Karpathy Deletes Anthropic Profile, Suspected to Have Left After 68 Days?

By: Henry, QuantumBit

Damn! Woke up one morning, is Karpathy also going to star in "Inside Anthropic"?

Just now, careful netizens discovered that the great Andrej Karpathy quietly modified his X bio, and the work information related to Anthropic has disappeared.

Thus, a bold speculation began to ferment:

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

This matter quickly exploded on X.

What's more, some netizens posted so-called "insider information," claiming Karpathy had submitted his resignation application a few hours ago and calmly and politely thanked Anthropic for the job opportunity.

Others stated that Karpathy's Anthropic email address also seems to have stopped working, further corroborating the resignation rumors.

However, this information currently remains at the level of netizen leaks and screenshots. Neither Karpathy nor Anthropic has yet made an official response.

It is worth mentioning that on Wikipedia, Karpathy's personal information still displays as "Anthropic (2026 to Present)".

So, whether Karpathy is plotting "Inside Anthropic" or Anthropic will come out to refute the rumors first, we'll just wait and see.

By the way, regarding Karpathy's writing style, those who know, know~

Too Ashamed to Be an Anthropic Employee

To be honest, the reason netizens quickly interpreted "deleting bio" as resignation is not only because Karpathy himself is too much in the spotlight, but also related to this subtle timing.

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

The open letter was initially co-signed by 25 companies and institutions including NVIDIA, Microsoft, and Meta; the list initially lacked the big three of large models simultaneously: OpenAI, Google, and Anthropic.

However, Altman quickly reposted the open letter.

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

As of now, among several head model companies, Anthropic is still absent.

So netizens began to let their imaginations run wild: Is it possible that Karpathy, as a famous open-source contributor, really couldn't hold back and hid his employer from his bio in a fit of anger?

Some even joked: "He might just be ashamed of his employer."

And what came later was even more outrageous... Facing external doubts, Anthropic employees personally entered the fray to push back against the wind:

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

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

Regarding this, Teacher Andrew Ng also made a reasonable point:

Of course, every company has the right not to open-source its own code; the problem is, you cannot choose to be closed-source while simultaneously trying to prevent others from open-sourcing.

For this Anthropic employee, the comment section was naturally full of criticism. It has to be said, deleting the Anthropic bio at such a critical juncture, no wonder netizens would joke:

Karpathy might really be too ashamed to be an Anthropic employee.

Of course, this is obviously just making memes. Between Karpathy deleting his bio and Anthropic not signing the open letter, there is currently no proven causal relationship.

But looking back, these 68 days after Karpathy joined Anthropic indeed did not pass as peacefully as imagined.

68 Days Inside Anthropic

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

His task was also very Karpathy-style: assemble a new research team to utilize Claude itself to accelerate pre-training research.

Simply put, it is letting AI help research the next generation of AI, moving towards more automated model development. Karpathy stated at the time that the next few years would be a critical stage for large model development, and he looked forward to returning to frontline R&D.

This news could be called shocking to the entire internet at the time.

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

The narrative of "Karpathy Defects to Anthropic" spread rapidly; his official announcement post has currently received over 27.7 million views.

But after joining Anthropic, Karpathy quickly fell into several public opinion controversies.

The first occurred during the release of Claude Fable 5.

Fable 5 officially launched on June 9, but within a few days, it was disabled for various reasons, then ups and downs, permissions turned on and off, we won't go into details here, those who know, know.

This twist of launch, takedown, and relaunch, of course, cannot be blamed on Karpathy, but it made the first Anthropic flagship model release after Karpathy joined seem somewhat like a bad start.

The second time was Claude Tag.

On June 23, Anthropic launched Claude Tag, putting Claude directly into Slack. Users just need to @Claude like @colleagues, and it can read channel context, break down tasks, call tools, and work continuously in team collaboration.

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

From the earliest web chat boxes, to Claude Code-style personal agents, to the current Claude Tag capable of entering team collaboration environments.

But in the eyes of many netizens who have followed Karpathy for years, this expression was somewhat exaggerated.

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

Sharper criticism was:

In the past, Karpathy was always teaching everyone how to implement neural networks, Transformers, and GPT from scratch, letting everyone see clearly what was happening inside the technology; but now, he is endorsing a closed-source product that does not disclose model weights, training data, and implementation details.

"You used to teach us build from scratch, but now you let us blind trust vendor?"

This wave of controversy also made many netizens start missing the Karpathy of the past who wrote nanoGPT, micrograd, and llm.c, leading everyone to hand-craft models from scratch.

Further back, there was this "Open Weight Alliance" initiated by Jensen Huang.

When Microsoft, Meta, NVIDIA, IBM, and even OpenAI which joined later all began to publicly endorse open models, Anthropic still did not appear on the list.

This time, Karpathy did not publicly express his attitude, but quietly deleted the Anthropic information in his bio.

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

What's more interesting is that as early as when Karpathy announced joining Anthropic, some netizens confidently predicted:

"He will leave within 6 months."

At the time, it looked like just a rash statement thrown out casually. Unexpectedly, only two months later, this sentence had already been dug out by netizens in advance to "manifest" it.

Why Karpathy of All People?

Of course, having said that, modifying a social media bio was originally not big news.

But placed on Karpathy, it is different.

As everyone knows, Karpathy is not only one of the most influential researchers and opinion leaders in the AI era, but also one of the few people who simultaneously span top-tier research, industrial practice, open-source communities, and public education.

He studied under Fei-Fei Li at Stanford University, mainly researching the intersection of computer vision and natural language processing during his PhD, and participated in designing and lecturing Stanford's famous deep learning course CS231n.

After graduating with a PhD, he became one of the founding members of OpenAI. In 2017, Karpathy joined Tesla as AI Head, leading the team responsible for Autopilot's computer vision, data labeling, neural network training and deployment, etc., until leaving in 2022.

In 2023, when ChatGPT ignited the large model wave, he returned to OpenAI again, assembling a new team responsible for midtraining and synthetic data generation. In 2024, he left OpenAI again and founded the AI education project Eureka Labs.

But compared to these resumes, Karpathy's truly special place may lie in his another identity:

AI Educator.

From Stanford CS231n, to Neural Networks: Zero to Hero on YouTube, to a series of minimalist open-source projects like micrograd, nanoGPT, llm.c, Karpathy has been trying to dismantle complex AI technology, flatten it, and rebuild it with as little code as possible.

These projects were not necessarily for direct industrial production, but allowed countless developers to truly understand for the first time what backpropagation, Transformers, and large model training are all about.

Until today, I still feel that "Deep Dive into LLMs like ChatGPT" released by Karpathy more than a year ago is currently the best introductory tutorial for large models, bar none.

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

Everyone is accustomed to viewing him as a representative figure of open technology, minimalist implementation, and inclusive education.

Therefore, after he entered a highly closed-source model company, his praise for every product, his silence on every industry controversy, even a missing line in his personal bio, will be taken out and repeatedly interpreted.

So, the truly interesting part of this turmoil may not just be whether Karpathy actually left Anthropic.

It is more like a tug-of-war surrounding Karpathy's public image:

On one side is entering frontier labs, re-participating in top-tier large model R&D; on the other side is the open-source educator in netizens' hearts who insists on build from scratch and can't wait to break down every technical detail to explain clearly.

As for this time, whether it was just Karpathy casually changing a bio, or he is really preparing to leave Anthropic, there is currently no answer.

But what is certain is that as long as the protagonist is still Karpathy, the internet will not lack drama. No sooner said than done. Just now, some netizens asked if Karpathy resigned? Karpathy stated: Surprisingly, such inexplicable fake news is spreading on Twitter. No, I didn't resign.  If I really resign, it definitely won't be in the form of changing a personal bio, I would at least write a 10-paragraph essay.