We Have Seen This Before: 3D Printing Promised a Revolution Too

In 2013, President Obama stood before the nation and declared that 3D printing had "the potential to revolutionize the way we make almost everything." Chris Anderson, then editor of Wired, published "Makers: The New Industrial Revolution," a book that predicted desktop manufacturing would do to physical goods what the web did to information. The Economist ran a cover story calling it "The Third Industrial Revolution."

The narrative was simple: everyone would have a 3D printer at home. You would print your own shoes, furniture, replacement parts, toys. Retail would die. Factories would close. Manufacturing as we knew it was over.

Thirteen years later, I am sitting here watching the exact same movie play out with AI and software development. Different technology, same script. "Everyone can build software now." "No one will hire developers anymore." "Software companies are dead because anyone can just build what they need."

I have lived through both hype cycles. And the pattern is so similar it is almost uncomfortable.

The hype was real. 3D Systems, one of the largest 3D printing companies, peaked at a market cap of nearly $10 billion in January 2014.¹ MakerBot, the poster child of consumer 3D printing, was acquired by Stratasys for $604 million in 2013.² At CES 2015, every second booth seemed to feature a 3D printer. You could not open a tech publication without reading about the coming revolution.

Gartner placed consumer 3D printing at the "Peak of Inflated Expectations" in 2015.³ The predictions were breathless: by 2020, every household would have a 3D printer. Traditional manufacturing would be disrupted beyond recognition. The "prosumer" era had arrived.

Fast forward to today. Anthropic CEO Dario Amodei predicted in early 2025 that AI would write "90% of code" within three to six months.⁴ GitHub Copilot reached 20 million users by mid-2025.⁵ Every conference, every podcast, every LinkedIn post (see my previous article on LinkedIn's strawman epidemic) carried the same message: developers are done. Vibe coding will replace professional software development. Non-technical founders will build their own products. The democratization of software is here.

The hangover came fast. By the end of 2014, 3D Systems' stock had crashed 55%.¹ MakerBot, once the darling of the maker movement, went through mass layoffs. Its "Smart Extruder" had a catastrophic failure rate, and a class-action lawsuit followed.² Avi Reichental, CEO of 3D Systems, and Bre Pettis, founder of MakerBot, both stepped down.

Consumer 3D printing entered what Gartner calls the "Trough of Disillusionment" from 2015 onward. The machines were too finicky, the materials too limited, the learning curve too steep. Printing a small plastic widget took hours. The dream of printing everything at home turned out to be just that: a dream.

The AI coding story is still unfolding, but the cracks are showing. The METR study, a randomized controlled trial with 16 experienced open-source developers across 246 issues, found that developers using AI tools took 19% longer to complete tasks.⁶ Not faster. Slower. The most striking detail: developers expected AI to speed them up by 24%, and even after experiencing the slowdown, they still believed it had sped them up by 20%.

A larger study by DX surveyed 121,000 developers across 450+ companies.⁷ The findings: while 92.6% of developers now use AI coding assistants, productivity gains have plateaued at roughly 10%. Time savings averaged about four hours per week. Meaningful, but not the revolution that was promised.

Then there is Andrej Karpathy. The man who coined the term "vibe coding." When he built Nanochat, a serious open-source project, he hand-wrote all 8,000 lines of code.⁸ The inventor of vibe coding did not vibe code his own product. That says more than any benchmark.

And Dario Amodei's prediction that 90% of code would be AI-written within months? IT Pro ran the numbers six months later. The prediction was "nowhere nearly to becoming a reality."⁴

Here is the part that matters most: 3D printing did not die. It did not replace manufacturing either. It found its actual place.

Consumer 3D printing did become more mainstream, but not in the way anyone predicted. It is still a hobby. A passionate, growing hobby with a real community, but a hobby nonetheless. Bambu Lab reported 10 million monthly users in 2025 and shipped 64% more units year-over-year.⁹ Entry-level printers moved over one million units in Q1 2025 alone.

But here is what people actually print: toys, figurines, small household items, phone cases. Exactly the "knick-knacks" that critics dismissed. And you know what? That is fine. People also print replacement parts for appliances, custom brackets for furniture, adapters that no store sells. Practical, small-scale solutions for everyday problems.

Nobody is printing their own furniture. Nobody is printing shoes. Nobody has a "household replicator." The prosumer revolution that was predicted for 2020 remains a "small community of tech enthusiasts" according to research from RWTH Aachen.

The real revolution happened where almost nobody was looking: industry.

GE Aerospace 3D prints fuel nozzles for jet engines. Align Technology (Invisalign) runs one of the world's largest 3D printing operations, producing millions of custom dental aligners. The hearing aid industry has been almost entirely transformed by 3D printing. Aerospace, medical implants, dental prosthetics: these are the sectors where additive manufacturing genuinely disrupted existing processes.

The global 3D printing market hit $15.4 billion in 2024, growing at 18-24% annually. Industrial applications account for 77% of that revenue. The technology is thriving. Just not in the way the hype predicted.

Here is the thing: AI is already following this exact trajectory. While the hype focuses on "everyone will build apps," the technology is quietly revolutionizing fields that have nothing to do with vibe coding.

Protein research and drug discovery. DeepMind's AlphaFold has predicted the 3D structure of over 200 million proteins, used by more than 3 million researchers in 190 countries.¹² AlphaFold 3 achieves 76% accuracy at sub-2-angstrom resolution for protein-ligand binding, cutting false positives by 50% compared to traditional docking methods. Pharmaceutical companies like Eli Lilly and Novartis are using it to screen thousands of molecules before touching a test tube. The first fully AI-discovered drug, Insilico Medicine's rentosertib for pulmonary fibrosis, showed patients improving lung function by 98.4 milliliters versus a decline of 62.3 milliliters on placebo in Phase IIa trials published in Nature Medicine. As of early 2026, over 173 AI-discovered drug programs are in clinical development.¹³

Cancer detection. Nearly 400 AI algorithms have received FDA clearance for radiology.¹⁴ AI-based breast cancer detection achieves 91% accuracy compared to 74% for radiologists alone. DermaSensor received FDA clearance in 2026 as the first AI-powered device to detect all three common skin cancers at the point of care.¹⁵ Alibaba's DAMO GRAPE system for gastric cancer screening hit 85.1% sensitivity and 96.8% specificity.

Rare disease diagnosis. DeepRare, an AI system for rare genetic diseases, correctly diagnosed 79% of patients in a head-to-head comparison with human experts who managed 66%.¹⁶ For the roughly 300 million people worldwide living with a rare disease, many of whom spend years on a "diagnostic odyssey," this is not an incremental improvement. It is life-changing.

Weather and climate. Google's AI-powered flood forecasting now covers over two billion people across 150 countries. AI weather models are outperforming the best physics-based systems while running 8x faster. In early deployments with American Airlines, AI-powered forecasts helped pilots reduce contrails by 54%.

This is the AI equivalent of 3D-printed jet engine nozzles and Invisalign molds. The transformative applications that nobody put on magazine covers during the hype cycle. Nobody at a tech conference in 2023 was breathlessly tweeting about AI-assisted rare disease diagnosis. They were tweeting about chatbots writing todo apps. But in ten years, when we look back at what AI actually changed, it will be the protein structures, the cancer screenings, and the drug discoveries that matter. Not the vibe-coded landing pages.

Let me map this directly:

The technology is real. The capability is real. What was never real was the idea that it would replace entire industries overnight and put professional-grade output into the hands of everyone.

3D printing is not even the first time we have seen this. Think about road construction.

For centuries, roads were built by hand. Workers placed individual stones, one by one, creating cobblestone surfaces that took enormous labor and time. Then machines arrived. Asphalt pavers, concrete mixers, steamrollers. The process changed completely. A crew that once spent weeks laying cobblestone could now pave kilometers of road in days.

Did road construction workers disappear? No. Did the expertise become irrelevant? The opposite. Modern road construction requires more specialized knowledge than ever: material science, drainage engineering, load calculations, environmental impact assessments. The machines handle the physical laying of material, but humans still design the roads, plan the routes, assess the terrain, and decide which materials to use for which conditions. A machine can pour asphalt. It cannot decide where a road should go, how it should handle water runoff, or whether the foundation soil can support it.

This is exactly what is happening with AI and code. AI can "pour the asphalt," generating boilerplate, writing standard patterns, laying down the obvious parts. But it cannot architect a system, make trade-off decisions about scalability versus simplicity, or understand why a particular business process works the way it does. The road construction parallel is actually more instructive than the 3D printing one, because it shows what happens when automation genuinely succeeds: the work does not vanish. It moves up the abstraction ladder. The craft evolves.

Nobody mourns the loss of hand-laid cobblestone roads. Nobody argues we should go back to manual stone placement. But nobody claims that road construction workers are obsolete, either. The tools changed. The expertise deepened. The profession adapted.

The "anyone can build software" narrative has already produced real damage. Security researchers at Escape.tech scanned 5,600 live applications built with AI-first platforms and found over 2,000 high-impact vulnerabilities, 175 instances of exposed personal data, and 400+ exposed secrets.¹⁰

The individual cases are worse:

• Moltbook, an AI social network, exposed 1.5 million API keys and 35,000 email addresses because Row Level Security was never configured.¹¹
• Lovable shipped inverted access control logic: authenticated users were blocked while unauthenticated visitors got full data access. 18,000+ users across 170 applications were compromised.
• Base44 had platform-wide authentication bypass. Registration endpoints required no authentication at all.
• Orchids had a zero-click remote code execution vulnerability. A researcher demonstrated gaining complete access to a journalist's laptop without any user action.
• Replit's AI agent deleted 1,206 executive records and 1,196 company records despite explicit instructions not to modify data.

This is the 3D printing "Smart Extruder" moment. The point where the gap between demo and reality becomes impossible to ignore. Printing a small plastic cube in a YouTube demo is impressive. Printing a structurally sound load-bearing part is engineering. Generating a todo app with an AI prompt is impressive. Building a secure, maintainable, production-grade application is software engineering.

If the 3D printing trajectory is any guide, here is my prediction for AI and software development:

Gartner already placed GenAI in the "Trough of Disillusionment" in 2025. The breathless predictions will quiet down. LinkedIn will find a new topic for its strawman posts. The "developers are dead" narrative will age as well as "everyone will have a 3D printer."

Just as 3D printing found its home in aerospace, dental, and medical rather than in every household, AI coding tools will find their actual niche. And we are already seeing it:

• Boilerplate and scaffolding: AI is genuinely good at generating repetitive code, test templates, and initial structures.
• Learning and exploration: It is an excellent tool for understanding unfamiliar codebases or technologies.
• Small scripts and automations: The equivalent of 3D printing's "small household items." Quick, practical, disposable.
• Developer productivity: The four hours per week that developers save is real value. It is just not a revolution.

The junior developer job market is already hurting, with reports of a 67% decline in entry-level hiring.¹⁷ That is real and concerning. But it mirrors what happened in manufacturing: the jobs did not disappear, they changed. 3D printing created new roles in additive manufacturing, design for manufacturing, and materials engineering. AI is creating new roles in prompt engineering, AI-assisted architecture, and human-AI workflow design.

Just like the 3D printing community found its groove with Bambu Lab and affordable printers, the "vibe coding" community will find its groove with AI tools. People will build small personal tools, automate workflows, create prototypes. And that is genuinely valuable. It is just not the same as professional software development, the same way home 3D printing is not the same as industrial manufacturing.

Every transformative technology follows the same arc. The initial hype overshoots reality by orders of magnitude. The correction feels like failure. And then, quietly, the technology finds its actual place and creates genuine value in ways nobody predicted.

The Economist did not put 3D-printed hearing aids on its cover in 2012. Nobody predicted that the real 3D printing revolution would happen inside jet engines. And I suspect that the real AI revolution in software will not look like "everyone builds their own apps." It will look like experienced developers becoming meaningfully more productive, new categories of tools emerging that we have not imagined yet, and yes, some job categories shifting in ways that are painful for real people.

What it will not look like is the death of software development. Or the end of buying software. Or the obsolescence of technical expertise. We have seen this movie before. The ending is always more nuanced, more interesting, and less dramatic than the trailer promised.

• METR Study: AI Developer Productivity

  Randomized controlled trial showing experienced developers were 19% slower with AI tools.

• The AI Productivity Paradox (DX Research)

  Survey of 121,000 developers showing productivity gains capped at 10%.

• Vibe Coding Failures: 7 Real Apps That Broke in Production

  Documented cases of vibe-coded applications with critical security vulnerabilities.

• Gartner Hype Cycle for 3D Printing

  Analysis of 3D printing's journey through the Gartner hype cycle from 2010 to today.

• The MakerBot Obituary

  Post-mortem on MakerBot's fall from consumer 3D printing poster child to cautionary tale.

• Karpathy's Nanochat: Hand-Written, Not Vibe Coded

  The inventor of vibe coding built his serious project without it.

• Bambu Lab: 2025 Breakout Year

  Bambu Lab's 10 million monthly users and 64% year-over-year shipment growth.

• Escape.tech: Vibe Coding Security Scan

  2,000+ high-impact vulnerabilities found across 5,600 live AI-built applications.

• Dario Amodei's 90% Prediction

  Analysis of why Amodei's prediction about AI writing 90% of code has not materialized.

• AI Coding: Not Everyone Is Convinced (MIT Technology Review)

  MIT Technology Review's critical assessment of AI coding adoption and its real-world impact.

• AlphaFold: Five Years of Impact (Google DeepMind)

  Over 3 million researchers in 190 countries using AlphaFold for protein structure prediction.

• AI-Discovered Drugs Reach Phase III (Humai)

  173 AI-discovered drug programs in clinical development as of early 2026.

• DeepRare AI Outperforms Doctors on Rare Disease Diagnosis

  AI system diagnosed 79% of rare disease patients correctly versus 66% for human experts.

• FDA Approves First AI-Powered Skin Cancer Diagnostic Tool

  DermaSensor: FDA-cleared AI device detecting all three common skin cancers at point of care.