Sam Altman has unleashed his ultimate move.
Early Friday morning, OpenAI officially released its latest generation of large model, GPT-5.5.
"Losing GPT-5.5 is like having a limb amputated," said an Nvidia engineer who participated in the internal testing.
On the same day, Sam Altman posted an email sent to him by Jensen Huang on social media, along with an internal memo from Nvidia.
Jensen Huang has mandated that all Nvidia employees, including those in legal, HR, and finance departments, use AI agents. This is an order, not a suggestion.
This article will tell you three things that Jensen Huang mentioned in his email that are relevant to you:
AI is an organ, not a crutch; legal and finance departments can use AI more and need it more than you think; embracing AI should be a requirement, not a suggestion.
I hope today's content has been inspiring for you.
I. GPT-5.5: From "Answering Questions" to "Completing the Work"
On one occasion, Dan Shipper, the founder of the AI writing platform Every, conducted an experiment.
After his app was launched, it encountered a persistent bug, so he hired a top engineer to refactor it.
The engineers spent several days and finally came up with a solution.
Then give the buggy code to the model and see if it can make the same decision as the engineer independently.
GPT-5.4 couldn't do it, but GPT-5.5 could.
Shipper said this was the first time he had ever felt true conceptual clarity in a programming model.
It's not about responding to a question, but about understanding the problem and figuring out how to solve it yourself.
MagicPath CEO Pietro Schirano had a similar experience.
One day, he used GPT-5.5 to merge a branch containing hundreds of frontend changes and refactoring changes with the main branch.
It only took 20 minutes to resolve the issue in one go, with almost no rework.
He said, " I really feel like I'm working with a higher intelligence." The key is not that it gets everything right the first time, but that it's easier to stay on the right track without having to frequently pull back along the way.
Behind these cases lies a fundamental change.
When OpenAI released GPT-5.5, its official positioning was only one sentence: "A new kind of intelligence for practical work and intelligent agents."
The key point of this statement is not intelligence, but work.
Over the past two years, the upgrade path for large models has been clear: stronger reasoning, longer context, and higher accuracy.
GPT-5.5 is still doing these things, but the focus has changed.
OpenAI emphasizes that models understand tasks earlier, rely less on hints, are better at using tools, and can keep moving forward until completion.
The goal of GPT-5.5 is no longer to answer correctly, but to complete the task.
From answering a question to completing a task, there's a difference of two words, but also a difference of an era.
The focus of the discussion has also shifted. From whether the answer is accurate, it has become how many times it needs to be revised and whether it can be run successfully on the first try.
Of course, this process is far from over, and multiple third-party reviews have mentioned that GPT-5.5 is more dependent on task boundaries.
If the requirements are not clearly described, it will not automatically fill in the blanks for you, but will execute based on the existing information.
But this precisely shows that it is becoming more like a collaborator in the real world.
2. Why did Huang Renxun send an email to all employees?
On the day GPT-5.5 was released, Sam Altman posted an email written to him by Jensen Huang on social media.
The email included an internal memo from Nvidia, the core message of which was simple: Jensen Huang required all Nvidia employees to use Codex (an AI programming assistant launched by OpenAI).
It covers all departments, without exception, including engineering, product, legal, marketing, finance, sales, human resources, operations, and development projects.
The core of the email, translated, can be summed up in one sentence:
"Codex isn't just for software teams; everyone should use Codex agents. They're our teammates, the superpowers that allow us to surpass ourselves. Better, smarter, faster."
Jensen Huang also revealed that NVIDIA and OpenAI have collaborated to build the Codex Lab, specifically to help employees get started. A series of online workshops will also be held in the coming weeks.
The email ended with the sentence: "Let's step into the era of light speed together. Welcome to the age of artificial intelligence."
As CEO, Huang sent an email to all employees requesting the use of the latest AI tools; this was not only a promotion but also a mobilization order.
Many people ask: What does legal review of contracts and HR's handling of recruitment have to do with writing code?
It's a very important matter.
Huang Renxun dared to let legal and HR departments use it because GPT-5.5 is essentially a work tool, not a programming tool.
Its core capability is to autonomously complete complex multi-step tasks.
You give it a vague instruction, and it plans the steps, calls the tools, checks the results, and keeps moving forward until the task is completed. You don't need to teach it step by step in this process.
It can recognize screen content on a computer, click buttons, fill out forms, and transfer information across tools.
These operations require legal review of contracts, HR screening of resumes, and financial reconciliation.
OpenAI's internal data confirms this judgment:
More than 85% of employees use Codex across departments weekly;
The public relations department used GPT-5.5 to analyze six months of speech invitation data and build a scoring framework;
The finance department reviewed 24,771 K-1 tax forms, totaling 71,637 pages, two weeks ahead of last year; the marketing team saved 5 to 10 hours per week.
Huang Renxun had already foreshadowed this in a speech on April 18th. He said:
"AI is a power multiplier, not a replacement tool. The key to determining your career destiny lies in your willingness to use it, not in AI itself."
Six days later, he put these words into action in an email to all employees.
Jensen Huang's logic is clear: using AI is your opportunity, not using AI is your loss. But if you're at Nvidia, you won't even be given the choice.
Nvidia's role is also noteworthy. In this GPT-5.5 release, Nvidia was not just a customer, but a co-designer.
GPT-5.5 and NVIDIA GB200, GB300 NVL72 systems were jointly designed.
You see, from training to deployment, the model and hardware have been working in tandem since their inception. NVIDIA provides the hardware, OpenAI provides the model, and everyone uses it.
This is a supplier relationship, but more importantly, it's a strategic alliance.
III. Three sentences related to you are hidden in Huang Renxun's email.
In fact, Huang Renxun's email only mentioned three things.
1. AI is an organ, not a crutch.
The Nvidia engineer's statement that "losing GPT-5.5 is like having a limb amputated" is not an exaggeration.
A cane is something you're comfortable using, but you can walk without it. An organ, on the other hand, stops functioning when removed and only survives when it's put back in.
Has any employee in your company ever said this?
If not, it means your company is still giving AI a try, but not that it can't live without it.
What's the difference?
The difference lies in whether your employees treat AI as a secretary or a colleague.
Only when they become colleagues will they become indispensable. This is the first step for your company to become an "AI-native organization".
2. Legal and finance departments can utilize AI more effectively and need it more than you think.
Huang Renxun included legal and HR departments in the all-staff email because GPT-5.5 capabilities already cover non-programming roles.
Legal contract review is essentially a multi-step information processing process; financial reconciliation is essentially the execution of rules across multiple tables; HR resume screening is essentially condition matching and batch processing.
These are things that AI excels at, and Nvidia has already proven them for you.
3. Huang Renxun made a request, not a suggestion.
You must understand this signal.
In a company, what becomes a requirement? It's not that doing it is better, it's that not doing it will lead to failure.
When Huang Renxun sent an email to all employees saying that everyone must use it, he wasn't promoting a benefit, but rather announcing a rule.
From this moment forward, AI proficiency is a basic requirement for Nvidia employees, not a bonus.
What about your company? Are you still suggesting that everyone learn it when they have time, or have you already decided who must use it, how to use it, and to what extent?
Please think about this carefully.
Conclusion
In the 1990s, Microsoft mandated the use of Office (Microsoft's office software suite) throughout the company.
In the 2010s, Google mandated the use of Google Docs (a web-based online document editing tool launched by Google) throughout the company.
In 2026, Jensen Huang required all Nvidia employees to use Codex.
Every time the entire company is forced to use it, it represents a fundamental shift in the way we work.
Nvidia is declaring: the AI era is not the future, it's the present.
Welcome to the age of artificial intelligence.
The question is, are you and your team ready?
This article is from the WeChat public account "Notesman" (ID: Notesman) , author: Lao Jia, and published with authorization from 36Kr.
