Where We Stand on AI

Our site leads with three lines now: Frontier technology. Applied intelligence. Human-centered. They are new. Most of what they stand for is not.

What’s changed is what’s possible. Over the last couple of years, what we’re able to build has expanded dramatically. Work that was out of reach not long ago is now well within it, and the range of challenges and opportunities we can take on has grown with it. When that much more becomes possible, the harder question is no longer whether something can be built, but whether it’s worth building. The three lines are how we describe where we put our effort now.

Frontier technology #

We have always worked with the most capable tools available; that part isn’t the change. “Frontier” points at the leading edge of the field, a moving target rather than a place you arrive, which is exactly the point. What’s different is how much those tools can now do, and how fast that keeps growing.

AI is clearly a powerful technology, and the world’s attention on it is warranted.

AI, like any tool, takes real skill to get the most out of, and we’ve been building that skill for more than two years now. You can watch our thinking on it evolve across the posts on our blog.

It’s worth meeting the skepticism head-on, because we hear it often: isn’t this a bubble? If what we mean is the financial question, which companies win, who the winners and losers turn out to be, what the eventual economic shape of all this is, then the honest answer is that we have no idea. Any thoughtful person would say the same. The future is far too hard to predict from here.

But the market and the technology are two different things. Part of why the skepticism lingers is that, for many, AI still means interacting with ChatGPT and not much beyond it. If that is your experience, then given the hallucinations, the overconfidence, and the weaknesses still visible in the models as of May 2026, it is reasonable to conclude they are not all that powerful. The real power, though, is in the agentic software: systems you give a goal and the room to pursue it, that work in loops and steps, check their own output, and finish the job rather than hand back a reply.

As practitioners who have followed this technology closely for the last few years, we have watched it revolutionize our own company, and the organizations of the clients we’ve had the good fortune to use it with so far. We are 100% certain this is game-changing for the entire world economy, in its ability to take on the thoughtful, judgment-heavy work, the decision intelligence, that was relegated to human minds alone until now. We want to state plainly how certain we are of the impact it has already had, and of the impact still to come. The forces behind it are enormous: humanity’s drive to create and to invent, the profit waiting for whoever gets it right, the sheer power, geopolitical and otherwise, in figuring these things out first. None of these forces is going to relent. They will keep driving this technology forward to the very limit of what any of us are capable of.

It’s tempting to believe there is some special human reserve the machines can’t reach, that real understanding is ours alone. We doubt it, and so do most of the people who built this technology. We take that argument up directly, in their words, in AI Is Not a Bubble.

Applied intelligence #

Research and evidence have always been how we work; “applied intelligence” is the name we’ve put to it. It is also the name of our newsletter.

It matters more now because these tools fail in a particular way. They are fluent and confident whether or not they’re right, and their output tends to look finished while being, on occasion, quietly wrong in ways that take real expertise to catch. The scarce skill is no longer producing a draft but judging one, and being willing to throw it out. That’s less impressive to watch than fast output, and considerably more valuable.

It also moves where the consequential work sits. As more of the building gets handled by the tools, the decisions that decide whether a project succeeds shift earlier: into strategy, discovery, and the framing of a problem before anyone writes a line of code.

Human-centered #

Human-centered means we care about the outcomes for the people our work touches. That has always been true of us, and it matters more now, not less. As the technology becomes able to take on, and take over, more of what people used to do, keeping the human stakes in view is part of the work, not a footnote to it. We build with these tools, and we stay deliberate about how and why, and about who they affect.

The same is true of our own team. We have always hired people we want to be around. Laughter sat close to the center of how we described ourselves for years, and while we’ve since gathered that into “delight,” the intent hasn’t moved: a team that genuinely enjoys each other’s company. The thing clients have told us for years, more than anything about the work itself, is that we were their people. We intend to stay that.

What does it all mean? #

The technology is moving faster than the institutions meant to govern it, and we’re writing about that as well.

Make no mistake: this is game-changing, unprecedented technology, and to treat it as anything less is to do so at your own peril. Like everyone, we cannot tell you where all this rapid change will lead. What we can tell you is this: we are getting an enormous amount of joy and satisfaction out of what we are now able to build, and out of the value it lets us bring our clients, whenever we find the time to do it. We take the societal impact seriously too, and we do our best to be engaged, positive contributors to the extent we can, even where there are no simple answers.

As practitioners, we’re invigorated by it: the chance to build past what was possible before, and the plain fun of doing it. As thoughtful citizens, we feel the obligation that comes with it, to help push toward the most beneficial future we can. Both are real for us, and we’d rather work out the how alongside you than pretend we have it figured.