Code Is Cheap Now: What Survives When Anyone Can Clone Your Software

October 25, 2026 · 13 min read

The weekend someone cloned the tool they were paying for

Here's a pattern I keep seeing. An indie hacker pays for a SaaS tool for a year or two, gets annoyed at the bill, opens Claude, and rebuilds the thing over a weekend. A week later it's live on a landing page, priced at a fraction of the original, and people are buying it.

I've spent about three years in SEO, and I'm now moving into building products myself. So I watched this with excitement and a knot in my stomach at the same time. If one person can rebuild in two evenings what a company spent years on, what exactly was that company selling? And, more uncomfortably, what am I about to sell?

This article is my attempt to answer that with data and history instead of vibes. The short version: code just became cheap, and that is the best and the worst news a software builder can get. Stay with me, because the history is the part that made me less scared.

The clone craze is real, and it's cheap

Developer Martin Alderson cloned most of Linear's core features in about 20 prompts with Claude Code. He did it from network traces alone, with no source code, over a couple of evenings. He's honest that it was incomplete, but that's not the point. The point is how little attention it took.

It's not just hobby projects. A 55-person real estate firm, Greenleaf Management, reportedly replaced Salesforce with a custom app and saved about $100,000 a year, because the app costs roughly $3,600 a year. And Retool's 2026 build-vs-buy report found that 35% of teams have already replaced at least one SaaS tool with a custom build. (You'll often see that number credited to Deloitte. It isn't.)

Even the people living inside this world are nervous. Pieter Levels, one of the best-known indie makers, argues indie hackers may be the first developers to go extinct. His logic: execution now costs about $20 a month, so anyone can build their own tool instead of buying yours, and building in public makes your idea easy to copy.

And it has reached big names. A four-person team used Claude and Rust to build subscription-free alternatives to Photoshop, Premiere and Excel. If it can happen to Adobe's lineup, nobody's product is automatically safe.

A real example: Marc Lou and his newsletter bill

The cleanest version of this story I've found belongs to Marc Lou, a well-known indie maker. By his own account, he paid beehiiv for two years while prices kept rising and it added no feature he used. When he saw how much he was paying, he built his own newsletter platform, MailCheap, and says it costs 83% less.

To be fair, this wasn't a feature-for-feature copy. He built what he needed, and the difference is where the lessons are.

The pricing flip. MailCheap charges for the emails you send, not for dormant subscribers. At 4,000 subscribers, it claims to be 83% cheaper than beehiiv and 75% cheaper than Mailchimp. One third-party review compares its $19 starter plan to beehiiv Lite at $109 in that scenario. These are mostly self-reported or from comparison sites, so treat the exact percentages as marketing until you test them.

Lowering the exit door. The project is open source, so you can self-host it or use the hosted version. It imports exports from beehiiv, Substack and Mailchimp as they are, which attacks the old SaaS moat of "nobody wants to migrate." It also ships a REST API and MCP server so AI agents can run the newsletter.

What it doesn't have. According to the same review, MailCheap lacks native paid subscriptions and beehiiv's built-in referral and ad ecosystem. A competing comparison site notes that a newsletter earning even $100 a month from beehiiv ads covers the price difference. For a newsletter that makes money, the "expensive" platform can pay for itself.

What I take from it:

He had distribution. A known maker with an audience launches with eyes on him. A stranger's clone starts at zero.
Pricing was the product. Paying for usage instead of list size is the same shift from seats to outcomes we saw in the numbers.
He copied the cost center, not the network. Ad networks and referrals get stronger with more users, and no weekend build replicates that.
His clone is already being cloned. Open-source forks of MailCheap exist, and his GitHub profile appears to show about 1.4k stars and 362 forks. If your product is easy to copy, publish it openly and compete on hosting, support and trust.

We've been here before (and it matters how it went)

Around 370 BC, Socrates told a story about the Egyptian god who invented writing. The king refused the gift, because writing would weaken memory and give people the look of wisdom without the real thing. Sound familiar? Every new tool that does some of our thinking triggers the same fear.

The pattern repeats, but the endings are not identical. That difference is the whole story.

Tool	The fear	What actually happened
Writing (~370 BC)	Memory would rot	Knowledge could travel and last, though the worry about shallow understanding never fully went away
Power looms (1811-13)	Weavers lose their livelihood	Luddites mostly fought how machines were used to cut wages, and the state sent about 14,000 soldiers against them
Calculators (1970s-80s)	Kids would forget math	72% of surveyed teachers and laypeople opposed calculators for seventh graders, yet by 1980 the main teachers' body endorsed them
Compilers (1957)	Hand-tuned assembly beats machine code	Programmer demand exploded, because far more people could program
ATMs (1970s-2010)	Bank tellers vanish	Teller jobs rose from about 500,000 to nearly 600,000 before mobile banking finally cut them
Spreadsheets (1979)	Accountants replaced	About 400,000 fewer clerks, but 600,000 more accountants

Two details I love. The famous 1986 calculator protest was real, but only about 15 of roughly 6,500 attendees joined it. And ATMs didn't kill tellers directly: each branch needed fewer, from 20 to 13, but banks opened so many more branches that total jobs went up.

The pattern is that cheap tools shift the scarce thing. Tellers moved from counting cash to selling and advising. Clerks became analysts. Assembly specialists moved up a level of abstraction.

But here's the part nobody likes to say. The Industrial Revolution had a long, painful middle. Economist Robert Allen's "Engels' Pause" describes decades when output per worker grew while real wages stayed flat. Historians still argue about how bad it was, and what eventually helped workers was institutions like factory laws and unions, not the machines alone. Tellers got decades to adjust. We may not.

What the numbers say (and what they don't)

The panic was real. In early 2026 the S&P 500's software and services index lost around $1 trillion in a few weeks, a selloff Jefferies traders nicknamed the "SaaSpocalypse." One trigger was a set of Claude Cowork plugins from Anthropic, the company that makes Claude. (I'm using Claude to build, so I'm not a neutral party here.)

But a panic is not a forecast. Here are the figures I'd actually trust, with their limits.

Number	What it tells us	Where it's from
$234 billion, about 20% of enterprise SaaS spend by 2030	Money exposed to agents doing the work instead of people clicking seats	Gartner
Over 40% of agentic AI projects canceled by 2027	The same firm warns plenty of AI projects will fail	Gartner, via a summary
At least five years to replace dashboards	2026 is early experimentation, not extinction	Deloitte, via a summary
92.6% of developers use AI coding tools monthly	Adoption is almost universal	DX survey, via an analysis
19% slower, while believing 20% faster	In a randomized trial, experienced open-source developers misjudged AI's effect on their own speed	METR study, via an analysis
40% of enterprise SaaS spend on usage- or outcome-based pricing by 2030	Seat pricing is under pressure	Gartner, via Deloitte

Read those together. The money at risk is large but not total, adoption is huge but gains are mixed, and even Gartner expects a lot of failures. "SaaS is dead" and "SaaS is fine" are both lazy takes.

The pricing shift is the quiet earthquake. When an AI agent does the work, charging per human login stops making sense. Zendesk now lists $1.50 per automated resolution and Intercom charges $0.99 per outcome. A survey of 300 SaaS CEOs found 97% plan to retire seat-based pricing within two years, yet 94% say seats still match their product's value. That contradiction tells you how unsettled this is.

And the coding side? Here's my honest read: the tools clearly help with some tasks, the hype about 10x is not supported by the best studies, and the skill of reading and judging code matters more, not less. That's why I still think learning to code is worth it. Not for typing speed, but for knowing when the AI is wrong.

Why your weekend clone probably won't save you

Building a clone for yourself is smart. Selling one is a different game, and this is where I'd slow down.

1. Your clone gets cloned. One analysis asks the question every clone seller should fear: why would anyone buy yours instead of the seven clones that exist by month three? It adds that churn rises because switching just got easier. Another essay puts it well: copying software is cheap now, and making it good isn't.

2. Security can sink you. Your customers' data is now your problem. A researcher found that 170 of 1,645 Lovable apps scanned (about 10%) had exposed databases, tied to a missing database permission rule. The vibe-coded social network Moltbook was reportedly breached within three days of launch, exposing 1.5 million API tokens. And Escape.tech's scan of 5,600 vibe-coded apps found over 2,000 high-impact vulnerabilities. Some of these figures come from security companies that sell scanners, and Lovable has disputed parts of the breach reporting, so treat the percentages as a warning sign, not gospel. The lesson is simple: shipping fast is easy, and shipping safely is the job.

3. Maintenance never ends. One commentator estimates lifetime maintenance runs two to four times the original build cost. I couldn't verify that number independently, but anyone who has run a paid product knows the feeling. The customer finds the bug before you do.

4. The legal and ethical gray zone. I'm not a lawyer, so take this as a heads-up, not advice. Commentators generally note you can't copyright an idea or a feature, only its specific implementation. Copying a product's look and feel is murkier, and even a clone that is legal can damage your credibility if people think you lifted someone's work. Also check the terms of service of the tool you studied.

So what does a clone seller actually have? Only what the code can't copy: support, trust and a clear reason to exist.

Open source is the canary in the coal mine

If you want to see where all this leads, look at open source. It's been hit twice, from two directions.

The business model got stress-tested. Tailwind CSS is one of the most popular CSS frameworks, with roughly 75 million downloads a month. In January 2026, its company laid off 75% of its engineering team. The founder said documentation traffic was down about 40% since early 2023 and revenue down roughly 80%, because AI tools now write Tailwind code so developers never visit the docs where the paid products were sold.

Was AI the killer? Drupal founder Dries Buytaert says AI didn't kill Tailwind's business, it stress-tested it, and AI commoditizes anything you can fully specify. Documentation, pre-built components and templates fall into that bucket.

The contribution model got flooded. Maintainers are drowning in low-quality AI-generated pull requests. The drawing-app project tldraw began auto-closing all external pull requests, and GitHub has said it is considering restricting or even disabling pull requests. The curl maintainer shut down his bug bounty over a flood of AI-generated reports.

There's an irony, though. The same tools also find real problems. Anthropic reports using Claude to find 500+ high-severity vulnerabilities in open source code, and AI companies put $12.5 million into open source security to help maintainers keep up. The tools create the burden and fund the cleanup.

So how does open source make money now? The recurring answer is by selling what code can't: managed hosting, compliance proof, support and operations. As one open source company put it, code has become a commodity, and what AI cannot generate is trust. Notice how similar that sounds to the clone seller's problem.

So what should we build?

The advice that keeps repeating across investors and founders is blunt: build what a model vendor can't ship in a release note. Thin wrappers get absorbed. One AI writing tool reportedly watched $50K in monthly revenue fall 70% in 60 days after a model update added a similar feature. (Wrapper failure statistics are repeated everywhere, but one analysis notes most trace back to secondary sources, so don't lean on them.)

What tends to last:

Own the workflow. Be embedded in how customers work, ideally the system of record for something critical.
Own live, exclusive data. But be careful: Chegg's content stayed proprietary and AI still delivered the outcome without it.
Own the transaction. One analysis argues that if you only rent the flow as a thin layer on someone else's rails, you're a feature.
Sell trust and accountability. AI lowered the cost of building, not the cost of earning trust.
Go vertical. Redpoint's 2026 update, reported by Crunchbase, has horizontal SaaS multiples down 35% in a year while vertical SaaS is roughly flat.
Be agent-ready. If agents become customers, expose your APIs to them through standards like MCP.

Here's a five-question test I'm using on my own ideas:

Does it still work the day a model company ships my feature for free?
Do I own data, a workflow or a transaction that others can't reach?
Would someone pay for accountability (support, security, someone to call), not just output?
Does it get better with use in a way a copy wouldn't?
Is the price tied to outcomes or usage, not just seats?

If you can't say yes to at least three, you're building a feature, not a business.

The takeaway

Go back to my opening question. If someone can rebuild a product in a weekend, what was the company selling? Turns out, never just the code. It was trust, support, data, distribution, and the promise that someone is on the hook when it breaks.

History says the tools that make a skill cheap don't erase the work. They move the value somewhere new, and the people who move with it do fine. History also says the transition can be rough, and it rewards those who adapt early.

So clone freely for yourself. Cancel the subscription you hate. But if you plan to sell, ask the uncomfortable question first: what do I have that a copy of my code can't give someone?

I'd love to hear your take. Have you replaced a paid tool with something you built? Did it hold up? Tell me in the comments.

Sources

Figures below come from search results and the pages linked in the text.