Back in the 1950s, a paper was published suggesting that there was an upper limit to the amount of oil that could be produced globally, which was expected to be reached in the early 2000s. After that point, it would all be used up, and oil as a resource would slowly and surely be exhausted, and with it, a society-collapsing apocalypse.
Of course, none of it came true. The idea of a production "peak" is a useful device for thinking about efficiency. The reality is that there are plateaus rather than a single peak. And I think we've jumped to the next one right now, and "is this the last one?" is the remaining question. Are we entering peak software?
The Last Plateau: Low-Code/No-Code
Over the last two decades, Low-Code and No-Code solutions have advanced significantly. Today, platforms are the names that dominate enterprise software: Microsoft, ServiceNow, Salesforce and there's a whole second-tier ecosystem of competing products right below them. These software abstraction layers, which use graphical elements and user interfaces to accelerate the development of software, have been around for a long time. And for a long time, they were underpowered, poorly designed, and typically created a lot more problems than they solved. As computing caught up, and the software matured, it got better - a lot better.
Since the beginning of the 2010s, Low-Code/No-Code really hit its stride. Organizations empowered teams to digitize their own work, even giving it various labels. It was a democratization of capability, although with price tags that strained budgets and made technology leaders look like they were in an episode of "Hot Ones" and had just bit into a wing with ghost reaper dust.
For the organizations that could lean into the technology with a mindfulness around process, it unlocked a lot of value. A golden time of software! It couldn't get any better than this. And throughout those years, I thought, "Surely we are at peak software, right now. It can't be any easier to deliver working software that provides value."
AI: the Next Plateau
It isn't hyperbolic to say that since the release of modern AI into the wild, led by ChatGPT's release, a whole new plateau has emerged, and we're rapidly climbing through the clouds to the next level. At first, it was AI playing at parlor tricks; here's a bash script to help you out. Here's a function to do a bubble sort. Here's a python version of snake. Within months, all-in companies were showing off truly crazy prototypes: entire software development and systems development lifecycle within a GUI. Every day, something new was rocking the software development community. To the point where luminaries like Gene Kim and Steve Yegge themselves spent as much time telling stories as sharing practice in their book Vibe Coding.
And now, in 2026, here we are with agentic models and software creating solutions that require solid engineering practice and discipline, and can compress software creation into minutes. Software that used to take months, even years, to write. The 10x engineer has finally met their match: the 100x agentic tooling. The ironic part is that none of these new tools has suddenly eliminated the need for the engineering that underpins them, the science that drives the interaction between electrons and logic gates, or the applied and practical approaches. Disciplinary mastery is more important than ever.
The migration to the next plateau is already happening. The DORA report, a survey of developers on practices and trends, shows that the software community is learning how to use these tools to increase quality, volume, velocity, and productivity. The new learning is how to ensure that, as a community, we are fixed on also providing value. When generating lines of instruction for a computer is commoditized, the importance of understanding their inherent value in solving real problems becomes the coin of the realm.
Is There Still Another Plateau?
Ironically, as much as I look at where we are today and think, "This. This must be the peak!" there may be yet another just a bit off in the distance. Quantum, while very early yet, threatens to upend how we perceive computing altogether. Instead, I've started to think and call our current path the "silicon plateau". While we may still see efficiencies and improvements, this transition into AI likely represents the final optimal configuration that silicon-based compute can offer. Maybe there are intermediary transitions just around the corner, pre-quantum photonic systems and the like, but for now, the ground beneath our feet will continue to look like a polished blue mirror.
For me, right now, I haven't had this much fun solving problems since I was a young adult. You see, in the early days of computing, and personal computing, software was simpler, and a single indivdual could solve problems without devoting all their time to the pursuit of it. We should embrace and revel in the idea that we've returned to software, albeit with all the complexity we've created over the past decades, that is built simply, that can be delivered by small teams, and realizes immediate value.
And wasn't that always the point?