I Spent My Day Off Building a File That Does Almost Nothing

Here's why. And why half the internet is making the opposite mistake.

The web is locking AI out. Cloudflare blocks AI crawlers by default now. 2.5 million sites have banned AI training. As of last week, bots officially outnumber humans online — 57.5% of all web traffic. Everyone's terrified AI is stealing their content.

I did the opposite. I opened my entire site to every AI crawler I could name — GPTBot, ClaudeBot, PerplexityBot, door's open, read everything. Because here's the part nobody wants to say out loud: when you're a solo dev with 57 stars, AI isn't stealing from you. It's the only free distribution you have left.

And Google? 60% of searches now end with zero clicks. 93% in AI Mode. Ranking #1 doesn't mean an AI will ever cite you. So I stopped optimising for the dying channel and rebuilt everything to be the thing the machine actually reads.

What shipped this week on pcworkman.dev

• 56 pages — the site went from one messy blog.html to a real structure.
• 35 build-in-public episodes rewritten 1:1 from my LinkedIn.
• Sitemap, schema, RSS, robots.txt wide open to AI. 0 validation errors.
• 40 PLN (9.5€) for the domain, for a year.

The useless file? It's llms.txt. The data says it does nothing today. But it cost half a day and it's a bet on a web where machines are the audience, not an afterthought. Same bet I made with PC Workman: build for what's coming, not what's dying.

This week's plan

• Close the 7 GitHub issues I fixed but never actually closed (yeah, I know).
• Extend the roadmap past v1.7.9, toward Microsoft Store.
• Back to the actual app. A week on infrastructure is enough.

Welding plastics by day. Building by night. Friday will tell the truth. What are you building this week?