America’s creative class just lost 200,000 jobs in four years. That’s not a rounding error. It’s a fundamental shift in how media, film, and the arts operate, and AI sits right in the middle of it.

TechCrunch launched an “AI graveyard” this week, tracking failed projects and startups. Apple’s repeatedly delayed Siri overhaul made the list. So did OpenAI’s messy super app launch. But the body count extends well beyond product flops. The real graveyard is in the employment numbers.

Here’s what’s actually happening. Streaming platforms and digital distribution already squeezed creative industries before AI entered the picture. Consolidation pushed margins down. Then AI arrived with a pitch: automate the expensive parts. Studios listened. So did publishers, production companies, and newsrooms.

The result isn’t just fewer jobs. It’s a collapse in how creative work gets valued. When a tool can generate rough cuts, draft articles, or produce background art, the market stops paying for those skills. The work still needs human refinement, but the economics shifted. Companies pay for the refinement, not the creation.

The Pattern Across Industries

Pennsylvania just required labels on AI-generated ads. Governor Josh Shapiro signed the Artificial Intelligence Deceptive Advertising Act on Thursday. Any ad significantly modified or generated by AI needs disclosure, whether it’s video, audio, images, or text.

That applies to everyone. Multi-billion-dollar TV spots and local restaurant flyers both need labels. The law recognizes something important: AI-generated content is flooding the market, and people can’t always tell the difference.

BBC’s Blue Planet III took the opposite approach. The nature documentary series deliberately kept imperfect shots, showing plankton and particles in the water. Executive Producer Elizabeth White explained they wanted to prove authenticity. Previous Blue Planet installments were known for impossibly crisp footage. Now that signals AI, so they’re embracing the rough edges.

Both moves, the Pennsylvania law and BBC’s production choice, point to the same problem. We’re in a credibility crisis. Too much content, not enough trust in what’s real.

What the Numbers Actually Mean

The 200,000 job losses break down across media production, film work, and arts jobs. Streaming drove part of it. The shift from cable to digital meant fewer people needed for distribution. Consolidation meant redundant roles disappeared after mergers.

But AI accelerated everything. Tools that automate editing, generate images, write copy, and produce music didn’t create new jobs at the same rate they eliminated old ones. That’s the core issue. The productivity gains went to companies and shareholders, not workers.

Meta built systems to flag school shooting threats using AI, routing high-risk cases to human reviewers and law enforcement. The New Yorker covered how it works and the tradeoffs involved. It’s one of the clearer examples of AI doing something humans probably can’t do at scale. Monitoring millions of posts for specific threat patterns requires automation.

That’s useful work. But it doesn’t employ 200,000 people. The AI applications that actually help, the ones with clear social value, tend to be narrow. The ones eliminating jobs are broad: anything involving pattern recognition, content generation, or basic creative tasks.

The Investment Question

Venture capital keeps pouring into AI startups. Funding rounds happen weekly. Valuations climb. But the TechCrunch graveyard shows what happens after the checks clear. Many AI companies can’t find product-market fit. Others solve problems nobody actually has. Some just run out of runway before the technology matures enough to matter.

Block released a voice model this week that runs locally on a MacBook M4 Max and sounds eerily human. That’s technically impressive. Whether it’s commercially viable depends on what people actually need realistic voice synthesis for, and whether they’ll pay for it.

A study found Grok’s AI overwhelmingly recommends Republican candidates while OpenAI’s GPT favors Democrats. The models disagreed in over 90% of nearly 500 races. That’s not a bug, it’s a feature of how these systems work. Training data, fine-tuning choices, and optimization targets all embed political assumptions. When companies claim neutrality, they’re either lying or ignorant about their own systems.

What Comes Next

The creative job losses won’t reverse. Those 200,000 positions are gone. The question is whether the industry stabilizes here or keeps contracting.

If AI tools plateau at their current capability, then studios, publishers, and production companies will adjust. They’ll hire smaller teams that use AI for grunt work. The jobs that remain will require higher skill levels. Entry-level positions, the ones that used to train the next generation, will mostly disappear.

If AI keeps improving, more jobs vanish. The threshold for “requires a human” keeps rising. Eventually you’re left with a handful of creative directors overseeing AI systems, and everyone else is out.

The funding will keep flowing until it doesn’t. Venture capital operates on power law returns. Most investments fail, a few pay for everything. AI fits that model perfectly. High risk, huge potential upside, lots of credible-sounding pitches.

But the workers who lost jobs don’t get power law returns. They get unemployment and a skills mismatch in a labor market that doesn’t value what they know how to do anymore.

Pennsylvania’s labeling law and BBC’s rough-cut documentary both try to address symptoms. The disease is economic. Creative work is being devalued faster than new opportunities are emerging. That’s the story the graveyard tells.