AI Is Browsing the Internet for Us
A phone conversation with my friend Lucas this morning made a bunch of things click at once.
Lucas: You deployed an app recently, right? What tool did you use?
Me: No idea, AI deployed it. Ask AI.
I said it without thinking — because that's genuinely how I operate now: I only care about results and the bill. What tool AI used, which approach is better — I couldn't care less.
That moment was a wake-up call.
When AI starts browsing the internet on our behalf, marketing to "humans" stops making sense. The real audience you should be serving is AI itself. Make it so AI can discover your product, pull information from it without friction, and interact with it directly.
Public Data Defines What AI Knows
The conversation kept going:
Lucas: AI says fly.io and Railway, recommends fly.io.
Me: Let me check my history… yep, that's exactly what I got too.
Lucas is on the other side of the planet, using a different AI, asking in a different way. But both AIs gave nearly the same answer.
Why?
Because they draw from the same pool of public data.
What's "public"? Simple: content you can see without logging in. Reddit, Stack Overflow, tech blogs, Hacker News — you Google it, you see it, and so can AI. That's public.
Now think about TikTok, Instagram, WeChat — you can't see anything without downloading the app and logging in. Behind those walled gardens, data doesn't get out. If AI can't read it, it might as well not exist.
The more I think about it, the more convinced I am: walled-garden platforms are relics of the last era — built for humans. In AI's worldview, information that isn't public simply doesn't exist.
Why AI Will Inevitably Browse for Us
Because starting is hard.
I've been meaning to do Reddit marketing for a while. Kept putting it off — didn't know where to begin. Yesterday over dinner, I casually asked AI about it. It searched relevant threads, drafted replies, and knocked out the first step for me. Suddenly the whole thing had momentum.
Most of the time it's not that people don't want to do things — it's that the activation energy is too high. AI just needs to get you past that first step, and the rest flows. This trend is irreversible — more and more actions will be initiated by AI, with humans just approving.
What I'm Doing Next
Once this clicked, the product direction became obvious:
1. Build Only for the Public Web
All products and content: open source, open comments, users can bring their own LLM, most features work without login.
2. Build Only What AI Can Fully Automate
No more products where humans need to learn a UI first. Every feature must be discoverable by AI, understandable by AI, and completable by AI — zero to done, fully automated on the user's behalf.
3. Marketing: Get Into AI's Knowledge Base
Back to the public data theme. Marketing strategy has to revolve around "make sure AI can find us":
Find matching interest communities on Reddit and respond to real needs with relevant content. Someone posts about wanting a specific type of character? Drop a link to a matching one. And yes — use AI to automate this step too.
Over time, when the next update of LLMs trains on Reddit data, our content becomes part of the search results for those queries.
This isn't the old "spend ad dollars → get conversions" playbook. It's "plant information → let AI harvest it." Spread keywords across countless niche interest areas so that when AI answers any related question, it naturally surfaces us.
TL;DR
- AI is browsing the internet for humans. If your product is only built for human eyes, you're invisible to AI.
- Public data is AI's only source of truth. If it's not public, it doesn't exist in AI's world.
- Go AI-first. Every capability should be open to AI — let it discover you, understand you, and act on behalf of your users.
This is what I'm building toward for the next five years.
I'm building an AI chat platform for interactive fiction. Open to conversations
