As I reflect on 2025, I think about the 2 characters that changed my entire mental model of the world: O3.

Released only 8 months ago, O3 is OpenAI’s first reasoning model that can produce genuinely useful legal work. ArcPrime's features, from claim charts to continuation recommendations, only work because of reasoning models. They are what give folks “tingles” on the future of the profession.

I felt "tingles" in 2019 when OpenAI beat the world’s best Dota 2 players. Any serious Dota player knows that requires serious intelligence because winning relies heavily on timely decision-making with imperfect information. I once spent months trying to improve and only made it to the top 10%. It only took OpenAI 6 years to translate that game intelligence into a product that produced useful legal work.

If trends continue, it seems inevitable the best lawyer in the world will be AI in the very near future. I don’t know whether that means fewer lawyers or different jobs as lawyers. But it does change the question. The question is no longer whether AI will become competent. It’s: what does great work look like when competence is abundant?

Great work can be subjective — pushing beyond what you thought you could personally achieve — or it can objectively benefit other humans. But either way, in a world with intelligent AI, I think self-expression becomes even more important than it is today: the ability to form an opinion about what the work should be and make it real.

And self-expression requires confidence. You’ll need confidence to express yourself in the work, form opinions about the shape of the work, and take responsibility for it. This is hard to do if you don’t trust your own thinking.

The most reliable way I’ve found to stay confident is to stay close with the details. Everyone starts at 0. You do the detailed work, and at some point, you become an an expert. Typically, your title grows, you get paid more, and you’re encouraged to delegate more.

I now think that is a critical moment. As you become senior, it’s easy to delegate everything. But if you do, you stop updating your sense of what's possible. Especially today, fields change quickly.

The antidote is to occasionally dive in and “feel the air”. This lets you apply new technology or perspectives, and you then see how boundaries can move. This likely compounds to more great work (and more confidence to further express yourself).

I’ve been told to uplevel, delegate, and manage. That’s necessary. But the way I’ve found to scale without losing confidence in my judgment is to reserve 10–20% of my time to stay in the craft. I’ll draft patent work product, ship a feature, or run a marketing campaign to “feel the air”. The hands-on pass reveals where workflows can be tightened and where issues are quietly emerging. Then I can more confidently delegate.