The Threshold Collapsed to Zero: Building in Public in the Age of AI Cloning

Arvid Kahl, one of the pioneers of the "build in public" movement in the software world, made a confession in recent months (April 2026) that fundamentally shook his own thesis. After years of advocating for sharing metrics, processes, and financial data transparently, Kahl stated that the "safe sharing threshold has collapsed to zero" due to the speed AI has reached. Now, every metric and architectural detail you share openly turns into a cloning blueprint that AI agents can replicate over a weekend.

When I first heard this, I stopped and looked at myself. Because I am building ProductLog, a platform designed for founders to share their product development processes in a structured way.

Which means I am not just an observer of this new threat; I am "Exhibit A" and the founder who hit his own foot with the heaviest axe.

How I Distributed My Own Product's Cloning Blueprint

When building ProductLog, my goal was to let founders transparently log their sprint goals, technical choices, and roadmaps. However, for AI assistants, this structured data isn't a community journal; it is a perfect, step-by-step product specification (blueprint) document.

What did I do on my own ProductLog profile? Using my own platform's format, I openly leaked the following:

• Go-To-Market (GTM) Strategy: In my "Business Strategy" board—before launching a single feature—I shared my targeted SEO keywords ("Canny alternative", "best changelog tools"), community-led launch plans, and backlink growth loops in plain text. A competitor could feed this strategy to an AI and beat me to my own game before I even write the first line of code.
• Technical Details and AI Confessions: While detailing backend architecture quirks like email delivery, queue management, and login errors, I openly stated that the code was mostly AI-written (via vibe coding) and that the backend suffered from major DRY/SOLID violations. This is equivalent to telling a potential cloner: "The backend is currently a spaghetti mess; you can use AI to rewrite it faster and cleaner than I did."
• Monetizable Features (The Trap): I explained in detail where competitors got it wrong with their pricing ($20-50/month) and which two features (widget integration and custom subdomains) anyone trying to clone ProductLog and monetize it faster should prioritize.

ProductLog's own format made my own product the easiest thing to clone. Instead of being a protective shield, transparency became a free development manual I handed to my competitors.

The Moat of the New World: Hide the Code, Build Connection

If AI has reduced the cost of writing and copying code to zero, what is the "moat" that protects a startup from being cloned?

Arvid Kahl's 2026 reversal points to exactly this: Software engineering capability or the code itself is no longer a moat.

When advanced models like GPT-5.6 and Fable 5 can reconstruct a functional codebase from a few screenshots and feature descriptions, the code ceases to hold proprietary value.

The only thing AI cannot clone is the human and social capital outside the codebase:

• Distribution Power and Founder Brand: Users choose a product not just for its technical specs, but for the connection they build with you and your honest struggle (storytelling) along the way.
• Unclonable Network Effects: Code can be cloned, but the community of people gathered around that code, the trust relationships, and the ecosystem cannot.

This shapes the future of ProductLog as well. If this platform remains just a place where founders display their technical showcases, it will turn into a hunting ground for AI agents. But if it evolves into a distribution network where founders support each other, grow together, and build trust-based relationships, even if AI clones the code, it can never replicate the ecosystem.

Conclusion: The New Rule of Building in Public

Building in public is still valuable in the AI era, but the motivation has changed. The new rule is this: Share the struggle openly, but keep the blueprint private.

Retreating entirely into your shell and hiding your code out of fear of being cloned might seem like a solution. But while hiding your code protects you from copycats, it won't prevent you from drowning in the deep loneliness where nobody even knows you exist.

Survival in the new world isn't about hoarding code; it's about building the human connection that lies far beyond the code. But how do we practically continue building in public under these constraints? I explore the new playbook and technical defenses in Build in Public 2.0: The New Playbook for Sharing Openly in the Age of AI Cloning.