Linus Torvalds had it backwards. In 2024, everyone can ship code. Not everyone can ship a story that sticks.

We're watching a weird inversion play out in real-time. The same week that yet another AI coding agent launched (which one was it? exactly), the deal that actually moved the needle was a founder who couldn't code their way out of a paper bag but absolutely crushed their Series A pitch deck. The technical co-founder is still listed on the website. Nobody's heard from them in six months. The CEO is doing three podcasts a week.

Here's what's actually happening: AI didn't just lower the barrier to building software. It obliterated it. Every Stanford sophomore and their ChatGPT instance can now ship a functional MVP in a weekend. GitHub is drowning in repositories with decent code and zero users. The marginal cost of "building the thing" is approaching zero, which means - and this is the uncomfortable part - building the thing barely matters anymore.

This is the third major platform shift where the valuable skill flipped. Web 2.0? Suddenly designers mattered more than backend engineers. Mobile? Distribution through the app stores became the game, not the elegance of your Swift. Now? The ability to articulate why your AI wrapper is different from the other 47 AI wrappers is worth more than your actual model fine-tuning.

Look at who's winning. Jasper didn't have better AI than anyone else - they had better positioning and enterprise sales motion. Character.AI's models weren't magic - their founder's pedigree and narrative about "AI companions" was. Even in the infrastructure layer, the companies breaking out aren't the ones with marginally better latency. They're the ones whose CEOs can explain their product category in a way that makes VCs feel smart for understanding it.

The venture math backs this up. YC's latest batch had something like 40% AI companies, most of them functionally identical under the hood. The ones that got term sheets weren't the ones with the cleverest RAG implementation. They were the ones who could tell you exactly which $500B market they were "Uber-ifying with AI" and why incumbents were structurally incapable of competing. Same code, different story, 100x valuation delta.

This isn't just about startups either. Look at the engineers-turned-influencers on Twitter with 200K followers. Their repos are fine, nothing special. But they've figured out that being able to explain technical concepts in viral threads is a more valuable skill than being able to implement them. One lands you consulting gigs and angel allocations. The other lands you a senior engineer salary, which - let's be honest - you could also get by being mediocre but great at Leetcode.

The really uncomfortable truth? This is probably correct market behavior, not a distortion. In a world where building is cheap, distribution is the moat. In a world where anyone can ship features, the company that controls the narrative controls the customer. Code is abundant. Attention is scarce. We've just been pretending otherwise because the tech industry spent 30 years venerating the builder and treating sales and marketing as necessary evils.

The old guard hates this. You can feel the resentment radiating from Hacker News every time a "non-technical" founder raises a big round. But that resentment is just cope. The game changed. The question isn't whether you can build it anymore - it's whether you can make people care that you built it. Whether you can position it before someone else claims the category. Whether you can tell a story that makes your solution feel inevitable.

So sure, talk is cheap. But in a world where GitHub Copilot and Claude can write your codebase, apparently code is cheaper.