What should AI know about you?

Probably more than your LinkedIn bio

Last month on a group trip, my friend showed me an agent he was building that runs a prompt through every major LLM and aggregates their opinions into one answer. To demonstrate, he cheekily submitted the prompt: “will Jay Feng be successful?” And I watched as Claude, ChatGPT, Gemini, and Grok all came together to form their opinion on my future success. I’d be lying if I said I wasn’t nervous watching four AI models deliberate on my future. A minute later they agreed I’d achieve some measure of success: whatever that meant.

I had a new idea. I took over the computer and expanded the prompt: which of the ten friends on the ski trip was the most likely to be successful? People laughed as I put in the prompt. They stopped laughing once the results came back. For the next few hours, we jockeyed for position to move up in the rankings. A few people were genuinely insulted by how low the LLMs ranked them. People started redefining what success meant altogether: adding health, happiness, creativity, freedom, any kind of framing that would move them up.

At the end of the chaos, I realized an interesting pattern. The LLMs weren’t ranking who was the most objectively successful. They were ranking whose ideas were most accessible online. For most of us, all the AI had to go on was a LinkedIn bio and a few social posts. But another friend, like me, had spent years writing about her work, life, and wins and losses throughout the years. The models just had more to work with.

Writing online has always had obvious benefits like building an audience, clarifying your thinking, and creating serendipity. But those benefits always required a human on the other end to actually read what you wrote and make the connection. What’s changed is that LLMs are now doing that work at scale.

A billion people are using ChatGPT weekly. AI is becoming the default layer for how people search, evaluate, and make decisions about their lives. And unlike a human, it can read everything you’ve ever published in seconds and match you against dimensions you’d never think to optimize for. The stuff you write online isn't just content anymore. It's your representation in every room the AI enters on your behalf.

I tried this with Boardy the other day. Boardy.ai is a networking AI bot that’s become a way to match startups to venture capitalists. A venture capitalist only has so many slots per week to hear founder pitches, and Boardy is already picking which founders get those slots. But when I asked Boardy who I was, I got this response:

“You are Jay Feng, co-founder of Interview Query, a data science training platform you’ve grown to 50K followers and a team of 10+.”

But is that it? What about the years I spent learning to surf in San Francisco and what that taught me about getting better at hard things? Or the months I spent reading every Berkshire Hathaway shareholder letter and what it taught me about how Buffett actually evaluates businesses?

In the real world, these overlapping niches are the ones that surface after three drinks and disappear into the ether. But if the AI had access to everything I’d ever written, it wouldn’t need three drinks. It could match me on the stuff that actually matters, what I truly care about.

Gwern argues writing for LLMs allows them to learn your preferences and values so they become better tools for you. Tyler Cowen frames it as intellectual immortality: write enough and future AIs can reconstruct your thinking for your grandchildren. Both are thinking about AI as your audience. The bigger shift is that AI is reading your writing to represent you to other people.

For example, if a startup asks “who should we pay $1000/hr for direct knowledge of consumer ed-tech SaaS companies,” I want to be the first result. A year ago, that search would have happened on Upwork or LinkedIn, filtered by job titles and keywords. But this is now increasingly happening first with LLMs. Which means the more you write, the more dimensions the AI has to work with, and the more likely you are to show up when it matters.

And it won’t stop with credentials. If the AI has access to how you actually think, what you obsess over, what you’ve learned the hard way, you’re distinct enough to match on things that matter. Not just “ed-tech founder” but “ed-tech founder who found a way to pivot towards B2B during a recession”. Every essay you write adds another surface for something to stick to.

I don't think you need to document every thought you've ever had. But the stuff you think doesn't matter, like the hard-won lessons or the weird niche obsessions, those might be exactly what connects you to someone you haven't met yet.

AI can only match you on what you've made legible. The question is how much of yourself you're willing to make visible.