TrafficScout

TrafficScout came out of a sprint I ran with a co-founder — our own take on the “build a startup a week” model. The idea was to compress the learning curve: ship something fast, try to get real traffic and revenue, and carry the lesson into the next one. Reddit kept coming up because it’s one of the best free sources of targeted traffic there is — if you can show up in the right threads with something actually worth reading.

The Thesis

Reddit is structurally hostile to most growth attempts — spam gets banned, and generic marketing gets ignored or reported. The bet was that an AI agent could do the tedious part — surface the genuinely relevant conversations — and help a founder respond in a way that added value instead of noise. Do that consistently and you’ve turned a chaotic channel into a dependable one.

What I Built

A tool that identified relevant subreddits and posts, and helped draft contextually appropriate, natural-sounding replies. I ran the growth for it — a launch that drove targeted traffic from the exact communities the tool was built for.

And it worked. There was clear, real demand: people installed it, and people paid for it. Willingness-to-pay was never the question.

What Happened

The problem wasn’t demand. It was the ground the product stood on.

Reddit’s anti-automation and bot detection are strict and have only gotten more sophisticated — even older, established accounts get caught. Keeping the tool working meant an escalating game of proxies and workarounds, and it put users’ own accounts at risk. That’s a cat-and-mouse fight you don’t win permanently; you just delay the next ban.

So despite paying customers, my co-founder and I made the call together to discontinue it. It sat in a gray area of Reddit’s terms of service, it wasn’t the kind of thing either of us wanted to build our name on, and it demanded constant upkeep just to stay alive — the opposite of a durable business.

What I Learned

Validated demand isn’t enough if the foundation underneath it is borrowed. TrafficScout had real paying customers and a real technical window — the window was just shorter than the demand suggested, because it depended on staying ahead of a platform’s own rules.

• A platform’s terms of service is a lease, not ownership — price that into how much you invest before you know how long you get to stay.
• Constant defensive upkeep (proxies, workarounds, staying ahead of detection) is its own tax on a business, even one that’s profitable.
• Willingness-to-pay answers “is there a market” — it doesn’t answer “is there a durable business.” Check both, separately.

Reddit growth is still a channel I think about a lot. If you’re working in that space, I’d like to trade notes.