About two weeks ago, the U.S. government directed Anthropic to suspend access to its latest frontier models. This week, OpenAI was required to limit the rollout of GPT-5.6 over cybersecurity concerns.
Whether you agree with those decisions or not isn't the point.
The point is that they highlighted something many of us have quietly ignored for years.
If your software depends entirely on somebody else's API, then part of your software is no longer under your control. Your application can be rate-limited. Features can disappear overnight. Pricing can change. Access can be restricted by policy, regulation, geography, or business decisions you have no say in.
That's not ownership. That's renting intelligence.
For years, local-first development was treated like a niche philosophy. People associated it with privacy enthusiasts, hobbyists, or developers who simply didn't want to manage servers. Today it feels less like nostalgia and more like common sense.
The conversation is no longer just about privacy.
It's about availability.
It's about resilience.
It's about building software that continues to work even when the internet, an API, or a provider decides otherwise.
This doesn't mean cloud computing is dead. Far from it.
The cloud will remain an important part of our industry.
But we've reached the point where depending on it for everything is becoming an architectural risk, not just a technical decision.
I don't think we're witnessing the end of the cloud.
I think we're being reminded of the value of a parallel ecosystem.
One where intelligence isn't rented—it runs locally.
One where software doesn't ask permission to keep working.
One where the browser is treated as a capable computing platform, not just a window into someone else's server.
No recurring API bills.
No mandatory internet connection.
No unnecessary data leaving the device.
We don't need billion-dollar server farms to solve every problem.
Sometimes a browser tab, local compute, and well-designed software are enough.
For the past year, that's the direction I've been exploring: treating the browser less like a document viewer and more like a sovereign operating system capable of running meaningful software entirely on the client. Maybe the future isn't cloud-first.
Maybe it's cloud-optional.
Because if recent events have shown us anything, it's this:
The more your software depends on someone else's infrastructure, the less independent it really is.
There's nothing as soothing as having software that works whether there's an internet connection or not. 😎