What I actually believe about technology, intelligence, and the work worth doing. Opinions held firmly but revised when the evidence demands it.
Not nations, not companies, not ideologies — humanity. The full, flawed, extraordinary sweep of eight billion people navigating existence on a pale blue dot with no instruction manual. Every technology worth building serves this project. Every system worth designing makes human lives more dignified, more capable, more free. When I lose the thread of why the work matters, I come back to this: the point has always been the people. Everything else is scaffolding.
Build for humans. All of them. Not just the ones who can afford it.
Einstein's insight that space and time are not separate fixed stages but a single interwoven fabric — curved by mass, dilated by velocity — is still not fully absorbed by most people who know the words. Time passes slower near a massive object. GPS satellites require relativistic corrections or they drift by kilometers per day. This is not abstract physics. It is the geometry of reality, and it is stranger than anything we invented. The universe does not owe us intuition.
Gravity is not a force. It is the shape of spacetime. Let that land.
Every system I have ever worked on that became unmaintainable did so because someone confused complexity with capability. The most powerful architectures are the ones a new engineer can understand in an afternoon. Elegance is not aesthetic — it is functional. A system that only its creator understands has a single point of failure: its creator.
The right abstraction does more work with less surface area.
The conversation about AI is still stuck on chat interfaces and question-answering. That is the wrong frame. Agentic systems — models that observe, plan, act, and reflect across multi-step tasks — represent a qualitative shift in what software can do. The unit of work is no longer the query. It is the goal. Organizations that internalize this early will build things that are simply impossible for those that do not.
We are not building smarter search. We are building systems that work.
This is not a metaphor. Architects who think in Terraform think differently about infrastructure than those who think in console clicks. Engineers who write in Rust think differently about memory than those who write in Python. The models, frameworks, and interfaces we use daily are not neutral — they carve grooves in how we approach problems. Choose your tools with the same care you'd bring to choosing your mental models.
You become what you repeatedly use.
A demo that works 90% of the time is impressive. A production system that works 90% of the time is broken. The hard 10% — the edge cases, the adversarial inputs, the cascading failures, the midnight pages — is where real engineering lives. I have seen too many "AI solutions" that are, under examination, carefully curated demos. The gap between a compelling prototype and a reliable system is wider than anyone admits in a sales meeting.
Ship to learn. But be honest about what you shipped.
Not because superintelligence is imminent, but because the systems we are building today are already consequential. A prior authorization model that denies care incorrectly, an NLP classifier that encodes historical bias, a recommendation system that optimizes for engagement over wellbeing — these are alignment failures. The question is not whether to care about alignment. It is whether you are paying close enough attention to realize you already need to.
Safety is not a constraint on capability. It is a precondition for trust.
There are more stars in the observable universe than grains of sand on every beach on Earth. The light reaching our eyes from the Andromeda Galaxy left before our species existed. Every heavy element in your body was forged in a stellar explosion. On the scale of cosmic time, the anxieties that fill human lives are the briefest flicker of disturbance in an otherwise unimaginable stillness. This is not nihilism. It is perspective — the most useful mental tool I carry.
Look up more often. The problems shrink.
Choosing not to understand something — a technology, a culture, a perspective, a system — is a choice with consequences. Intellectual incuriosity is not a neutral stance. It concentrates the future in the hands of those who did bother to understand. Twenty years of driving this country has been, at its core, an exercise in refusing to let my mental model of America be secondhand. The road is a commitment to primary sources.
Go see for yourself. Form your own view. Then update it.
Any organization whose AI strategy is entirely dependent on a single vendor is one pricing change, one outage, or one policy shift away from a crisis. Platform independence is not about distrust — it is about negotiating leverage, resilience, and the freedom to use the best tool for each job. The abstraction layer is not technical overhead. It is strategic optionality.
Never let a vendor make a decision for you by default.
The incentive structure of most technology companies optimizes for engagement, retention, and revenue — measured in quarters. The technologies we are building will shape how billions of people think, work, and understand the world — measured in generations. These time horizons are not aligned. The engineers and architects building these systems have more moral responsibility than the current incentive structures acknowledge. Act accordingly.
Build as if the person who inherits your system will have to live in the world it created.
Architecture Decision Records exist because the second most common failure mode in software (after the initial design being wrong) is forgetting why you made the decisions you made and reversing them for the same reasons they were made. Undocumented decisions are decisions that will be relitigated. Write them down. Note the context, the options, the choice, the consequences. Future you is a stranger. Leave a note.
The decision costs 30 minutes to write. The lack of it costs months to recover from.
A great novel does something no other medium can: it places you inside another consciousness so completely that their memories feel like yours, their grief lands like yours, their joy is indistinguishable from your own. Code transfers logic. Music transfers feeling. Literature transfers being. The empathy, historical range, and moral imagination it builds are not soft skills — they are the difference between an engineer who ships features and one who understands what the features are for. Read widely, read slowly, read things that are nothing to do with your field.
Every hour spent reading serious literature is an investment in the quality of your thinking.
The reason a Wi-Fi card made in South Korea works with a router made in Finland is because thousands of engineers sat in committee rooms and argued, at great length, about packet formats and timing tolerances and error correction schemes — and then wrote it all down. Standards are not bureaucracy. They are the accumulated interoperability of the world. Every time you send a packet, stream a video, or charge a device, you are standing on the work of standardization bodies most engineers have never heard of. IEEE, IETF, ISO, W3C — these are the unglamorous foundations that make everything else possible.
The boring work of standardization is some of the most consequential engineering ever done.
"Build things that last, question things that don't, and stay curious long enough to tell the difference."