When every purple cow looks the same

Week #19 in Bangkok

A small heads-up before we start: this week I'm trying something different. Instead of the usual travel blog format, with Bangkok anecdotes, expat life, condo stories, I want to write something more useful. More "indie hacker lessons learned in the trenches" than "here's the pad thai I ate for lunch".

If you prefer the other blog format, please tell me. Drop a comment here, reply to my post on X, send me a message. I genuinely want to know what you find valuable here.

Now let’s get into it.

Purple cows?

Seth Godin's purple cow concept is marketing gospel: in a field full of normal cows, be the purple one. Be remarkable. Stand out.

The problem? In 2026, everyone is a purple cow. And when everyone is purple, purple becomes the new black and white.

I had a long conversation this week with Nico, the famous French maker. We ended up dissecting this exact pattern. The conversation started with something he threw at me: statistically, as an indie maker, you have almost zero chance of making it. Not a new idea. But hearing it laid out plainly forces you to sit with it for a minute.

My answer was simple: I know. But that doesn't stop me from trying.

Let me explain everything to you in detail.

The AI wrapper epidemic

There is a thing that a lot of people in this space know but don't say out loud: the current wave of indie makers is largely building the same things. AI wrappers. GPT on top of a niche. "I built a SaaS in 72 hours with Claude Opus 4.6 / GPT 5.4 / Gemini 3.1" content every single day.

Some of these work. Most don't.

The people who struggle are almost always the same profile:

No real backend, no defensible moat
Building things that anyone with a free Lovable or Google Studio account can replicate over a weekend
Posting demos constantly, launching on Product Hunt every other week
And stagnating

If you recognize yourself here, this post is for you.

The problem isn't effort. The problem is differentiation.

When your product can be copied in two days, and your marketing is "I built this with AI", you're not a purple cow. You're a black and white cow with purple paint.

The "move to Asia" meta

There's a second version of this problem, more specific to the X indie hacker bubble: everyone is living the same life.

Everyone left their job. Everyone moved to Southeast Asia. Everyone posts from a co-working space with their OpenAI Codex setup. Everyone documents their "journey". And then they're surprised when their audience stagnates at 1,000 followers despite grinding daily.

These days, dropping everything to go to Asia is no longer extraordinary. Three years ago, it was a statement. Now it's a template.

What still works on X? Drama. Real moments. MRR that actually moves. Failures. Controversies. The stuff that creates an emotional reaction. Pieter Levels barely talks about his products. He talks about politics, provokes people, shares raw opinions. His engagement is insane.

X is an attention machine, not a sales tool. You sell in the long run, because you post things that make people feel something. Product demos from AI wrappers that everyone has already seen? Nobody cares.

Many makers believe that X's algorithm has changed. It's not the algo itself that has changed. It's that what was once remarkable isn't anymore.

The hard stuff nobody wants to build

Here's where my story gets relevant.

For the past few weeks, I've been deep in something unsexy: building a scraping infrastructure for Google SERPs from scratch. No third-party APIs. No SerpAPI. Just me, a VPS, some proxies, stealth browsers and a growing pile of failed experiments.

Why? Because the SEO tool I'm building needs this data, and the API providers charge a fortune at scale. My goal was to hack the system so my entire infrastructure costs the price of a VPS ($5/mo) instead of hundreds of dollars. That asymmetric cost structure is what makes a SaaS with a healthy margin possible.

Accepting the verdict

After two weeks of this, I need to face the reality: maybe I might not crack it.

So, at this point, the most sensible solution is to set a deadline, give myself one more week, and if that doesn't work, switch to SerpAPI and do something else. I absolutely must avoid spending 3 or 4 months on this.

My honest assessment? I can probably build something that scrapes a few hundred queries a day without getting blocked. Useful for bootstrapping a free tool with low traffic, like SnappyName right now.

Note: You probably don't make the connection between domain name searches and Google's SERPs, but there is one. I can't wait to show you what I've cooked up!

But at large volume? Google has put up too many walls. It would require industrial-grade infrastructure: dedicated servers, clickstream data purchased at scale, the kind of setup that costs millions like Ahrefs. Not my level.

The conclusion is almost certain: large-scale Google scraping isn't viable for a solo indie hacker in 2026. It's not impossible. It's just prohibitively expensive if you do it the right way, and unstable if you try to hack it. I've tried nearly every low-cost option on the table. I'm close to the end of the list.

The deadline is setting itself.

Why it was worth it anyway

No regrets though. Here's what this two-week detour actually gave me:

I now know exactly what's feasible and what isn't. I know the architecture. I know the weak points in Google's bot detection. I built something that partially works. And critically, I built something that nobody can copy in two days with an AI prompt.

That's the actual purple cow. Not the output. The process.

This is pure R&D. It's unglamorous. It makes zero followers on X. Nobody claps for you when you spend 4 days going down a rabbit hole that ends in failure. But this is exactly the kind of work that creates tools with real competitive moats.

Some others indie makers I know here in Bangkok put it well: you have to launch multiple projects, because you never know which one will take off. Some successful products that reached the holy grail of 10k MMR were launched as side projects that no one believed in, and shipped in a weekend.

Key takeaways

The infrastructure I'm building is my bet. Maybe I lose. Maybe the whole thing collapses and I pivot to SerpAPI next week. But I'll have no regrets. I'll have tried everything.

And when it does work, even partially? That's not just a competitive advantage. That's a launch post that actually has something to say.

This was week #19 in Bangkok. The scraping verdict is almost in. The differentiation game is ruthless. And the hard stuff is still the only stuff worth building 💪