Jeff Johnson, the indie developer behind LapCat Software and known for apps like Checked!, has published what he calls 'The Obligatory AI Post' on his blog—a scathing critique of how Large Language Models have infiltrated software development without delivering demonstrable improvements to end users.

The Productivity Theater Problem

Johnson's core argument centers on an uncomfortable question: if AI-generated work is so valuable, why wasn't it being done by humans before? 'If there was money to be made, then humans had financial incentives to do the work,' he writes. He points to App Store data that illustrates this tension—app submissions have surged dramatically in recent months (a trend Apple itself has confirmed), yet downloads and revenue remain essentially flat. 'Everyone is making video-coded apps, but nobody is buying them. In my opinion, that's productivity theater.' The Cupertino-based developer has spent decades writing code the old-fashioned way, manually typing into older versions of BBEdit and Apple Xcode for his own projects. He notes that while countless developers claim AI makes them dramatically more productive, he hasn't seen or enjoyed any tangible benefits from these alleged efficiency gains in the software he consumes.

Speed Is a Vice

Beyond the productivity theater critique, Johnson brings a perspective that's increasingly rare in tech discourse: that moving faster isn't inherently better in software development. 'I've always felt that speed is a vice rather than a virtue with regard to software development,' he argues, drawing on his previous writing about Mac OS X Snow Leopard's focused release cycle. His analogy cuts to the heart of the matter: AI produces source code at an unprecedented rate, like a hare racing a tortoise. But software, Johnson contends, is more akin to climbing a mountain—when you're near the summit, there's not much altitude left to gain but a long way to fall. 'Innovative software is complex and unpredictable. It's ready when the engineers have solved the problems, not on some arbitrary date preannounced by a glory-hungry CEO.'

The Resource Paradox

Johnson also tackles the irony of AI's environmental footprint. The technology that promises to make computing easier for the masses has triggered mass shortages of computing components and increased electricity costs for consumers. Training and running LLMs requires enormous power, contributing to climate instability at a time when industry should be reducing its carbon impact. He doesn't shy away from the intellectual property controversy either, calling out LLM vendors directly: 'The LLM vendors have perpetrated mass theft of intellectual property, aided by regulatory inactivity.' The workers who remain employed are increasingly burned out dealing with AI-generated busywork and insatiable demands for productivity gains that don't translate to better products.

The HAL 9000 Precedent

Johnson ends with a reference that's both literary and cautionary: Stanley Kubrick's 2001: A Space Odyssey. He notes the famous bone-to-spaceship cut represents his view on the 'just a tool' defense of AI. Yes, it's a tool—but one that fundamentally changes the species using it. The problem in Kubrick's film wasn't HAL's intelligence; it was humans placing too much trust in AI with free agency. 'It has always been due to human error.'

Key Takeaways

  • Johnson questions whether AI-generated code represents genuine value or just cheaper work that wasn't worth doing before
  • He argues software maturity requires deceleration, not acceleration—mature code is battle-tested and dangerous to change rapidly
  • The environmental and resource costs of LLMs contradict their promise to democratize computing
  • LLM vendors have faced criticism for intellectual property practices, raising ethical concerns about training data

The Bottom Line

Johnson's contrarian take serves as a necessary counterweight to the relentless AI boosterism flooding developer communities—if the apps being built with these tools aren't delighting users or generating revenue, perhaps the productivity gains are more illusion than substance.