The open core business model is in trouble. Not theoretically. Right now.
Anthropic purchased Bun and rewrote it from Zig to Rust, file by file, using AI-assisted tooling, in a timeframe that would have been laughable to attempt with human engineering alone. The project is large. The languages are genuinely different. It worked. That is the proof of concept everyone in open source should be paying attention to.
What just became possible at Anthropic's scale is now possible at yours.
The Moat You Thought You Had Is Gone
MIT and Apache licenses were always a trade. You open your code, the community contributes, and in return you get distribution, credibility, and a head start that any competitor would need serious resources to close. The bet was that the cost of forking and maintaining a complex codebase was prohibitively high for most actors. You could monetize the enterprise wrapper, the consulting, the managed service, the support contract, because building your own version of your thing was just too expensive for most companies to justify.
That bet is now off.
Think about what AWS did to Elastic. It was controversial, it was aggressive, and it required a hyperscaler's engineering capacity to pull off. Elastic responded, changed its license, sparked an industry conversation about what open source even means in a cloud world. The whole saga felt like something only a company with hundreds of engineers could attempt. That was the operating assumption.
The Anthropic-Bun rewrite dissolves that assumption entirely. The question is no longer whether a competitor has the resources to fork your project. The question is whether they have a reason to. And increasingly, the answer is yes.
If you are building on an open source commercial model where you give the project away and charge for the enterprise layer, the gravitational pull just shifted. Your enterprise customer now has a credible alternative to your paid tier: build it themselves, in the language they prefer, with the dependencies they want, hosted the way they want it. The engineering cost of that decision just dropped by an order of magnitude.
What This Means Depending on Which Side of the Table You Sit
This is a two-sided shift, and the two sides have almost nothing in common.
If you are running a company built on the open source commercial playbook, the upsell just got harder. Your managed service, your premium support, your enterprise features: those were priced against the cost of self-hosting your complex system. That cost is falling. You should be thinking hard about what you are actually selling, because "the managed version of the thing you could now build yourself in a sprint" is not a durable value proposition.
If you are a solo founder building something new, this is close to a gift. The tools that used to require a full engineering team to assemble from open source primitives are now reassembleable by one person with clear intent and the right AI tooling. You can take a battle-tested backend service, one that has supported production systems for years across the industry, and make a version of it that fits your stack, your architecture, your constraints. Off the shelf still, but yours.
The solo builder just got access to leverage that used to belong exclusively to well-funded teams. You do not need to buy the SaaS subscription for the thing you can now fork, modify, and own outright. That changes the build vs. buy calculation in ways the SaaS pricing models have not caught up to yet.
The Definition of "Derivative" Is About to Cause Problems
There is a legal dimension here that nobody has fully stress-tested yet. GPL licenses, in particular, carry obligations around derivative works. If an LLM rewrites a project file by file into a different language, is the output a derivative? The clean-room proof of concepts that already exist suggest people are thinking about this. But thinking about it and having case law around it are very different things.
MIT and Apache are more permissive and will be the easier targets first. But the GPL question is real, and it is coming. The rewrite-by-LLM approach does not obviously map onto the legal frameworks that were written when "porting" meant a team of humans painstakingly translating logic by hand. Courts and regulators are going to have to catch up, and in the meantime, founders using this approach are operating in genuine uncertainty.
That uncertainty cuts both ways. For the open source companies trying to defend their moat with a GPL license, the protection may be less solid than they think. For the founders doing the forking, the legal risk is not zero.
Where This Lands
The Anthropic-Bun rewrite is not an isolated stunt. It is a demonstration that the engineering cost of building on top of, or diverging from, an established open source project is now within reach of a small, motivated team. That changes the competitive dynamics of the open source commercial model in a way that a license change alone will not fix.
The founders who built companies around open source and consulting revenue need to be honest with themselves about what that model looks like in three years. The founders who have been told they need to buy the enterprise SaaS because building the alternative is too expensive should revisit that assumption.
The gates were always theoretically open. Now they are actually open. That is a very different situation.