AI Made Coding Easy. Vision Is Still the Hard Part.
By Neil Rose, CFA
Notes from the First Regency Hack-a-Thon
Arthur Mallet and I recently spent two and a half days tapping into our inner AI engineers—the first of what will be many rounds of getting deep into systems design, writing code (okay, more like pasting in code that AI fed us), and building things. Useful, productive things enhancing our daily work in small and big ways.
With a whiteboard standing over a long conference table and a venous spread of cords connecting several computers and even more monitors, we hacked away, feeling part The Social Network (without the sociopathic behavior), part Weird Science (without Kelly LeBrock, of course).
More than a week later, what’s front and center in my mind isn’t AI itself or what it’s capable of. It’s that I’m still reconciling an overwhelming barrage of what we are now capable of with a new nervousness about getting right just how to wield it.
Funny how things turn on their head. The scarcity used to be an insurmountable technical barrier. Five years ago, it seemed like every single day someone was warning us our kids would be left behind if they didn’t learn to write computer code. It was plausible enough. Software was eating the world, and computer science majors and tech bros were on their way to becoming the world’s new ruling class.
I bought books with titles like Teach Your Kids to Code and Python: The Complete Course. I learned to clumsily explain the differences between database architectures. I knew enough to correctly say “sea-quill” and “non-sea-quill” instead of reading them as written: “SQL” and “NoSQL.”
Then AI commoditized the one thing that was so good at commoditizing almost everything else: coding. AI took away the need for fluency. Problem solved. Today you describe in detail what you want, and—poof!—out come lines of code and an infinite resource to guide you.
There is still a learning curve. Using a computer’s terminal, cloning Gits and repos, building skill.md files, wiring up connectors—these are all pieces of knowhow you have to pick up. But they’re far lower barriers than coding, and learning them has been surprisingly fast. AI is a splendid teacher.
What hit me hard is that the coding part is just a part—and not the important part—of building automated workflows, apps, and agents (or at least of doing it right). In building actual tools—research agents, alerts, data gatherers and refiners, a dashboard, even a grumpy agent who argues the other side of any idea I have with zeal—I realized the hardest, most important work still lies in vision and systems engineering.
Nothing has changed about the fact that what and why matter more than how. We all know this intuitively. Think about a great restaurant and everything required for it to succeed and last. How produces the great ingredients and parts—a perfectly filleted fish, a complex sauce, perfectly emulsified egg whites—but what and why create the actual dishes, the harmony, the executed vision that delights guests. (And that’s just the food part of the restaurant business.)
No chef would turn down an innovation that offers more and better things to work with. But that doesn’t make the chef’s job irrelevant, or even easier—far from it, because every other kitchen now has the same ingredients to work with.
How a chef’s kitchen stands out is by getting the vision and the plan right, with well-honed systems behind them. Technology can help with execution, but less than I think most of us realize.
AI is still the shiny object. Trillions are being spent acquiring its power and buying more hows. The spending is input-focused: “The more we spend, the more competitive and profitable we’ll be.” In its most recent quarter, Alphabet posted negative free cash flow for the first time since its 2004 IPO—roughly $5.9 billion in the red—as AI infrastructure spending outran even its enormous cash generation. Almost all of the biggest tech companies are pouring money in the same way, and many have taken on huge debt to fund the binge.
The biggest and smartest companies in the world are input-focused. It’s a gold rush. And yet I doubt many of them will ultimately earn a return on the investment, because they’ve forgotten the vision thing—the whats and the whys.
I’m looking forward to our next hack-a-thon, and not just for the building and the practical solutions we’ll move forward. I’m looking forward even more to having gained some wisdom between now and then—clearer whats and whys, and a little less vulnerability to the shiny hows.
About the Author
Neil Rose, CFA, is the founder and CEO of Regency Capital Management.
