What Makes a Good Coder in a World Where Anyone Can Code?

Notes on programmas, LLMs, and the future of computer science

In a fireside chat with the editor of Fortune, maverick investor Cathie Wood made a bold proclamation regarding the future of work. Wood predicted that in the coming years, “We’re all going to become programmers.”

If Wood’s predictions prove correct and large language models (LLMs) continue to mature, computer programming will evolve from code-based to prompt-based. As a consequence, the barrier to entry to coding will be lowered dramatically and the boundary between technical and non-technical knowledge workers may soon blur, perhaps even to the point of dissolution. 

So, what will constitute programming ability in a future when anyone and everyone can code? 

Researchers at LinkedIn recently produced their own Future of Work Report in which they found that 96% of software engineering skills were “potentially augmentable” by generative artificial intelligence (GAI) – by far the highest number of any occupation in the study. The question is whether “potentially augmentable” is in this case a euphemism for “fully replaceable.” And if so, what does the precious 4% remainder consist of? 

It’s worth noting here that no-code platforms have been around for a good while now and have yet to imperil the livelihoods of software engineers (SWEs). There is, after all, much more to engineering acumen than mere fluency in programming languages. The best SWEs are highly analytical, “first principles” thinkers who know how to boil problems down to their rudiments and schematize elegant solutions leveraging technological advances. While GAI may take over much of the legwork on the solutions end of this spectrum (e.g. templating, debugging, or learning new languages), the ability to identify problems in the first place is more resistant to automation. Indeed, if there is one skill that separates exceptional founders from the pack, it is surely this capacity for problem identification (call it problematizing). 

While some problems are glaringly obvious to pretty much everyone, others fly under the radar, hide in plain sight, or masquerade as unsolvable. These as-yet-unidentified problems are the embryos of billion-dollar businesses and category-creating innovations. If Wood’s intuitions are valid, the great founders of tomorrow will not be great programmers so much as great queryists. 

So what does querying aptitude entail? Where will future founders identify the most intractable problems of their era? And what are the attributes that differentiate tireless problematizers from passive inhabitants of the status quo? 

First principles thinking will likely always confer an edge. For proof, just consider the results of a recent Endeavor report on the career journeys of 200 unicorn founders. Unsurprisingly, researchers found significant convergence around attributes such as entrepreneurial teeth-cutting, international work experience, and STEM degrees, indicating that unicorn founders are seasoned, worldly, analytical thinkers, adept at approaching problems with fresh eyes. Yet the clearest takeaway from the report is the sheer variety of paths to unicorn status. There is no tried-and-true formula, only a series of gnomic wayposts. 

In our experience, the attributes that differentiate exceptional founders from merely good ones are ambient, resistant to analysis, and hard to pin down. They are, in a word, soft skills, however hard-won. They are skills developed through first-hand experience; data that takes the form not of numbers in a spreadsheet, but of implicit insights and glimpsed connections. These skills resist formulas and formalizations.

In a future where everyone can code, it’s those skills that will tilt the scales. To nurture them, our educational landscape could be due for a course correction. Since the mid-2000s, the percentage of undergraduates enrolled in STEM degrees has surged, rising from 22 percent in 2006 to 30 percent in 2015 – the highest peak since researchers began tracking national enrollment data in 1987 – while enrollment in the humanities has plummeted. No STEM field has surged as dramatically as computer science, and CS intro classes have grown so overcrowded that some universities have made concerted efforts to weed out new entrants.

CS degrees are widely considered safe bets as far as job prospects and earning potential are concerned, and LLMs are unlikely to change that in the near future. At the very least, the analytical mindsets fostered by STEM classes are sure to retain their value. Yet the skills required to sift signals from noise and make meaning of our increasingly complex world need not stem from STEM (or any other academic domain, for that matter). They are just as likely to come from serendipitous insights while solving real-world problems on the ground.

Once AI can both articulate problems and solve them for us, we will presumably have reached the point where Wood’s prophecy rings true. Yet that point still seems very remote on the horizon. Curiosity is an attribute that is irrevocably rooted in lived experience. It feeds off of patience, slowness, and idleness, characteristics that are seldom found in the lightning-fast realm of ones and zeroes.

The programmers of the future may not possess anything resembling what we currently think of as coding skills. But we can bank on them being inquisitive types who approach familiar problems with defamiliarized eyes. Hence why we at Armory remain as excited as ever about outsiders, immigrants, and denizens of overlooked markets. Innovation can come from anywhere and everywhere, and in our view, a grandma unicorn founder isn’t far off.