The interesting part of this incident may not just be whether Karpathy has left Anthropic, but rather a tug-of-war around Karpathy's public image.

Written by: Henry, Quantum Bit

What a surprise! Waking up, is Karpathy about to play "Inside A"?

Just recently, observant netizens noticed that the great Andrej Karpathy quietly modified his X personal profile, and the work information related to Anthropic has disappeared.

As a result, a bold speculation quickly began to ferment:

Having joined Anthropic for just over two months, Karpathy may have already resigned.

This news quickly exploded on X.

What’s more, some netizens shared what they called "internal news," claiming that Karpathy submitted his resignation a few hours ago, and calmly and politely thanked Anthropic for the job opportunity.

Others indicated that Karpathy's Anthropic email also seems to have stopped being used, further corroborating the resignation rumors.

However, all this information remains at the level of netizen rumors and screenshots for now. Neither Karpathy nor Anthropic has made a formal response to this.

It’s worth mentioning that on Wikipedia, Karpathy's personal information still shows "Anthropic (2026 to present)."

So, whether Karpathy is brewing "Inside A" or if Anthropic comes out to dispel the rumors, we will just have to wait and see.

By the way, everyone knows about Karpathy's writing style～

No face to continue as an A family member

To be fair, the reason netizens quickly interpreted "deleted profile" as resignation, besides Karpathy being under too much scrutiny, is also related to this subtle timing.

Just two days ago, Jensen Huang posted his first post after joining X: a joint open letter supporting open-weight AI models.

The open letter was initially signed by 25 companies and institutions including NVIDIA, Microsoft, and Meta, but notably lacked the big three in large models: OpenAI, Google, and Anthropic.

However, Ultraman quickly retweeted the open letter.

Subsequently, OpenAI and Google were also added to the list of signatories.

As of now, Anthropic is still absent among the leading model companies.

Thus, netizens began to let their imaginations run wild: could it be that the well-known open-source contributor Karpathy was unable to hold back and, in a fit of anger, hid his workplace from his profile?

Some even joked, "He may just be embarrassed by his employer."

And then came something even more outrageous… In the face of external doubts, Anthropic employees also personally took to the frontlines:

Why doesn't NVIDIA open-source CUDA and GPU drivers?

Why doesn't Microsoft open-source Windows and Office?

In response, Andrew Ng also offered a candid remark:

Every company certainly has the right not to open-source their code, but the problem is you cannot choose to keep your code closed and then try to stop others from open-sourcing.

As for this Anthropic employee, the comment section was naturally filled with criticism. It’s no wonder that at such a crucial moment, deleting the Anthropic profile led netizens to tease:

Karpathy may really feel ashamed to be an A family member.

Of course, this is clearly just a joke. There is currently no confirmed causal relationship between Karpathy deleting his profile and Anthropic not signing the open letter.

But looking back, the 68 days since Karpathy joined Anthropic have indeed not been as peaceful as one would imagine.

68 days inside A

On May 19, Karpathy officially announced his joining of Anthropic as a technical employee, entering the pre-training team led by Nick Joseph.

His task was also very much in line with Karpathy's style: to build a new research team, leveraging Claude itself to accelerate pre-training research.

In simple terms, it was to let AI help research the next generation of AI, moving towards more automated model development. Karpathy stated at the time that the coming years would be a critical phase for the development of large models, and he was looking forward to returning to frontline research.

This news was shocking throughout the internet at that time.

After all, on one side was a founding member of OpenAI and former head of AI at Tesla, while on the other was one of OpenAI's most direct competitors.

The narrative of "Karpathy defecting to Anthropic" quickly spread, and his official announcement post has currently garnered over 27.7 million views.

However, after joining Anthropic, Karpathy soon found himself embroiled in several controversies.

The first occurred during the release of Claude Fable 5.

Fable 5 officially went live on June 9 but was disabled a few days later for various reasons, then experienced ups and downs with permissions being toggle on and off; we won't elaborate here, those in the know will understand.

This series of ups and downs surrounding its launch can't be blamed on Karpathy, but it made the release of Anthropic's first flagship model after his joining somewhat less than successful.

The second controversy was about Claude Tag.

On June 23, Anthropic launched Claude Tag, integrating Claude directly into Slack. Users just need to @Claude like they would a colleague, and it can read the channel's context, break down tasks, call tools, and continuously work within team collaboration.

Karpathy rated this product highly, calling it the "third significant redesign" of large model interaction:

From the earliest web chat box, to Claude Code-style personal agents, and now to Claude Tag that can enter team collaboration environments.

However, for many netizens who have followed Karpathy for years, this expression seemed somewhat exaggerated.

Some felt that a Claude integrated into Slack seems unable to bear the title of "third paradigm shift"; others began to joke that after joining Anthropic, Karpathy had transformed from an educator into a company man.

Even sharper criticisms included:

The past Karpathy always taught everyone how to implement neural networks, Transformers, and GPT from scratch, making sure everyone understood what was happening inside the technology; but now, he was standing up for a closed-source product that does not disclose model weights, training data, and implementation details.

"You used to teach us to build from scratch; now you want us to blind trust the vendor?"

This wave of controversy also led many netizens to reminisce about the previous Karpathy, who wrote nanoGPT, micrograd, and llm.c, guiding everyone to build models from scratch.

Then came Huang's initiated "Open Weights Alliance."

When Microsoft, Meta, NVIDIA, IBM, and even the later joined OpenAI started publicly supporting open models, Anthropic was still absent from the list.

This time, Karpathy did not publicly express his stance but quietly deleted the Anthropic information from his profile.

The timing was simply too coincidental, and netizens naturally began to connect the dots.

Interestingly, even when Karpathy announced his joining Anthropic, some netizens confidently predicted:

"He will leave within six months."

At the time, it seemed like just a throwaway remark. Who would have thought that just two months later, this prediction would be brought back up by netizens as "evidence"?

Why is it precisely Karpathy?

Of course, to put it in perspective, modifying a social media profile isn't a big news story.

But when it comes to Karpathy, it is a different story.

Everyone knows that Karpathy is not only one of the most influential researchers and opinion leaders in the AI era but also one of the few individuals who simultaneously spans top-notch research, industrial practice, open-source community, and public education.

He studied under Fei-Fei Li at Stanford University, primarily researching the intersection of computer vision and natural language processing during his doctoral studies, and participated in designing and lecturing the renowned Stanford deep learning course CS231n.

After earning his doctorate, he became a founding member of OpenAI. In 2017, Karpathy joined Tesla as the head of AI, leading a team responsible for Autopilot's computer vision, data labeling, neural network training, and deployment until he left in 2022.

In 2023, as ChatGPT ignited the wave of large models, he returned to OpenAI to form a new team responsible for mid-training and synthetic data generation. In 2024, he left OpenAI again to create an AI education project called Eureka Labs.

However, compared to these credentials, what truly sets Karpathy apart is perhaps his other identity:

An AI educator.

From Stanford CS231n, to YouTube's Neural Networks: Zero to Hero, to a series of ultra-simple open-source projects like micrograd, nanoGPT, and llm.c, Karpathy has been trying to break down and flatten complex AI technologies, and then rebuild them with as little code as possible.

These projects may not be aimed at direct industrial application, but they have allowed countless developers to truly understand how backpropagation, Transformers, and large model training work for the first time.

To this day, I still believe that Karpathy's "Deep Dive into LLMs like ChatGPT," published over a year ago, is the best introductory tutorial on large models, bar none.

It is precisely because of this that many people’s expectations of Karpathy are no longer as simple as "just a company employee."

People are used to viewing him as a representative of open technology, minimal implementations, and inclusive education.

Therefore, when he enters a highly closed model company, every praise for products, every silence on industry controversies, and even the missing line from his profile will be subjected to repeated interpretation.

Thus, the most interesting aspect of this incident may not just be whether Karpathy has truly left Anthropic.

It seems more like a tug-of-war around Karpathy's public image:

On one hand, entering a cutting-edge lab to re-engage in top-tier large model development; on the other, the image netizens have of him as a stubborn advocate for building from scratch who wants to clarify every technical detail.

As for whether this time was just Karpathy casually changing his profile or if he is genuinely preparing to leave Anthropic, there is still no answer.

But one thing is certain: as long as the protagonist is still Karpathy, the internet will never lack drama. Just as it happens, just now, a netizen asked if Karpathy had resigned? Karpathy stated: "I can’t believe such absurd fake news is spreading on Twitter. No, I haven’t resigned. If I were to resign, I certainly wouldn’t do it by changing a personal profile; I’d at least write a 10-paragraph essay."