The Outcome Era: Manifesting Software to move electrons in the final Layer
SaaS was just a workaround for expensive engineers
Once we developed silicon capable of computing bits, we found a way to write custom scripts to make these machines do our custom work. This impacted every industry horizontally. Entertainment grabbed attention, which opened doors for marketers, better systems emerged to record transactions, store data, and calculate. This was the birth of the software era. The hurdle was to create the scripts. To create digital products, you needed brilliant people with a high capacity for abstraction, people who could write efficient code. This was a scarce resource, requiring years of learning and experimentation. Consequently, software became the most lucrative field of the last decade. "Lucrative" because software is massive leverage, a simple script running 24/7 creates a lot of economic value, and the people writing it are paid accordingly. Because these human resources were so scarce, the best way to build a business was to create one specific tool and tweak it slightly for a larger audience. We called this "Software as a Service". It was more efficient to capture a market vertically than to build bespoke solutions for every individual. Even though code is ultimately meant to "move electrons" per your wish, the human bottleneck forced us into one-size-fits-all products. Then mathematical models capable of predicting the next token in natural language began to "understand" the world through diverse data, self-attention architectures, and massive computation. Whether these models truly "understand" the human world or are simply mimicking it is beside the point. In human history, only a few people have made truly creative contributions, meaning something different from their own "training data." Most of us, while capable of uniqueness, are bound by mimicry. We live in echo chambers, we are the average of the five people we spend time with. Even social media algorithms have figured us out, displaying content just to keep us hooked. This means most tasks can now be automated. We don’t need to invent new physics to understand this revolution. Arguments about needing a change in architecture to achieve AGI and sentience are a separate matter. In fact, if we achieve a sentient AGI with its own identity, we couldn't just "shut it down" without raising moral issues. Let’s not go there. The point is: we have solved the bottleneck. We no longer strictly need humans to write the code that moves those electrons. Critics argue that AI only makes "toy scripts" that can't scale. But remember: enterprise software was born because scaling one workflow was the only way to justify the human cost. If there is no longer a scarce resource bottleneck, a company (or an individual) can simply write their own custom scripts in-house. Instead of having the power to make custom functions in Excel, you now have the power to build Excel itself using natural language. Most engineering hardship comes from scaling a script for millions, but do you really need to scale if you are just using it for yourself or a small team within your org? We can now move electrons however we want. We don't need to know how to move them, we only need to know what the outcome should be, how to evaluate if that outcome has been achieved, and how to account for edge cases. I may be wrong, but this suggests there is no longer a need to study software engineering in extreme depth. We should understand how things work at a high level, but the deep technical knowledge is becoming secondary. The valuable skill is now the ability to manifest an idea into reality efficiently. Build for yourself, and build for others, at least until the loop closes. Once AI can figure out "what to build" and evaluate the outcomes on its own, I’m not sure what happens next.