March 2026 was a rocky month for big tech. On March 18, Meta decided to wind down support for Horizon Worlds, the last bastion of its five-year, $70 billion-plus virtual reality bet and the flagship project of its rebrand from Facebook. Eight days later, OpenAI announced it was shuttering Sora, its AI video platform, at a much smaller loss (most estimates put it at a mere $1 billion).
Between the two announcements, Meta and Alphabet were found culpable for teen addiction related to the design of their social media algorithms, verdicts that could open the floodgates for tens of thousands more. By then, localities across the U.S. had halted or delayed over $155 billion in new data center projects due to concerns about skyrocketing water use and electricity prices.
The losses, lawsuits, resistance, and rein-ins reveal something the techno-optimist set keeps getting wrong. While technical breakthroughs open up new possibilities, breakthroughs alone—even well-resourced and/or well-intentioned ones—aren’t enough to create meaningful, lasting progress. True progress includes the decision about what to keep and what to refuse. True progress is built on constraints.
The word constraints conjures up images of limits: budgets, timelines, and resources. Those realities count, but the constraints that make things worthwhile are the ones we choose—the lines we draw about what to refuse and what to insist on.
The question isn’t whether or not we should embrace constraints. The real question is when. And when makes all the difference. If we draw the line early, breakthroughs can create value for people. When we draw it late, they become a cost for society. And that’s not progress.
Nearly every beloved product is beloved because its designers committed to constraints early on—Apple committed to craft and simplicity, Costco to living wages and $1.50 hot dogs. Nearly every unfolding crisis, from the smog-filled LA basin in the 1970s, to body dysmorphia and political unrest in the 2020s, came from ignoring constraints until the breakthrough turns into an issue.
Today we’re in the early stages of a new breakthrough with artificial intelligence, and it’s the right moment to take a good, hard look at what counts as progress. In short, progress should be judged by the fruit it produces, not simply by the pace of productivity. Breakthroughs naturally outpace our ability to make good use of them—that’s what makes them breakthroughs. But breakthroughs without constraints leave us sifting through a smoldering pile of what we can do (good and bad) instead of creating a world built upon what’s actually worth doing. Constraints are not antiprogress; they are key to it.
Design is a field well acquainted with the value of constraints. The proliferation of the printing press in the 1400s spawned typography standards in the 1500s. Bauhaus principles of the early 1900s aimed to tame the industrial chaos of the late 1800s by reviving a sense of craft and humanity. The 1990s internet sprawled until UI/UX design brought order in the 2000s. Breakthroughs expand possibilities. Constraints turn those possibilities into the things we actually value.
We’re in the same cycle now, with two distinct differences. First, the pace is picking up. So much so that even the people inventing and shaping AI aren’t able to plan more than three months ahead. (I’ve spoken to leaders at large language model companies who say they are on two-week timelines and others at software firms who count their cycles in days.) Possibilities—for better and worse—are expanding like mad.
Second, with each breakthrough, we transcend our prior constraints. Old buildings look lovely in part because building techniques and materials were constrained to natural elements like brick and wood and stone. Architects of today have much more freedom, but the ones who create timeless structures understand they have to make highly restrained choices. Not only do we need constraints around emerging tech, we have to get better at choosing them. Progress by design is saying yes to some things and no to others.
As technological breakthroughs continue apace, our ability to design worthwhile constraints has to catch up. It’s not hard to imagine; commitment is the issue. It starts with making commonsense choices that address effects, agnostic of tech—like say, don’t break things you can’t fix, sell things you wouldn’t give your own family, addict kids, or profit from despair. From there, it’s about developing tactics to anticipate side effects—like the confusion that AI-generated content and artificial personalities are likely to produce—and design constraints before they arrive.
It’s helpful to look to the past for guidance and nuance. Like today, past designers had different relationships with breakthrough technologies of their era. Some, like Charles and Ray Eames, embraced emerging technologies and used them to make better things. In their words, to create “the best [products], for the most [people], for the least [cost].” Others, like Victor Papanek, rejected most technology in favor of simple methods to empower DIY counterculture.
For those who choose to buy into the current round of breakthroughs, and want to improve on them, there is plenty of tuning to do to make these technologies more useful. Now is the time to design better UI for AI and—god forbid—help people collaborate rather than compete. Today’s ubiquitous, single-player chat window is like the blue links of the early internet: convenient, but nothing close to the potential.
A skeptical approach brings attention to undesirable side effects that seem to be showing up almost instantly alongside AI breakthroughs, like AI-induced psychosis in people with no prior indications or the growing concern of ”human enfeeblement.” Others are trying to reject the tech altogether, as in-person, digital-free experience design is already undergoing a renaissance and techniques for AI abstinence are also on the rise. Regardless of the approach (buy-in, skepticism, or rejection), each can create value via constraints.
None of this is to say the build-and-breakthrough ethos is entirely wrong. It’s just incomplete. Allow me an allegory. In Raiders of the Lost Ark, Indiana Jones and his rival Belloq are racing to find the Ark of the Covenant. The key is a medallion that, mounted atop a staff of just the right height, refracts sunlight onto a map to reveal the Ark’s location.
Both men have a medallion, but Belloq’s is a copy, with instructions only on the front. Indy’s original has an inscription on the backside too.
Belloq builds his staff according to the front-side instructions which say the staff should be 6 kadams tall (about 72 inches). He shines the light, marks the spot, and starts digging. Meanwhile, Indy gets the backside translated. It includes another instruction, “and take away one kadam to honor the God whose Ark this is.” The proper height is 5 kadams, not 6. With only half the instructions, Belloq built the staff too tall. He’s digging in the wrong place!
With their advocacy of breakthroughs as the one path to progress, hardcore techno-optimists are digging in the wrong place. They confidently—albeit naively, and often willingly—follow half of the instructions as if they have the whole picture. They believe technological breakthroughs are all that’s needed for progress. Actually, breakthroughs are necessary but not sufficient for progress. Real progress requires raising the bar and trimming it down.
Constraints don’t need to be aimed at the tech itself (let development continue apace). They should aim at the effects (define what we don’t want now so if and when it happens, we have a ready way to rein it in). But to do so, we need to stop pretending technologists own the path to progress. Though breakthroughs may lay the groundwork, even the people helming the companies that are building foundational AI models readily admit they have no idea what they are dealing with.
The job of inventors is to create new realms of possibility. We can love them for that. But we can’t leave progress in the hands of the technically savvy alone. Technical capability is not wisdom, engineering is not understanding, feasibility doesn’t guarantee desirability—and possibility alone is not progress. Most importantly, thoughtful constraints are not antiprogress; they are provalue.
Build. Abundance. Sure. But, embrace constraints as a key part of the path to progress. Well-designed constraints don’t impede progress—real progress depends on them.
