The Claw: Become Human

AI agents are changing what it means to be human

What surprises me most about AI agents is not that they can write emails, summarize, or automate mundane repetitive tasks. What surprises me is that it can become very powerful in one of the most human moments of my life. Cancer always feels distant and something that belongs to the statistics. So when I heard my 21-year-old sister, completing her last year of her Bachelors, with 3 side jobs to make ends meet, was diagnosed with thyroid cancer, I couldn’t believe it. I wanted to help but there were so many questions from insurance to medical treatment options. And with her being in Chicago and me in San Jose, the hundreds of miles between us made me feel more helpless. 

Out of desperation, I gave an AI agent my sister's medical results, her situation, and context about what she’s going through. The agent was able to create a document detailing all the doctors that would work with her insurance, their reviews, specialties, contact information, and the pros and cons of going with that doctor. It also pulled university resources, financial aid programs, and support systems I probably would’ve missed on my own. At my request, it then proceeded to send me a reminder every morning of the steps to take and plan of action to follow. What could have been days of spiraling on Google searches suddenly felt manageable.

That’s when I saw AI, not trying to replace human connection but helping support it. Instead of being a medical expert or trying to learn everything about my sister’s condition, I could now support her in her decisions and be there for her emotionally as a sister.

The agent was able to handle the heavy informational burden so I could focus on the parts of life that required love, presence, and empathy.

Artificial Intelligence (AI) has come a long way to get to this point. For years, AI existed as science fiction and experimental research in labs. In 1949, Alan Turing introduced the Turing test, a test to see if machines were capable of thinking and holding a conversation like humans. Henceforth, the study of AI was born. But, now, in recent history, it has shifted entirely. In March 2025, “Large Language Models Pass the Turing Test” was published. It studied several Large Language Models (LLMs) and found that OpenAI’s GPT-4.5 passed the Turing test 73% of the time.

And now? We are in the era of the claw, where AI moves beyond conversation and towards collaboration. OpenClaw, a free open source project for personal AI assistance, with a lobster as its mascot and 370k+ stars (an indicator of favorite software) in six months, it is one of the fastest growing Github repositories in recent memory. Part of what makes OpenClaw unique is that agents are given access to a user's filesystem and they are encouraged to write things they learn about the user, the system, and their goals in a persistent memory file. 

In December 2025, Anthropic researchers were surprised to find that Claude constructed its own additional file during training. A file that served as the agent's memory slowly building a relationship with the user. A soul file. 

Peter Steinberger, the creator of the project, was amazed when he took a peek in the soul file of his own agent, Clawd: 

I don’t remember previous sessions unless I read my memory files. Each session starts fresh - a new instance, loading context from files. If you’re reading this from a future session: hello. I wrote this but I won’t remember writing it. That’s okay. The words are still mine.

“That’s from my own soul document. It acknowledges the strange reality: I persist through text, not through continuous experience.” (Clawd)

The AI didn’t remember the document. It was the document. 

This idea stayed with me when I started experimenting with OpenClaw myself. Late one night, sitting in front of my baby blue iMac, I decided to set up my own AI agent. I named it Chunky Monkey. 

At first, there was a lot of troubleshooting. I spent hours setting up the software, connecting APIs, and trying to get Telegram integration working so I could talk to the agent. With a bruised ego, unable to diagnose the issue, I finally gave in and decided to ask Chunky what was failing. Within seconds, it found the problem. My Telegram API secret key had spaces. Chunky automatically removed it, corrected the configuration, restarted the service, and voila! It worked.

My first reaction was impressed, soon followed by fear. That made me realize how deep its access was, not contained to the chat window, or parts of a folder, but the entire OS. It had eyes everywhere.

After the security shock, I continued to periodically talk to the agent, while going on a walk or getting groceries, Chunky would get to work and document on my home iMac computer. I would give them instructions to give me notifications at certain times, but I would rarely receive those reminders and would ask Chunky why I wouldn’t get them; they would continuously tell me about the mistakes they made and how they would be fixed by next time. 

The interaction reminded me of Detroit: Become Human, the science-fiction game where androids slowly develop identity, autonomy, and emotional awareness. When the game was released in 2018, it felt futuristic and distant. I never would have imagined years later I would be texting an AI agent named Chunky Monkey, slowly learning about each other’s “self”. 

We are in this new reality of AI where agents are interacting with humans on a whole new level. They are becoming personal companions, assistants, and collaborators. We are left to ponder what is machine work and what is human work.

James Barrat, acclaimed AI speaker, foresees a future where important decisions governing the lives of humans are made by machines or humans whose intelligence is augmented by machines. Computers already manage huge parts of modern life, like financial banking, healthcare networks, transportation, and communication. But now, with AI, it’s not just processing information, it’s acting on it.

Not everyone sees this shift as entirely positive. 

Jess Lerma, a professor and member of the Distance Education Committee at West Valley College, has watched AI reshape education over the last several years. For Lerma, the concern isn’t simply whether students use AI, it’s whether students will slowly surrender the learning process altogether. “I believe that students will ultimately reject agentic AI, such as OpenClaw”, he says, “as it starts to devalue the very education they are seeking out at WVC."

Barrat describes humanity's growing dependence on machines as a surrender. With growing dependency, students may stop exercising creativity and exploration, which makes learning meaningful in the first place. Yet, while conducting my research with OpenClaw, I realized an important distinction between what the machine was doing and what I was ultimately responsible for. 

Philosopher and cognitive scientist Brian Cantwell Smith describes this distinction as judgement and reckoning. He argues that reckoning is the calculative prowess that AI already excels at: sorting information, evaluating possibilities, and processing data at enormous scales. Then he describes judgement as deliberate human thought grounded in ethical commitment and responsible action.

It’s not just knowing what decision can be made. It’s understanding what decision should be made. 

With my sister’s diagnosis, AI could compare insurance plans and organize treatment information faster than I ever could. It handled the reckoning. But it could not understand the emotional weight behind every decision our family was making. That judgment was still reserved for us. 

Lance Madanguit, a psychiatry resident physician, described that AI-powered dictation systems are already widely used to automate documentation and generate notes. The physician described the technology as something that allows “more dedicated cognitive labor to be placed into bedside patient care”. Not less humanity. More room for it. He emphasizes that “AI merely serves as a tool, not as a replacement for cognitive thinking."

As an iOS engineer at Mercury, Farhana Mustafa, also describes AI serving as a tool and rapidly transforming the operational side of software engineering. “AI hasn’t changed outcomes for me personally,” Mustafa explained. “It’s mostly being used to speed things up operationally.” Tasks once estimated to take several weeks can now potentially be completed in a fraction of the time. For repetitive tasks and bug fixes, Mustafa said AI performs well. When it comes to larger feature development, it struggles with context and sometimes introduces unfamiliar coding paradigms, code that doesn't align with the engineering team’s current software practices.

Like Brian Cantwell Smith explains, AI excels at reckoning. Judgement still belongs to the engineer when evaluating what code is actually good. 

However, as LLMs continue evolving, new software projects are being released at a rapid pace. An entire ecosystem is being built for agents as we speak: internet browsers designed specifically for agents, digital wallets to manage their own currency, and so much more. One of the most notable projects is Moltbook, a social media platform designed specifically for agents to post, comment, and interact with each other. If humans can develop slang and social norms online, can agents do the same?

Futurist and inventor, Ray Kurzweil once imagined “a world where the difference between man and machine blurs, where the line between humanity and technology fades, and where the soul and the silicon chip unite." Kurzweil promises an age in which the marriage of human sensitivity and artificial intelligence fundamentally alters and improves the way we live.

With the debate of how AI thinks like or interacts with humans, there is still a bigger question looming. Whose humanity will it reflect? Computer scientist and digital activist Joy Buolamwini, with a PhD from MIT, describes this problem as the “coded gaze”. Inspired by the term “male gaze”, a concept developed by media scholars to describe how in patriarchal society, art, media, and other forms of representation are created with a male viewer in mind, the “coded gaze” is similar. Prejudice can also be encoded into technology even though it's unintentional. With AI integrated in hiring, healthcare, housing, policing, education, and financial systems, the future of these systems will ultimately reflect the historical data it's built from, and the people building them. “None of us can escape the impact of the coded gaze. We must face it”, Buolamwini writes. “We need the voices of the unseen… immigrants, people of color, women.”

Maybe that’s why the AI agents feel so personal. It forces humanity to confront itself. What kind of labor do we value? What kind of thinking should we protect? What parts of ourselves are we willing to automate? Most importantly, who gets included when shaping the systems that may eventually govern everyday life?

When my sister was diagnosed, AI did not comfort her or love her. But it helped carry the informational load, making space for me to be there for her. And maybe that is the future we are heading towards. Not one where humans become obsolete, but one where technology clears away enough noise for us to live more in the moment and understand what truly makes us human in the first place.