You Can’t Prompt Your Way to a Product
Early in Morpho’s life, we had a client come to us. A startup founder who’d raised $1.5 million.
Before finding us, he’d worked with a fancy San Francisco design firm. They charged him $250,000.
Partly because they were a fancy San Francisco design firm, and partly because, well, they probably knew how much he’d raised.
He came to us and the product was beautiful. It really was. Until our engineers looked at it and realized it was, to put it plainly, a pretty picture that couldn’t be made.
I found that pretty funny at the time. Of course, sad as hell for the guy, but funny. Because now I’m watching the same thing play out.
Except people aren’t spending $250K to get there. They’re spending 25 cents on ChatGPT credits.
The parallel is almost too on the nose. CAD files that can’t be manufactured, that haven’t been through DFM, are the same as a rendering.
Maybe they get you a little closer, but the ability to actually build the product? Same thing. And now we’ve got people burning through both the $250,000 and the 25 cents and getting similar results.
To be fair, the San Francisco firm was world class. Most people suck at prompting. But either way, you end up with something that looks incredible and can’t be made.
And here’s the thing: that’s actually incredible. It’s a stunning technological advancement that you can get to that point in minutes, sometimes seconds.
But people have to know the difference between having that and what the reality is. If it pushes you in the right direction, great. If it doesn’t, you’re in for a world of hurt.
The barrier to invention just collapsed. The barrier to execution didn’t.
Hardware used to filter itself.
Getting a product to the point where it looked credible — real CAD files, a working prototype, a BOM — required enough effort that it weeded out people who weren’t serious. That filter is gone.
AI for manufacturing and design has compressed what used to take a small engineering team months into something a single founder can do in a week.
That’s genuinely useful. It accelerates concept validation, helps catch obvious design problems early, and makes it cheaper to explore more ideas before committing to tooling.
But manufacturing doesn’t care about your render. Steel molds don’t care how clean your AI-generated geometry looks. A PCB supplier in Shenzhen isn’t going to quote you on a concept board.
The physical world has its own constraints: draft angles, parting lines, wall thickness, component lead times, assembly sequences… and none of those are solved by better software.
Now, I’m not naive. I really believe in tech. I believe that one day you’ll be able to use AI to design an entire product end to end.
But the questions are: when is that day, how available is that tech, and how accessible is it to startups right now?
Because right now, no matter who you are, no matter what AI you’re using, you cannot produce a fully DFM’d product from start to finish. That’s what matters, and that’s why this conversation is important.
The bottleneck has always been execution. AI just made that more visible by removing every bottleneck upstream of it.
More ideas, fewer products
There’s a parallel worth paying attention to in software.
The tools for building apps got so accessible that almost anyone can throw something together.
But the number of apps with meaningful use cases, the number people are actually using — that hasn’t kept pace.
Just to paint a picture, 73% of AI startups funded in 2023 have already pivoted or died. The total capital destroyed: $2.3 billion.
10 million projects have been created on Lovable alone. 41% of all code written globally is AI-generated.
As mentioned in the article above, the volume of products being generated right now is unprecedented. Most of them never shipped. Most never mattered.
The advances in software development from AI are probably a hundred times greater than what’s happened in hardware. And even in software, the product graveyard is full. Hardware is heading the same direction, with a harder floor.
The number of people who will be able to design something that looks like a real product is about to explode. The number who take it to mass production won’t keep pace. Not because manufacturing is getting harder — because it’s not — but most people will stop when the hard part starts.
The founders who close the distance between idea and physical product are going to be rarer, and more valuable, than they’ve ever been.
The tools lower the floor. They don’t raise the ceiling.
“Smart” is no longer a feature
I’ve been saying this for a while now: there’s no need to call your product “smart” anymore.
If it’s not smart, you should probably just call it dumb. Because that’s the reality.
The Internet of Things was so buzzy for so long. “Smart home.” “Smart device.” Now it should just be expected.
If you’re launching a connected product and leading with the fact that it’s smart — act like you’ve been here before. Of course it’s smart.
In ten years, a product without embedded intelligence won’t be a simpler product. It’ll just be a worse one.
Edge AI, local on-device processing, sensor miniaturization, better batteries… These aren’t exotic features anymore. They’re the baseline the market is moving toward.
The founders building now need to be thinking about where that baseline lands in three to five years when their product actually reaches scale, not where it is today.
The products that win are going to be the ones where intelligence is integrated so naturally that users don’t even think about it. It disappears into the object.
The companies that win won’t be AI companies
A lot of people are waiting for some AI company to come solve hardware. It won’t happen that way.
At Morpho, we use AI for renderings all the time. It’s a great way to conceptualize ideas, explore colors and finishes, get alignment fast. I’ve been doing what people are now calling “vibe coding” product concepts for two years. We’re at the front of this. But we know exactly what it is and what it isn’t.
What’s going to win is a hardware company that has a real grasp on AI for manufacturing.
One that understands mechanical and industrial design, knows how to run a supply chain, knows what actually happens on a factory floor and uses AI to move faster through work it already knows how to do.
That’s different from AI solving hardware. It’s hardware people using AI the way a good craftsman uses a better tool. The craft still has to be there.
A small team today can do what took a hundred people ten years ago. But the teams doing it are the ones who understood the work well enough to know where AI actually helps, and where it hallucinates a BOM or misses a DFM constraint that would have cost them in tooling.
What this means for founders building now
If you’re developing a hardware product today, a few things are worth being clear on.
The render is not the product.
AI tools can get you to something that looks finished before you’ve solved a single manufacturing problem. That’s amazing. Use it. However, treat the render as the starting point of the real work, not proof that the work is done.
Intelligence in your product is not optional much longer.
You don’t have to ship an AI-native product on day one. But if your roadmap has no path toward embedded intelligence, you’re designing for today’s market, not the one that exists when you’ve reached scale.
Execution is the differentiation.
The founders who figure out how to move from idea to physical product reliably, without burning capital on avoidable mistakes, are going to win. The ideas are getting commoditized, so it’s not about who develops the better one. It’s about who can deliver.
Use AI as a boost, not a substitute.
Everyone’s saying this. I’m saying it because I believe it. You’re not going to get replaced by robots. You’re going to get replaced by people who use robots better than you.
Learn it. Stay up with it. Use it as a launching pad and get there quicker, cheaper, faster. But it does not solve the problems for you. It probably never will. Don’t use it as a crutch. Use it as an accelerator.
AI is going to democratize invention. It is not going to lend a hand in democratizing operational excellence. That gap is where the real companies get built.
How Morpho works with founders in this environment
The engineering and manufacturing realities haven’t changed. Tooling still needs to be designed for the right process. Components still need to be sourced and controlled. QC still needs to happen at the right points in production, not just at the end.
What’s changed is the pace at which good founders can move through concept development before they get to those gates.
We work with founders at the point where design meets production reality — DFM reviews, prototyping, tooling, mass production runs with our Component Control Method to protect IP and hold quality across supplier relationships.
If you’re at the stage where the idea is real enough to start asking manufacturing questions, book a discovery call. That’s where the conversation actually starts.
Til next month.
Peace,
Ready to start manufacturing your product?
Whether you are already dealing with a failed production run or preparing for your first batch, the rule is the same: do not let the factory’s confidence replace a controlled process.
Book a call with Morpho’s manufacturing experts today. We’ll help you identify where the risk sits, what needs to be validated, and how to move into production without turning your first shipment into the test batch.
What is a Discovery Call?
- A free and no-commitment 30-minute session to assess your product, understand Morpho’s process, and map your next steps.
- We’ll validate if your design is manufacturable and compliant.
- You’ll gain guidance on risks, timelines, and our proven premium-build process.
Book a Discovery Call and learn how Morpho can help you move confidently from idea to production.
FAQ
Will AI for manufacturing make hardware product development faster for founders?
Yes, in specific ways. AI for manufacturing tools accelerate concept design, generative CAD, simulation, and early-stage iteration. A founder today can get to a credible product concept much faster than was possible even three years ago. But AI doesn’t solve tooling, component sourcing, DFM, supply chain coordination, or QC. Those parts still require real engineering judgment and manufacturing experience. AI speeds up the front end of development significantly. The manufacturing execution remains where it’s always been.
What does “intelligent hardware” actually mean for consumer product founders?
It means physical products that process data locally, respond to their environment, and integrate into users’ lives without requiring conscious interaction. Sensor miniaturization, edge AI chips, and better battery technology are making it possible to embed meaningful intelligence into products at price points that weren’t viable a few years ago. For founders building today, the question isn’t whether to add intelligence. It’s how to do it in a way that’s reliable, manufacturable, and doesn’t add unnecessary cost or complexity to the BOM.
How does Morpho approach hardware development differently for AI-enabled products?
The core process is the same — DFM, prototyping, tooling, controlled mass production, QC — but AI-enabled products introduce additional considerations at the component level. PCBs, sensor packages, wireless modules, and edge compute chips have their own sourcing dynamics, lead times, and quality verification requirements. Morpho’s Component Control Method (CCM) handles these components separately, sourced from specialized suppliers, with individual QC before assembly. That’s not a workaround for complexity — it’s how you protect IP and hold quality when the product has multiple technically sensitive components.
Can a small team actually compete with larger companies in hardware now?
More than ever, yes. The combination of AI-assisted design tools, rapid prototyping, and a strong manufacturing partner compresses timelines that used to require large, specialized teams. What used to take a hundred-person team a year can now be done by a small founding team working with the right partners. The constraint is execution quality. A small team that understands the manufacturing process and moves fast through iteration can outrun a slower, larger competitor.
When should a product founder start talking to a manufacturer?
Earlier than most do. The most expensive mistakes in manufacturing happen when constraints are discovered after tooling decisions are already made. Getting a DFM review before you’ve committed to a tool design can save months and real capital. Morpho’s discovery call is the starting point. A conversation about your product, where it is in development, and what the realistic path to production looks like.
Author & Last updated
Josh Fairbairn is the CEO and founder of Morpho, a product design and contract manufacturing company in Hong Kong working across mainland China, and official Kickstarter Partner Program member.
- Last updated:
Table of Contents
