Imagine a new habitable planet suddenly came within reach: a vast world, rich with possibilities, with extraordinary value waiting to be discovered. Passage is available, but you have to recognize the opportunity, learn to pilot a new kind of vehicle, and venture beyond the landing site.
That is what agentic AI has opened up in the space of ideas.
We can ask questions, analyze data, build prototypes, and run experiments that were previously too expensive or time-consuming to attempt. Things that once required teams, specialized expertise, or months of work can increasingly be explored by a curious person willing to learn. Entire regions of possibility have become accessible.
When the boundaries of what we can build change, we should question what we have already built. How much of the world do we accept as inevitable simply because changing it once seemed too expensive?
Calling for a rewrite is not a demand to discard everything. It is an invitation to reconsider everything we learned to tolerate: the cumbersome tool, the missing feature, the workflow nobody has time to fix, the idea that never got past “who would build it?” That includes the governance models and fundamental architectures we inherited as computers, and the internet emerged, often without ever choosing their terms for ourselves.
Old software embeds hard-earned knowledge. Preserve the lessons, requirements, and edge cases. But preserving what we learned does not require preserving every implementation through which we learned it.
And a bad first draft of AI-generated code is not the boundary of this new world. Spend the tokens on rigor. Demand clear requirements, maintainable designs, automated tests, security checks, benchmarks, and repeated review. Do not merely ask an agent to build. Require it to demonstrate that what it built works.
Where you lack expertise, use the same tool to help acquire it. Have it teach you. Have it critique its own work, but do not let its confidence substitute for independent tests and experienced review. Learn from other builders and exchange best practices. The opportunity is to make expertise more accessible, not unnecessary.
For anyone with curiosity and an entrepreneurial heart, this changes what is worth trying. An idea no longer has to justify an enormous investment before you can begin discovering whether it works. Explore, test, revise, discard, and try again.
But rebuilding software means more than changing what it can do. It means reconsidering what it should be allowed to do, whose permission it needs, and whose interests it serves.
That includes how our personal information is leveraged to create intelligence, personalize our experiences, and generate economic value.
Having someone’s information should not imply permission for every conceivable use. Agents built to serve us should not treat another person’s privacy, consent, or dignity as worth less than our own.
That is why I’m helping build wetheusers.ai. Join us in writing the principles and building the tools for an internet that respects our privacy, our agency, and the value we create together.
Who gets a say? Who controls personal information? Who shares in the rewards? Cheaper tools won’t answer those questions for us. We should answer them together, not simply inherit arrangements.
This is the golden age of building. Not because every constraint has vanished, but because the distance between imagining something and bringing it into the world is shrinking dramatically.
The opportunity is not merely to build faster. It is to attempt what we would never have attempted before.
So learn the new vehicle. Leave the landing site. Explore.
The world is ready for a rewrite. Let’s build what we once could only imagine, on terms we choose together.
Originally published on X on September 18, 2026.