May 30, 2026 May 30, 2026 2026-05-30

In January 2010 I wrote a post asking whether you were running the best version of yourself, built around a software metaphor — defrag your mind, update your drivers, stay virus-free, embrace occasional shutdown. The metaphor was meant to make the self-maintenance idea more concrete, a verifiable parallel with computer maintenance.

Sixteen years later I think the metaphor stopped being a metaphor – it is happening.

AI tools now literally do some of what the 2010 post described as internal work. They patch cognitive weak spots. They handle tasks that used to require specific human skills. They run in the background while you focus on something else, just like a computer daemon.

The question “are you running the best version of yourself?” has gotten 10x more interesting — and more complicated — because the definition of “yourself” is blurring.

Which of the Five Actually Stuck

The 2010 post offered five upgrade strategies. Sixteen years of testing them gives me a usable data set.

The defrag — clearing cognitive clutter through meditation or GTD — is the one I’ve practiced most consistently and gotten the most from. Not because it works the way the post implied (it’s not a one-time cleanup, it’s ongoing maintenance), but because the discipline of regular mental clearing is itself the upgrade. You don’t defrag once and run clean forever. You build the habit of noticing when things are cluttered and doing something about it. It’s the equivalent of taking the trash out every single evening.

The shutdown — releasing control periodically — took me the longest to actually practice. The post described it well theoretically. Doing it consistently required accepting that the work doesn’t stop because I stop, and the world doesn’t end because I’m unavailable. That took years, not months, or weeks.

The virus-free section has aged the most interestingly. In 2010 the threat was bad influences and groupthink. In 2026 the equivalent is algorithmic content curation and AI-generated persuasion at scale — systems specifically optimized to install beliefs and preferences without your awareness. The threat vector is the same (external manipulation of your mental operating environment) but the delivery mechanism is orders of magnitude more efficient. Staying virus-free in 2026 requires more active maintenance than it did in 2010, not less.

Where the Self Ends

The genuinely new question the 2010 post couldn’t have asked: when you’re using AI tools as cognitive extensions — for drafting, for research, for pattern recognition, for tasks that used to live entirely inside your head — where does the self end and the extension begin?

I’ve built ten iOS apps in four months with significant AI assistance. The ideas were mine. The judgment calls were mine. The final decisions about what to build, what to cut, what mattered — mine. But large portions of the execution happened through a collaboration with a tool that didn’t exist in 2010. Is the result “me at my best” or “me plus a capable assistant”?

I think this, as popular and intriguing as it is these days, is a false distinction. Humans have always extended themselves through tools — writing, calculators, search engines. The self isn’t what you can do unaided; it’s what you choose, what you direct, what you take responsibility for. The extension is part of the system. The question isn’t whether you’re using AI — it’s whether you’re using it in a way that amplifies your actual judgment or substitutes for it. If it substitutes, by the way, I think you already surrendered and you’re running your life on auto-pilot.

Undocumented Features, Revisited

The 2010 post closed with the concept of “undocumented features” — the unique potential that makes each person genuinely irreplaceable. In a world where AI can replicate many outputs, the question of what’s actually undocumented and irreplaceable has suddenly claimed the top spot.

My honest answer: it’s not any specific skill or output. It’s the particular combination of experience, judgment, and values that makes someone’s choices distinctively theirs. AI can produce content that looks like mine. It can’t produce the specific sequence of lived experience — communist Romania, two decades of blogging, crypto cycles, learning and teaching tango, ultra running, parenting at fifty — that informs why I make the choices I make. That particular combination doesn’t exist anywhere else.

The best version of yourself in 2026 is the version that knows what its undocumented features are and is actually running them. It’s the one that produces bio content, not AI-generated slop. It is not outsourcing your actions, it is not suppressing your features, it is not waiting until the conditions are better.

It’s just using its extensions well, and always knows which parts of the work only you can do.