We’ve become very good at building things. We’re becoming considerably less interested in whether we should.
I think there’s a particular kind of madness in the way we talk about progress, both in the tech industry and in our larger society. We celebrate disruption, reward speed, and admire people who take enormous risks, challenge convention, and refuse to accept that something can’t be done. We build entire business philosophies around the idea that standing still is tantamount to dying.
“Move fast and break things.”
“Fail fast, fail often.”
“Innovate or die.”
These have become more than Silicon Valley slogans. They’ve become a kind of secular religion for business leaders, investors, and a large portion of the people responsible for deciding what our collective future looks like. And like most religions, there are some inconvenient questions we’re apparently not supposed to ask:
Who gets to decide what’s worth breaking? Who pays for the failure? Who benefits from the innovation? And, perhaps most importantly, what happens when the things we break are things other people depend on?
I’ve been chewing on this for a while, and a conversation at a leadership offsite last week brought a lot of it into sharper focus. We spent a considerable amount of time discussing artificial intelligence, its influence, its potential, and the seemingly unstoppable trajectory of its development. There was a fairly consistent sentiment that AI is coming whether we like it or not, that the benefits will ultimately outweigh the risks, and that our job is to figure out how to adapt rather than question the direction of travel.
I understand the argument. I really do. I’m just not sure I buy it.
Not because I think technology is inherently bad, or because I believe we should stop experimenting, building, or finding better ways to solve difficult problems. And certainly not because I think the world should return to some imagined golden age before computers and the internet. I’ve spent most of my adult life working in technology. I’ve built systems, operated infrastructure, led engineering organizations, and watched technology fundamentally change the way people live and work. I have no particular interest in pretending innovation is the enemy. But I think we’re ignoring the underlying problems.
AI is still very much a technology problem. There are substantial technical challenges around reliability, security, safety, and the ability to understand and govern increasingly capable systems. But those aren’t the only problems, and they may not even be the most consequential ones. AI is also a symptom of something considerably larger: an economic system that rewards growth without adequately accounting for responsibility, a culture that mistakes wealth for wisdom, and a leadership class that seems increasingly comfortable confusing what is possible with what is actually good for us.
AI didn’t create those problems. It’s simply giving them a much larger surface area on which to operate.
Which leads to a much more concerning question.
The people we choose to admire
As a society, we’ve come to treat economic success as a reasonable proxy for moral authority. Whether that association took root through the religious and cultural shifts of the Protestant Reformation, the industrial wealth and paternalism of Carnegie’s Gospel of Wealth, or the economic and political changes of the Reagan/Thatcher era is a subject for another time. The connection has been developing for considerably longer than any of those individual milestones.
But the consequences are difficult to ignore. If someone builds a billion-dollar company, we assume they must know something the rest of us don’t. If they accumulate enough wealth, we start treating their opinions as wisdom. If they disrupt an industry, we call them visionaries. If they challenge established norms, we call them courageous. We’ve built a pantheon of modern heroes out of people who are extraordinarily good at accumulating money, and somewhere along the way, their success became evidence of their judgment, their wealth evidence of their intelligence, and their influence evidence that they deserve to have influence.
We rarely stop to ask whether they’re actually making the world a better place.
There is a distinction between being successful and being worthy of admiration, just as there is a distinction between being extraordinarily capable and being a particularly good steward of that capability. We’ve become very comfortable collapsing those distinctions. The problem isn’t simply that we admire wealthy people. It’s that we’ve allowed economic success to stand in for the legitimacy that should come from proven judgment, accountability, and responsibility to others.
Having the resources and technical capability to build something doesn’t automatically give someone the authority to decide whether it should exist, how it should be deployed, or who should bear the consequences. Yet those distinctions have become increasingly blurred.
The people making decisions about the future of AI aren’t working in a vacuum. They’re operating within a system that rewards market dominance, shareholder returns, competitive advantage, and the ability to get there before someone else does. The incentives aren’t particularly mysterious: build bigger models, spend more money, acquire more customers, capture more market share, and get there first. Figure out the consequences later. And if you happen to create something genuinely transformative along the way, all the better.
The problem is that the consequences aren’t necessarily something you can figure out later. Some may be irreversible, while others affect people who never consented to participate. Many will be borne by people who never see a fraction of the benefits. We’ve arrived at a point where some of the people building systems with potentially civilization-altering consequences openly discuss catastrophic possibilities while continuing to race toward deployment.
Think about that for a moment. We’ve normalized a business environment in which even the possibility of extraordinary harm can become part of an acceptable risk calculation. The disagreement over how likely those outcomes are doesn’t make the underlying question any less important: what responsibility do the people making these decisions have to everyone else?
Imagine applying the same attitude to another industry. Imagine an airline announcing that it intends to deploy a new aircraft design despite unresolved safety concerns because the competitive advantages of getting it into service sooner outweigh the risks. We wouldn’t call that innovation. We’d call it negligence.
The comparison isn’t perfect. Aviation has established safety frameworks and measurable operational risks, while many of the most consequential AI risks remain uncertain and contested. But that’s part of the point. Uncertainty should create an obligation to understand and manage risk, not an automatic justification for ignoring it. Yet somehow, when the product is software, the rules of accountability become considerably more flexible.
The problem with breaking things
There is nothing inherently wrong with experimentation. Failure is a necessary part of learning. Prototyping, testing, challenging assumptions, and being willing to abandon an idea that isn’t working are all valuable practices. I’ve spent much of my career advocating for exactly those things.
But there is an enormous difference between creating a safe environment in which to experiment and creating an environment in which the consequences of experimentation become somebody else’s problem. That distinction seems to have disappeared somewhere between the first Silicon Valley startup and the latest trillion-dollar technology company.
“Move fast and break things” sounds wonderfully entrepreneurial when the things you’re breaking are your own assumptions, prototypes, or business models. It sounds considerably less inspiring when you’re breaking public infrastructure, livelihoods, democratic institutions, or the systems people depend on to get through their day.
And here’s the thing: the people doing the breaking aren’t necessarily the people who experience the consequences. A founder gets a valuation. An investor gets an exit. A company gets market share. Executives get bonuses. Shareholders get returns. The displaced worker gets a severance package, if they’re lucky. The customer gets a degraded service. The public gets to absorb the consequences, and society gets to figure out what to do with the mess.
There’s a term for part of this in economics: externalizing costs. It’s what happens when the people making decisions don’t bear the full costs of those decisions, shifting them instead onto customers, workers, communities, or the public. We’ve just become very good at dressing it up as innovation.
Reality has a rather inconvenient habit of intervening
Consider the recent nationwide grounding of flights by the FAA.
On November 13, a telecommunications problem in the FAA’s air traffic control network resulted in a nationwide ground stop. A routine infrastructure replacement, involving the transition from aging copper communications lines to fiber, was associated with a surge of connection attempts that destabilized the network and disrupted communications and radar feeds across multiple air traffic control facilities.
The FAA grounded flights as a precaution. The disruption affected air traffic operations across the country, and the recovery continued well beyond the initial incident. The precise root cause is still under investigation. The FAA identified excessive connection attempts originating from two locations as the immediate problem and said contractor work involving new routers was among the potential contributing factors. Importantly, this wasn’t a complete loss of communications or radar. Some systems remained operational, allowing controllers and pilots to continue operating safely under established procedures.
If we take the FAA’s initial explanation at face value, this wasn’t an AI failure or some futuristic threat. It was an infrastructure problem involving aging systems, routine modernization, maintenance, and the complicated reality of keeping critical services running while changing the systems underneath them. In other words, it was a very human problem.
And that’s precisely why it’s relevant.
The FAA’s response was, appropriately, to stop. To protect the people who depended on the system, stabilize the infrastructure, and establish that it was safe to resume operations. Nobody stood up and argued that we couldn’t stop the airspace because doing so might compromise our competitive advantage in aviation or travel. Nobody suggested that the benefits of maintaining flight schedules outweighed the risk of continuing to operate with compromised communications.
We recognized that there are circumstances in which continuing forward isn’t courage. It’s recklessness.
Now consider Microsoft’s recent global outage. Outlook, Teams, Copilot, and a range of other Microsoft 365 services experienced widespread disruption, affecting organizations and users around the world. Microsoft attributed the incident to a technical problem involving Exchange Online, in which an earlier network capacity upgrade and a subsequent network change exposed a latent defect that overwhelmed affected systems. The disruption lasted more than 24 hours before Microsoft reported that services had fully recovered.
This occurred against the backdrop of Microsoft’s announcement of a voluntary exit program for up to 60,000 eligible employees in participating organizations, as the company continued making enormous investments in AI infrastructure.
I’m not suggesting the outage was caused by AI, or that the employees who chose to leave were responsible for the failure. Microsoft’s incident reporting attributed the technical problem to its network and Exchange Online systems, not its workforce decisions or AI investments. But the juxtaposition is difficult to ignore. We’re watching one of the largest technology companies in the world invest extraordinary sums in the future of artificial intelligence while simultaneously offering a voluntary workforce exit program, and then experiencing a global outage in the very services millions of people and organizations rely on to do their jobs.
There is an uncomfortable question buried in that sequence of events: what exactly are we optimizing for?
We seem perfectly comfortable treating experienced people as costs to be eliminated, institutional knowledge as an inconvenience, and operational resilience as something that can be addressed after the next quarterly earnings call. The irony is that the systems that make these companies valuable don’t maintain themselves. They don’t magically become reliable because the organization has a more compelling AI strategy, and they don’t become less complicated because someone has decided that a smaller workforce can accomplish more with better tooling.
Technology is built, operated, maintained, and recovered by people. We seem to keep forgetting that.
The FAA and Microsoft incidents aren’t evidence that AI is inherently dangerous, nor are they proof that workforce reductions cause outages. They’re reminders that the systems underpinning modern life are complicated, interconnected, and surprisingly fragile. More than that, they expose a broader problem with the way we think about progress: we’re focused on what we can accelerate, while paying considerably less attention to what we need to preserve.
The missing ingredient: knowing when to stop
I recently read an essay about liminal thinking that has stuck with me. The author made an argument for embracing the spaces between what we know and what comes next: those uncomfortable periods of uncertainty when the old assumptions no longer quite work, but the new answers haven’t fully emerged. The idea resonated with me, particularly because it offers an alternative to the relentless insistence that we must always be moving forward.
We have become extraordinarily uncomfortable with uncertainty. In business, it’s something to eliminate. In technology, it’s something to engineer away. In leadership, it’s often treated as a failure of vision or decisiveness. We want roadmaps, projections, measurable outcomes. We want to know what comes next and how quickly we can get there. And when we don’t know, we often pretend that we do.
But perhaps the most important leadership skill isn’t knowing how to move faster. Perhaps it’s knowing when the circumstances demand that we stop moving long enough to reconsider where we’re going. I don’t mean stopping as in abandoning progress. I mean creating the space to ask questions that are inconvenient, uncomfortable, and frequently unwelcome.
What problem are we actually trying to solve? Who benefits from this decision, and who bears the cost? What assumptions are we making about the future, and what happens if they’re wrong? What are we sacrificing in the pursuit of something we haven’t fully considered? And perhaps most importantly, if we have the ability to do something, what obligation do we have to consider whether doing it is the right thing?
That last question seems to have become almost quaint in some circles. We’ve replaced it with a much simpler one: if we don’t do it, someone else will.
That argument tells me more about the person making it than anything else. It takes a decision that should involve ethics, judgment, and responsibility and turns it into a competitive inevitability. It absolves the decision-maker of responsibility by suggesting that the outcome is predetermined. It’s not a moral argument. It’s an excuse, and an extraordinarily convenient one for people who stand to benefit from the outcome.
The difference between progress and momentum
One of the things I’ve learned over the course of my career is that there is a considerable difference between making progress and generating activity. In infrastructure, we understand this intuitively. A system that is constantly changing isn’t necessarily improving. A deployment pipeline that moves faster isn’t necessarily delivering better software. A platform that can scale isn’t necessarily resilient, and an organization that can produce more output isn’t necessarily creating more value.
Sometimes the most important work is invisible. Maintenance, documentation, redundancy, testing, recovery planning, and investing in the people who understand how things work. Making sure that when something inevitably breaks, the people responsible for fixing it have the knowledge, authority, and resources to do so.
None of those things make very compelling keynote presentations. They don’t generate the same excitement as a new product launch or a revolutionary technology announcement, and they don’t lend themselves to breathless predictions about the future. But they’re the things that make the future possible.
This is where we’ve gotten ourselves into trouble. We’ve built an economic system that disproportionately rewards the visible, the immediate, and the scalable. We’ve created incentives that favor disruption over stewardship, growth over sustainability, and short-term returns over long-term resilience. We’ve taught leaders that their job is to accelerate, optimize, and maximize, while spending comparatively little time teaching them how to preserve, protect, and exercise restraint.
We’ve also created a culture in which saying “I don’t know,” “I’m not convinced,” or “perhaps we should reconsider” is often interpreted as a lack of ambition. I’d argue the opposite. It takes a certain amount of courage to question the direction everyone else is running, and even more to do so when the people running fastest are the ones with the most money, influence, and institutional power.
There is nothing particularly courageous about following the incentives of a system that rewards you for doing so.
We need a different definition of success
I don’t have a problem with ambition. I don’t have a problem with building extraordinary things, making money, or creating businesses that change the world. I do have a problem with treating those outcomes as sufficient evidence of success.
What if we measured success differently? What if we valued organizations that created durable value without destroying the people who built them? What if we admired leaders who understood the limits of their own knowledge, could recognize when their decisions were causing harm, and were willing to change course even when doing so was financially inconvenient? What if we treated the preservation of critical infrastructure, institutional knowledge, human dignity, and public trust as measures of progress rather than obstacles to it? And what if we stopped treating people as interchangeable inputs in an optimization exercise?
Perhaps we should also recognize that some things are worth doing slowly, carefully, and deliberately, even when we have the technological capability to do them faster.
This isn’t an argument for stagnation. It’s an argument for intentionality. There is a place for experimentation, calculated risk, and moving quickly when the circumstances warrant it. But those things need to exist within a framework of responsibility. The fact that we can move faster doesn’t mean we should. The fact that we can automate something doesn’t mean we should eliminate the human judgment surrounding it. The fact that we can build something doesn’t mean we have adequately considered its consequences. And the fact that someone can make an extraordinary amount of money doing something doesn’t mean that thing is good for the rest of us.
I suspect we’ll spend the next several years arguing about artificial intelligence as though the technology itself is the central issue. We’ll debate model capabilities, regulation, copyright, employment, safety, and whether AI will eventually make us all more productive. Those are important conversations, and we should have them. But I think we’re avoiding the harder conversation: who gets to decide what progress looks like? What obligations come with the power to shape it? And why have we become so comfortable allowing economic incentives to answer questions that should require moral judgment?
AI isn’t making those decisions for us. People are. Companies are. Investors are. Governments are. And the rest of us are being told that the outcome is inevitable, that resistance is futile, and that our only sensible option is to adapt.
I disagree.
We are not passengers on some unstoppable technological train. We are the people building the tracks, fueling the engine, writing the checks, setting the regulations, and deciding which destinations are worth pursuing. We have choices, we have agency, and we have an obligation to exercise both.
Perhaps what we need most right now isn’t another breakthrough, another model, another platform, or another declaration that the future has arrived. Perhaps we need a little more willingness to stand at the threshold, acknowledge that we don’t have all the answers, and take the time to figure out what kind of future is actually worth building.
Not everything that can be built deserves to be built, and not everything that can be accelerated should be accelerated. And not every opportunity to make money represents an opportunity to make the world better.
Progress without responsibility isn’t progress. It’s just momentum. And momentum, left unchecked, has a nasty habit of carrying us somewhere we never intended to go.
