TOGETHER WITH REFORGE BUILD
If you're a founder with a new project, use one of the many AI app builders. They're perfect for when you are starting from scratch.
But if you're a PM you already have a product and need prototyping built for product teams. That's what Reforge Build does.
Generate prototypes that look like your product
AI that knows your customers, product, and strategy
Explore multiple variants easily
Team collaboration built-in
Validate with customers quickly (coming soon)
Please support our sponsors!
GEEK OUT
The DIY Threat
For the last twenty years, the line between software companies and their customers was clear.
We built the software, and customers bought it.
They bought it because building it themselves was too hard.
To build even a simple workflow tool, you needed engineers, you needed a product manager, and you needed time.
Most companies didn't have those things to spare.
That wall is crumbling.
And it’s crumbling because the cost of producing code is dropping toward zero.
Now your customers can build a lot of the stuff you used to sell them.
This isn't entirely new.
We saw a preview of this with no-code tools a few years ago. But no-code tools still required you to learn a tool.
You had to learn how to drag the blocks and wire the logic.
It was easier than learning to code, but it still required a builder mindset.
AI is different.
The interface for AI is natural language.
And that changes the physics of the build vs. buy calculation.
Consider a typical B2B SaaS application.
If you strip away the branding and the sales team, what is it?
Often, it’s just a database wrapper with some specific logic on top.
It takes data from point A, formats it, maybe sends an email or updates a record in database, and shows you a dashboard.
In 2020, if a customer wanted that specific custom workflow, they had two choices: buy a SaaS product that did it, or hire a dev shop to build it for $50,000.
In 2025, they can paste a CSV into an LLM and say, "Write me a python script that takes this data, formats it like X, and uploads it to Y." And ten seconds later, they have the script.
This creates a new and dangerous competitor for software startups: your own customers.
If your product is essentially a utility - if you are selling "glue" or simple logic - you are in trouble.
You’re not just competing with other startups anymore.
You are competing with a customer who says, "I could probably hack that together this afternoon."
Disposable Software
The standard objection to this theory is "maintenance."
I hear this often, especially from folks who most need to acknowledge the shift.
They say, "Sure, a customer can write the code, but they won't want to maintain it. They don't want to deal with API changes, or server updates, or debugging. They will pay us to handle the headache."
This is a dangerous assumption.
It assumes that software in the future will be built the way software was built in the past.
Historically, software was like a building.
You built it, and then you had to maintain it.
If the roof leaked, you had to fix it. If you didn't, the building fell down.
But AI-generated utility software is more like a paper cup than a building.
It’s disposable.
If you generate a script to automate a workflow and it breaks six months later because an API changed, you don't hire an engineer to debug it. You just copy the error message, paste it back into the AI, and say "Fix this." Or you just delete it and ask the AI to write a new one from scratch.
When the cost of creation is near zero, the cost of maintenance matters less, because replacement is cheaper than repair.
Boundary: paper cups are fine for ad‑hoc workflows, one‑user jobs, and non‑regulated data. They’re much less suitable for boiling liquid: regulated data, SLAs, many concurrent users, granular permissions, audit trails, or change control.
Moving Up the Stack
So what happens to the software industry? Does it disappear?
No. But the water level is rising.
If you are a founder, you have to look at your product roadmap and ask a hard question: "Could a smart intern with the latest frontier coding model build this in a week?"
And if it breaks, could they just regenerate the code to fix it?
If the answer is yes, you don't have a product. You have a feature that is about to be commoditised.
To survive, you have to move up the stack.
You have to provide value that is genuinely hard to replicate.
The definition of "hard" is changing, but here are a few things that still seem safe:
Deep Complexity
LLMs are getting better every week in their ability to write working software quickly, but still can't easily architect a system with fifty moving parts that all need to stay in sync with 99.999% reliability. If your product solves a problem that is genuinely complex - like managing global payroll compliance or orchestrating Kubernetes clusters at scale - you are probably safe. The customer can’t script their way out of that.
Proprietary Data
Classic moat, now more important. If your value comes from the code, you are vulnerable. If your value comes from the data that flows through the code, you are stronger. AI can build a Zillow clone in an afternoon, but it can't give you the historical housing prices for every home in America unless it has the data.
Taste and Opinion
AI is a mirror. It gives you the average of what it has seen. It is very good at doing exactly what you ask, but it is bad at telling you what you should do. Software that embodies a strong opinion about how a workflow should work - software that teaches the user how to be better at their job - is harder to replace. You are selling expertise, not just utility. Think Linear’s constraints or Superhuman’s shortcuts
Trust and Compliance
If buyers need audit logs, SOC 2, granular permissions, and data governance, DIY gets harder. If you standardize trust, you stay relevant even when code is cheap.
Embeddedness and Integration Depth
Being the default in a workflow beats DIY. If your product is wired into procurement, SSO, finance, and audits, DIY looks risky.
You Can't Outrun the DIY Trend
Don't fall into the trap of thinking you can just out-ship these user-builders. Velocity is no longer a moat.
Even if your engineering team is the fastest in the world, you cannot beat the feedback loop of a user solving their own problem. You have a communication lag: the user has to explain what they need, and you have to understand it. The user-builder has zero lag. They feel the pain and generate the fix in the same minute.
When everyone uses AI, speed becomes table stakes, not a differentiator. You need velocity to stay alive, but it won't protect you.
The New Default
In the past, only developers were builders.
Everyone else was a consumer.
Now, the middle ground is vanishing.
Your customers are becoming builders.
This is good for the world. It means more problems get solved.
If a marketing manager can build their own tool to analyze campaign data, that’s a win for productivity.
But for a startup founder, it is a call to arms.
The era of easy SaaS is over.
You can’t build a company anymore by spotting a simple gap in the market and filling it with a simple solution.
The friction of building software was your moat.
Now that friction is gone, make the problem itself your moat.
Concrete actions you can take:
1. List the top 5 workflows your product automates. Mark the ones an LLM could replicate in a day.
2. For each marked workflow, ask: what data, trust, or opinionated version would make this non‑trivial?
3. Add audit trails, permissions, SLAs, and schema stability where buyers care about risk.
4. Ship “teach the user” features: templates, guardrails, or an in‑product guide that embodies your taste.
5. Build for AI assembly: clean APIs, stable schemas, great docs, usage‑based pricing that welcomes building around you as a foundation.
Enjoying this content? Subscribe to get every post direct to your inbox!
If you’re already a subscriber, consider upgrading to a VIG Membership to get access to every post in the archive.
BEFORE YOU GO…
Book a 1:1 consultation call with me - I keep a couple of slots open each week for founders and product growth leaders to explore working together and get some free advice along the way.
Sponsor this newsletter - Reach over 7800 founders, leaders and operators working in product and growth at some of the world’s best tech companies including Paypal, Adobe, Canva, Miro, Amplitude, Google, Meta, Tailscale, Twilio and Salesforce.
