The Future is Undefined

What's Changed

As a second time founder, people often ask me what's different this time around. My typical answer revolves around how the second time is easier because you have better instincts, you know who to hire, you can think on longer time horizons, you know what mistakes to avoid, etc.

Startups are a game of minesweeper, and the second time around, you have a better sense of the map.

But there's another aspect of entrepreneurship that is radically different today. Welcome to 2026, where the calculus for how to build has been turned upside-down. Anyone today who claims they have a rulebook for success is not embracing how utterly chaotic and unpredictable the current AI tech landscape is. Their rulebook is, at best, obsolete.

A prime example. Here are the things I knew for sure as recently as two years ago, based on my experience founding and scaling a consumer tech company:

1. Every decision comes down to a cost-benefit analysis - The things worth prioritizing are the ones that have the highest ratio of benefit to cost. It's the only reliable method for determining what's worth doing with limited resources.
2. Cost is synonymous with the resource-hours it would take to build something - Nothing matters more than the seconds in your day, and nothing eats up those seconds more than allocating effort to a particular initiative at the expense of every other possible initiative.
3. Innovation comes from working within constraints - If you want to come up with a creative, new solution to a problem, resource limitations force you to develop innovation solutions. It's one of the only ways in which small companies have an advantage over big ones.

But in 2026, in a world in which coding agents are increasingly supplanting software engineers and anyone can build impactful software, here's what has happened: The cost piece of a cost-benefit analysis has dropped precipitously. In the coming years, it will asymptotically approach zero.A very inaccurate graph representing the time needed to implement complex featuresA very inaccurate graph representing the time needed to implement complex features

In other words, the cost piece of the cost-benefit analysis, on which (most) founders historically based (most) decisions has disappeared or will soon disappear.

What's a cost-benefit analysis without the cost half? A benefit-only analysis?

What's Really Changed

So as an entrepreneur faced with the question of what to prioritize, I'm feeling more and more that my barometer for worthwhileness needs recalibration. So much for the pros of being a second time founder.

One might think a benefit-only analysis is preferential. After all, it's all upside, isn't it?

But day-to-day, that's far from the truth. Let's revisit those three axioms of prioritization that have since lost their axiomatic-ness.

1. Once upon a time, the ratio of benefit to cost determined what was worth building. That was a relative ratio. But without cost, there is no counterforce against which to assess benefit. The analysis ceases to become relative and instead becomes absolute. Prioritization actually becomes harder because of the lack of a tangible yardstick. Plus, when every option suddenly becomes plausible (or at least more plausible), the list of options grows too numerous to manage. Picture a spaceship that suddenly enables travel—without regard for fuel or distance—to the farthest stars in our galaxy. Where do you go first? Our galaxy contains hundreds of billions of stars.
2. To the extent that development cost is still a factor, its definition certainly seems to have changed. Speed of implementation is a fraction of a fraction of what it once was. (I like pondering how long I would have spent a decade ago prototyping things that today take me five minutes. A week? A month?) And, dollars needed for implementation are similarly a fraction of a fraction of what they once were. So if cost of development still exists, it will soon be neither of time nor money. What then will differentiate the winners from the losers?
3. To the extent constraints still exist, they certainly aren’t the ones that existed before. Compute is widely accessible. The most innovative technology now gets commoditized within days. Expertise in any domain is no longer a limiting factor (with LLMs being, in effect, the combined knowledge of humankind; speaking of which, who else is watching

Soon, the only cost will be drive, and the only limiting factor will be imagination.

Tl;dr

To put this all a bit differently, benefit-to-cost used to look like this:

But as the future nears and cost approaches zero, this ratio approaches infinity:

But as you may remember from math class, anything divided by zero is undefined:

Ergo, the future is undefined: