $1.8 million) plus equity and medical benefits.
△Images generated by AI
Think of Grok's multimodalling capabilities as a worldwide mouthpiece.
While horizontal fists are better than GPT and footsteps are better than Gemini, everyday photo questions are enough, and tied to X real-time data flows, low latency of voice interaction, short video generation with synchronized audio, and relatively loose auditing are ideal for social media creativity.
In addition to Agent's move towards operating software, understanding on-site and completing tasks is on the rise, and the demand for multimodalls is growing, so it's easy to think about how old horses can't do without this fat piece of meat.
It's a thunderstorm, it's a mask, it's too hot.
From Tesla to SpaceX, and then to Twitter's bloody beginnings, the kind of crap he felt wasn't good enough to build up and dismantle the crap of managing aesthetics finally came to xAI's head.
Musk has been playing for so many years, cutting off, cleaning up, re-inventing tricks, skillfully making people feel bad (bushi).
△Images generated by AI
From the founding executive to the R&D front line
Do not underestimate the intensity of the shaking within xAI this year!
Throughout 2026, xAI has undergone a comprehensive cleanup from management to core technology backbone.
After SpaceX announced its acquisition of xAI in February, the founding team began to dissolve at a noticeable rate. Co-founders Tony Wu, Jimmy Ba, and others left.
According to Reuters at the time, by mid-February 2026, about half of the 12 co-founders of xAI had left completely.
And then there's Musk.The company's organizational structure has been adjusted sharply.The four main business lines are: Grock Basic Models and Voice; Coding Models and Machine Learning Infrastructure; Imagine Multimodal Generation; and Macrohard Enterprise Process Automation.
The re-appointment of the heads for each line of business and greater management authority for some technical heads.
In March, Old Horse said in a statement on X:
The first time xAI was not built correctly, it is now being rebuilt from the ground up.
It was later fulfilled.
As the integration of SpaceX and xAI progressed, Musk began to draw direct line managers from SpaceX and Tesla, heavily involved in the day-to-day management and business review of xAI. The original management relationship and reporting system also changed, with the line managers reporting directly to Musk.
This severe blow directly resulted in a massive loss of talent.
According to The Information, by mid-May, all of xAI's original co-founders had left, with the exception of Musk.
Since SpaceX's acquisition of xAI, more than 50 researchers and engineers have left (including active layoffs, layoffs and layoffs).
The heads of coding, world models, Grok Voice, etc. have been stepping out, and even the underlying team responsible for core pre-training has reportedly suffered a sharp decline.
I didn't find it.XAI's round of cleaning goes from the initial level of founding executives to the core line of business research and development.
In November, the fire finally burned the multi-modal team.
After the demolition of the Masks, a number of departments were rebuilt.
It's a script that has been rebuilt, demolished and rebuilt over the years, and Musk doesn't know how many times he's done it.
As long as the pace is not right, the cost overruns or the technical route is not as expected, his first reaction is to directly remove the brake barrier from the physical sense.
Moderate optimization?
I've never heard of it!
△Images generated by AIHe has used the organization's adjustment method several times in different companies, but not much in the way of specific tools.
At times, when the project is not progressing as expected, he takes the lead and regains command; at other times, he simply disconnects the entire department; he encounters changes in the technical direction, and even years of R&D teams can be pushed back.
As for the companies that were acquired, that's all that matters.
We looked for a few more typical instances, and found that the old horse's operation was very much like a flower.
One word doesn't match, the whole team pulls out.
The core logic:Business must continue to advance rapidly, but when the leaders of the existing sector form objections and are unable to fully implement the will of the old man, even if the business is in an absolute leadership position in the industry, it is better to dismantle the entire power system and recruit people later.
A case in point is the Tesla Supercharger team incident in 2024.
The team, previously responsible for the construction and operation of Tesla's global supercharging network, has partnered with several automakers, including Ford, GM, to build the best recharging network in the U.S. at the time.
In April 2024, Musk and Rebecca Tinucci, the head of the overloading business, spoke out on the scale of the layoffs.Tinucci was fired at the meeting and decided to disband the entire 500-strong team.
Following a Reuters investigation, Tinucci had previously sacked 15 to 20 percent of his team members, and when the two sides reported the business plan to Musk, disagreements over further personnel cuts led to the result.
But business can't stop!
Tesla first handed over the charging network to the original energy sector, but the new team was not familiar with the construction and operation of the charging network, and some new projects were subsequently requested to be suspended, with suppliers and partners affected.
A few weeks later, Tesla began recruiting the supercharging team again, during which time it recruited some core members of the original supercharging team.In the United States, the number of charging stations in the North American charging market has risen dramatically in recent years.
2 The cost of remaining silent after the original route was overturned
Core logic: Once it is determined that the input output of a technology or product line is no longer established, no matter how many years of research and development have been invested, how many assets have been accumulated, even the physical evaporation of the project with software hardware research and development specialties.
Another scenario is that Musk makes new judgments about the original technological path, and then aligns the project with the team.
A typical example is the Tesla Dojo supercomputer team event in 2025.
Tesla has been developing Dojo since 2019 and hopes to provide computational support for autonomous driving model training through its self-developed chips and computing clusters.
After years of service with Tesla, Musk has repeatedly publicly stressed the importance of Dojo.
In July 2024, he also said that Tesla would increase its investment in Dojo.
In August 2025, the direction suddenly changed.
Following a change in the calibration pattern and the price of the self-developed chip, Musk immediately announced the dismantling of the Dojo team, ending the project, and concentrating all of the calibration on the AI6 chip and general computing architecture.
Musk subsequently described Dojo 2 as a dead end on the evolutionary ladder, arguing that it was no longer necessary to continue investing. Some employees switched to other computing projects and the original team was no longer retained.
XAI's AI Tutor experienced a similar bloodbath early last year.
Because xAI announced adjustments to training strategies, reduced general-purpose AI Tutors, shifted to professional AI Tutors, and plans to expand the size of the professional team by 10 times.
On the one hand, there has been a sharp reduction in the general training team, on the other hand, there has been an expansion in the demand for professional trainers, as well as a change in the structure of the original staff, which has resulted in the removal of at least 500 data labeling employees.
3 Companies rebranded
Core logic: Facing an established organization or system of mergers and acquisitions, through surgical layoffs, it directly breaks the original corporate culture, division of functions and compliance boundaries, pushing the remaining people into the hardcore model.
How could this not have been done by Timmagor after the acquisition of the original Twitter X?
Following the completion of the acquisition that week, core executives such as former CEO, CFO, general counsel were fired immediately, and Musk became Twitter's sole owner.
In less than a week, Twitter launched massive layoffs. The original machine learning ethics team, human rights team and content curator team were uprooted, and the external trust and security committee dissolved on the spot.
The acquisition was completed in less than a month, and nearly 3,700 people were laid off. About half of the company's employees lost their jobs overnight.
Musk also sent an internal email to the rest of the staff asking them to choose whether or not to join the Twitter 2.0 campaign he envisioned.
If you want to stay, you have to work longer hours and work harder; if you don't want to, you have to take three months' salary.
In the end, a large number of employees chose to leave, and some core engineering and infrastructure teams were also affected.
△Images generated by AIAs you can see,In Musk's management logic, organizational architecture always seems highly subordinate to current strategic objectives.
The number of achievements the team has achieved, the amount of experience it has accumulated, and even the number of mature management systems it has formed, is hardly a reason to keep them.
As long as business priorities change or existing teams do not comply with new requirements, management, personnel and resources may be reassigned.
When necessary, even the entire company can be pushed back.
Perhaps the best thing about the old horse is to keep a company in a state of restart forever.
